zephyr/scripts/dts/testdtlib.py

2249 lines
44 KiB
Python
Raw Normal View History

dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#!/usr/bin/env python3
# Copyright (c) 2019, Nordic Semiconductor
# SPDX-License-Identifier: BSD-3-Clause
import os
import shutil
import sys
import dtlib
# Test suite for dtlib.py. Run it directly as an executable, in this directory:
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# $ ./testdtlib.py
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
# TODO: Factor out common code from error tests
def run():
"""
Runs all dtlib tests. Immediately exits with status 1 and a message on
stderr on test suite failures.
"""
# Note: All code is within this function, even though some triple-quoted
# strings are unindented to make things less awkward
def fail(msg):
sys.exit("test failed: {}".format(msg))
def parse(dts, include_path=()):
open(".tmp.dts", "w").write(dts)
return dtlib.DT(".tmp.dts", include_path)
def verify_parse(dts, expected, include_path=()):
# The [1:] is so that the first line can be put on a separate line
# after """
dt = parse(dts[1:], include_path)
actual = str(dt)
expected = expected[1:-1]
if actual != expected:
fail("expected '{}' to parse as '{}', parsed as '{}'"
.format(dts, expected, actual))
return dt
def verify_error(dts, msg):
prefix = "expected '{}' to generate the error '{}', generated" \
.format(dts, msg)
try:
parse(dts[1:])
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
# These might already exist from failed earlier runs
try:
os.mkdir(".tmp")
except OSError:
pass
try:
os.mkdir(".tmp2")
except OSError:
pass
#
# Test cell parsing
#
verify_parse("""
/dts-v1/;
/ {
a;
b = < >;
c = [ ];
d = < 10 20 >;
e = < 0U 1L 2UL 3LL 4ULL >;
f = < 0x10 0x20 >;
g = < 010 020 >;
h = /bits/ 8 < 0x10 0x20 (-1) >;
i = /bits/ 16 < 0x10 0x20 (-1) >;
j = /bits/ 32 < 0x10 0x20 (-1) >;
k = /bits/ 64 < 0x10 0x20 (-1) >;
l = < 'a' 'b' 'c' >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""",
"""
/dts-v1/;
/ {
a;
b;
c;
d = < 0xa 0x14 >;
e = < 0x0 0x1 0x2 0x3 0x4 >;
f = < 0x10 0x20 >;
g = < 0x8 0x10 >;
h = [ 10 20 FF ];
i = /bits/ 16 < 0x10 0x20 0xffff >;
j = < 0x10 0x20 0xffffffff >;
k = /bits/ 64 < 0x10 0x20 0xffffffffffffffff >;
l = < 0x61 0x62 0x63 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""")
verify_error("""
/dts-v1/;
/ {
a = /bits/ 16 < 0x10000 >;
};
""",
".tmp.dts:4 (column 18): parse error: 65536 does not fit in 16 bits")
verify_error("""
/dts-v1/;
/ {
a = < 0x100000000 >;
};
""",
".tmp.dts:4 (column 8): parse error: 4294967296 does not fit in 32 bits")
verify_error("""
/dts-v1/;
/ {
a = /bits/ 128 < 0 >;
};
""",
".tmp.dts:4 (column 13): parse error: expected 8, 16, 32, or 64")
#
# Test bytes parsing
#
verify_parse("""
/dts-v1/;
/ {
a = [ ];
b = [ 12 34 ];
c = [ 1234 ];
};
""",
"""
/dts-v1/;
/ {
a;
b = [ 12 34 ];
c = [ 12 34 ];
};
""")
verify_error("""
/dts-v1/;
/ {
a = [ 123 ];
};
""",
".tmp.dts:4 (column 10): parse error: expected two-digit byte or ']'")
#
# Test string parsing
#
verify_parse(r"""
/dts-v1/;
/ {
a = "";
b = "ABC";
c = "\\\"\xab\377\a\b\t\n\v\f\r";
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""",
r"""
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
/dts-v1/;
/ {
a = "";
b = "ABC";
c = "\\\"\xab\xff\a\b\t\n\v\f\r";
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""")
verify_error(r"""
/dts-v1/;
/ {
a = "\400";
};
""",
".tmp.dts:4 (column 6): parse error: octal escape out of range (> 255)")
#
# Test character literal parsing
#
verify_parse(r"""
/dts-v1/;
/ {
a = < '\'' >;
b = < '\x12' >;
};
""",
"""
/dts-v1/;
/ {
a = < 0x27 >;
b = < 0x12 >;
};
""")
verify_error("""
/dts-v1/;
/ {
// Character literals are not allowed at the top level
a = 'x';
};
""",
".tmp.dts:5 (column 6): parse error: malformed value")
verify_error("""
/dts-v1/;
/ {
a = < '' >;
};
""",
".tmp.dts:4 (column 7): parse error: character literals must be length 1")
verify_error("""
/dts-v1/;
/ {
a = < '12' >;
};
""",
".tmp.dts:4 (column 7): parse error: character literals must be length 1")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# Test /incbin/
#
open(".tmp.bin", "wb").write(b"\00\01\02\03")
verify_parse("""
/dts-v1/;
/ {
a = /incbin/ (".tmp.bin");
b = /incbin/ (".tmp.bin", 1, 1);
c = /incbin/ (".tmp.bin", 1, 2);
};
""",
"""
/dts-v1/;
/ {
a = [ 00 01 02 03 ];
b = [ 01 ];
c = [ 01 02 ];
};
""")
open(".tmp/in_subdir", "wb").write(b"\00\01\02")
verify_parse("""
/dts-v1/;
/ {
a = /incbin/ ("in_subdir");
};
""",
"""
/dts-v1/;
/ {
a = [ 00 01 02 ];
};
""",
include_path=(".tmp",))
verify_error(r"""
/dts-v1/;
/ {
a = /incbin/ ("missing");
};
""",
".tmp.dts:4 (column 25): parse error: 'missing' could not be found")
#
# Test node merging
#
verify_parse("""
/dts-v1/;
/ {
l1: l2: l1: foo {
foo1 = [ 01 ];
l4: l5: bar {
bar1 = [ 01 ];
};
};
};
l3: &l1 {
foo2 = [ 02 ];
l6: l7: bar {
bar2 = [ 02 ];
};
};
&l3 {
foo3 = [ 03 ];
};
&{/foo} {
foo4 = [ 04 ];
};
&{/foo/bar} {
bar3 = [ 03 ];
l8: baz {};
};
/ {
};
/ {
top = [ 01 ];
};
""",
"""
/dts-v1/;
/ {
top = [ 01 ];
l1: l2: l3: foo {
foo1 = [ 01 ];
foo2 = [ 02 ];
foo3 = [ 03 ];
foo4 = [ 04 ];
l4: l5: l6: l7: bar {
bar1 = [ 01 ];
bar2 = [ 02 ];
bar3 = [ 03 ];
l8: baz {
};
};
};
};
""")
verify_error("""
/dts-v1/;
/ {
};
&missing {
};
""",
".tmp.dts:6 (column 1): parse error: undefined node label 'missing'")
verify_error("""
/dts-v1/;
/ {
};
&{foo} {
};
""",
".tmp.dts:6 (column 1): parse error: node path 'foo' does not start with '/'")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_error("""
/dts-v1/;
/ {
};
&{/foo} {
};
""",
".tmp.dts:6 (column 1): parse error: component 'foo' in path '/foo' does not exist")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# Test property labels
#
def verify_label2prop(label, expected):
actual = dt.label2prop[label].name
if actual != expected:
fail("expected label '{}' to map to prop '{}', mapped to prop '{}'"
.format(label, expected, actual))
dt = verify_parse("""
/dts-v1/;
/ {
a;
b;
l2: c;
l4: l5: l5: l4: d = < 0 >;
};
/ {
l1: b;
l3: c;
l6: d;
};
""",
"""
/dts-v1/;
/ {
a;
l1: b;
l2: l3: c;
l4: l5: l6: d = < 0x0 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""")
verify_label2prop("l1", "b")
verify_label2prop("l2", "c")
verify_label2prop("l3", "c")
verify_label2prop("l4", "d")
verify_label2prop("l5", "d")
verify_label2prop("l6", "d")
#
# Test offset labels
#
def verify_label2offset(label, expected_prop, expected_offset):
actual_prop, actual_offset = dt.label2prop_offset[label]
actual_prop = actual_prop.name
if (actual_prop, actual_offset) != (expected_prop, expected_offset):
fail("expected label '{}' to map to offset {} on prop '{}', "
"mapped to offset {} on prop '{}'"
.format(label, expected_offset, expected_prop, actual_offset,
actual_prop))
dt = verify_parse("""
/dts-v1/;
/ {
a = l01: l02: < l03: &node l04: l05: 2 l06: >,
l07: l08: [ l09: 03 l10: l11: 04 l12: l13: ] l14:, "A";
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
b = < 0 > l23: l24:;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
node: node {
};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""",
"""
/dts-v1/;
/ {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
a = l01: l02: < l03: &node l04: l05: 0x2 l06: l07: l08: >, [ l09: 03 l10: l11: 04 l12: l13: l14: ], "A";
b = < 0x0 l23: l24: >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
node: node {
phandle = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""")
verify_label2offset("l01", "a", 0)
verify_label2offset("l02", "a", 0)
verify_label2offset("l04", "a", 4)
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_label2offset("l05", "a", 4)
verify_label2offset("l06", "a", 8)
verify_label2offset("l09", "a", 8)
verify_label2offset("l10", "a", 9)
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_label2offset("l23", "b", 4)
verify_label2offset("l24", "b", 4)
#
# Test unit_addr
#
def verify_unit_addr(path, expected):
node = dt.get_node(path)
if node.unit_addr != expected:
fail("expected {!r} to have unit_addr '{}', was '{}'"
.format(node, expected, node.unit_addr))
dt = verify_parse("""
/dts-v1/;
/ {
no-unit-addr {
};
unit-addr@ABC {
};
unit-addr-non-numeric@foo-bar {
};
};
""",
"""
/dts-v1/;
/ {
no-unit-addr {
};
unit-addr@ABC {
};
unit-addr-non-numeric@foo-bar {
};
};
""")
verify_unit_addr("/no-unit-addr", "")
verify_unit_addr("/unit-addr@ABC", "ABC")
verify_unit_addr("/unit-addr-non-numeric@foo-bar", "foo-bar")
#
# Test node path references
#
verify_parse("""
/dts-v1/;
/ {
a = &label;
b = [ 01 ], &label;
c = [ 01 ], &label, <2>;
d = &{/abc};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
label: abc {
e = &label;
f = &{/abc};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""",
"""
/dts-v1/;
/ {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
a = &label;
b = [ 01 ], &label;
c = [ 01 ], &label, < 0x2 >;
d = &{/abc};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
label: abc {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
e = &label;
f = &{/abc};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""")
verify_error("""
/dts-v1/;
/ {
sub {
x = &missing;
};
};
""",
"/sub: undefined node label 'missing'")
verify_error("""
/dts-v1/;
/ {
sub {
x = &{/sub/missing};
};
};
""",
"/sub: component 'missing' in path '/sub/missing' does not exist")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# Test phandles
#
verify_parse("""
/dts-v1/;
/ {
x = < &a &{/b} &c >;
dummy1 {
phandle = < 1 >;
};
dummy2 {
phandle = < 3 >;
};
a: a {
};
b {
};
c: c {
phandle = < 0xFF >;
};
};
""",
"""
/dts-v1/;
/ {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
x = < &a &{/b} &c >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dummy1 {
phandle = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
dummy2 {
phandle = < 0x3 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
a: a {
phandle = < 0x2 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
b {
phandle = < 0x4 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
c: c {
phandle = < 0xff >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""")
# Check that a node can be assigned a phandle to itself. This just forces a
# phandle to be allocated on it. The C tools support this too.
verify_parse("""
/dts-v1/;
/ {
dummy {
phandle = < 1 >;
};
a {
foo: phandle = < &{/a} >;
};
label: b {
bar: phandle = < &label >;
};
};
""",
"""
/dts-v1/;
/ {
dummy {
phandle = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
a {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
foo: phandle = < &{/a} >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
label: b {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
bar: phandle = < &label >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""")
verify_error("""
/dts-v1/;
/ {
sub {
x = < &missing >;
};
};
""",
"/sub: undefined node label 'missing'")
verify_error("""
/dts-v1/;
/ {
a: sub {
x = /bits/ 16 < &a >;
};
};
""",
".tmp.dts:5 (column 19): parse error: phandle references are only allowed in arrays with 32-bit elements")
verify_error("""
/dts-v1/;
/ {
foo {
phandle = [ 00 ];
};
};
""",
"/foo: bad phandle length (1), expected 4 bytes")
verify_error("""
/dts-v1/;
/ {
foo {
phandle = < 0 >;
};
};
""",
"/foo: bad value 0x00000000 for phandle")
verify_error("""
/dts-v1/;
/ {
foo {
phandle = < (-1) >;
};
};
""",
"/foo: bad value 0xffffffff for phandle")
verify_error("""
/dts-v1/;
/ {
foo {
phandle = < 17 >;
};
bar {
phandle = < 17 >;
};
};
""",
"/bar: duplicated phandle 0x11 (seen before at /foo)")
verify_error("""
/dts-v1/;
/ {
foo {
phandle = < &{/bar} >;
};
bar {
};
};
""",
"/foo: phandle refers to another node")
# Test phandle2node
def verify_phandle2node(prop, offset, expected_name):
phandle = dtlib.to_num(dt.root.props[prop].value[offset:offset + 4])
actual_name = dt.phandle2node[phandle].name
if actual_name != expected_name:
fail("expected {} to be a phandle for {}, was a phandle for {}"
.format(prop, expected_name, actual_name))
dt = parse("""
/dts-v1/;
/ {
phandle_ = < &{/node1} 0 1 >;
phandles = < 0 &{/node2} 1 &{/node3} >;
node1 {
phandle = < 123 >;
};
node2 {
};
node3 {
};
};
""")
verify_phandle2node("phandle_", 0, "node1")
verify_phandle2node("phandles", 4, "node2")
verify_phandle2node("phandles", 12, "node3")
#
# Test mixed value type assignments
#
verify_parse("""
/dts-v1/;
/ {
x = /bits/ 8 < 0xFF 0xFF >,
&abc,
< 0xFF &abc 0xFF &abc >,
&abc,
[ FF FF ],
"abc";
abc: abc {
};
};
""",
"""
/dts-v1/;
/ {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
x = [ FF FF ], &abc, < 0xff &abc 0xff &abc >, &abc, [ FF FF ], "abc";
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
abc: abc {
phandle = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""")
#
# Test property deletion
#
verify_parse("""
/dts-v1/;
/ {
keep = < 1 >;
delete = < &sub >, &sub;
/delete-property/ missing;
/delete-property/ delete;
sub: sub {
y = < &sub >, &sub;
};
};
&sub {
/delete-property/ y;
};
""",
"""
/dts-v1/;
/ {
keep = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
sub: sub {
};
};
""")
#
# Test node deletion
#
verify_parse("""
/dts-v1/;
/ {
sub1 {
x = < 1 >;
sub2 {
x = < &sub >, &sub;
};
/delete-node/ sub2;
};
sub3: sub3 {
x = < &sub >, &sub;
};
sub4 {
x = < &sub >, &sub;
};
};
/delete-node/ &sub3;
/delete-node/ &{/sub4};
""",
"""
/dts-v1/;
/ {
sub1 {
x = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""")
verify_error("""
/dts-v1/;
/ {
};
/delete-node/ &missing;
""",
".tmp.dts:6 (column 15): parse error: undefined node label 'missing'")
verify_error("""
/dts-v1/;
/delete-node/ {
""",
".tmp.dts:3 (column 15): parse error: expected label (&foo) or path (&{/foo/bar}) reference")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# Test /include/ (which is handled in the lexer)
#
# Verify that /include/ searches the current directory
open(".tmp/same-dir-1", "w").write("""
x = [ 00 ];
/include/ "same-dir-2"
""")
open(".tmp/same-dir-2", "w").write("""
y = [ 01 ];
/include/ "same-dir-3"
""")
open(".tmp/same-dir-3", "w").write("""
z = [ 02 ];
""")
verify_parse("""
/dts-v1/;
/ {
/include/ ".tmp/same-dir-1"
};
""",
"""
/dts-v1/;
/ {
x = [ 00 ];
y = [ 01 ];
z = [ 02 ];
};
""")
# Test tricky includes and include paths
open(".tmp2.dts", "w").write("""
/dts-v1/;
/ {
""")
open(".tmp3.dts", "w").write("""
x = <1>;
""")
open(".tmp/via-include-path-1", "w").write("""
= /include/ "via-include-path-2"
""")
open(".tmp2/via-include-path-2", "w").write("""
<2>;
};
""")
verify_parse("""
/include/ ".tmp2.dts"
/include/ ".tmp3.dts"
y /include/ "via-include-path-1"
""",
"""
/dts-v1/;
/ {
x = < 0x1 >;
y = < 0x2 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""",
include_path=(".tmp", ".tmp2",))
verify_error("""
/include/ "missing"
""",
".tmp.dts:1 (column 1): parse error: 'missing' could not be found")
# Verify that an error in an included file points to the right location
open(".tmp2.dts", "w").write("""\
x
""")
verify_error("""
/include/ ".tmp2.dts"
""",
".tmp2.dts:3 (column 3): parse error: expected '/dts-v1/;' at start of file")
# Test recursive /include/ detection
open(".tmp2.dts", "w").write("""\
/include/ ".tmp3.dts"
""")
open(".tmp3.dts", "w").write("""\
/include/ ".tmp.dts"
""")
verify_error("""
/include/ ".tmp2.dts"
""",
"""\
.tmp3.dts:1 (column 1): parse error: recursive /include/:
.tmp.dts:1 ->
.tmp2.dts:1 ->
.tmp3.dts:1 ->
.tmp.dts\
""")
verify_error("""
/include/ ".tmp.dts"
""",
"""\
.tmp.dts:1 (column 1): parse error: recursive /include/:
.tmp.dts:1 ->
.tmp.dts\
""")
#
# Test /omit-if-no-ref/
#
verify_parse("""
/dts-v1/;
/ {
x = < &{/referenced} >, &referenced2;
/omit-if-no-ref/ referenced {
};
referenced2: referenced2 {
};
/omit-if-no-ref/ unreferenced {
};
l1: /omit-if-no-ref/ unreferenced2 {
};
/omit-if-no-ref/ l2: unreferenced3 {
};
unreferenced4: unreferenced4 {
};
unreferenced5 {
};
};
/omit-if-no-ref/ &referenced2;
/omit-if-no-ref/ &unreferenced4;
/omit-if-no-ref/ &{/unreferenced5};
""",
"""
/dts-v1/;
/ {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
x = < &{/referenced} >, &referenced2;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
referenced {
phandle = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
referenced2: referenced2 {
};
};
""")
verify_error("""
/dts-v1/;
/ {
/omit-if-no-ref/ x = "";
};
""",
".tmp.dts:4 (column 21): parse error: /omit-if-no-ref/ can only be used on nodes")
verify_error("""
/dts-v1/;
/ {
/omit-if-no-ref/ x;
};
""",
".tmp.dts:4 (column 20): parse error: /omit-if-no-ref/ can only be used on nodes")
verify_error("""
/dts-v1/;
/ {
/omit-if-no-ref/ {
};
};
""",
".tmp.dts:4 (column 19): parse error: expected node or property name")
verify_error("""
/dts-v1/;
/ {
/omit-if-no-ref/ = < 0 >;
};
""",
".tmp.dts:4 (column 19): parse error: expected node or property name")
verify_error("""
/dts-v1/;
/ {
};
/omit-if-no-ref/ &missing;
""",
".tmp.dts:6 (column 18): parse error: undefined node label 'missing'")
verify_error("""
/dts-v1/;
/omit-if-no-ref/ {
""",
".tmp.dts:3 (column 18): parse error: expected label (&foo) or path (&{/foo/bar}) reference")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# Test expressions
#
verify_parse("""
/dts-v1/;
/ {
ter1 = < (0 ? 1 : 0 ? 2 : 3) >;
ter2 = < (0 ? 1 : 1 ? 2 : 3) >;
ter3 = < (1 ? 1 : 0 ? 2 : 3) >;
ter4 = < (1 ? 1 : 1 ? 2 : 3) >;
or1 = < (0 || 0) >;
or2 = < (0 || 1) >;
or3 = < (1 || 0) >;
or4 = < (1 || 1) >;
and1 = < (0 && 0) >;
and2 = < (0 && 1) >;
and3 = < (1 && 0) >;
and4 = < (1 && 1) >;
bitor = < (1 | 2) >;
bitxor = < (7 ^ 2) >;
bitand = < (3 & 6) >;
eq1 = < (1 == 0) >;
eq2 = < (1 == 1) >;
neq1 = < (1 != 0) >;
neq2 = < (1 != 1) >;
lt1 = < (1 < 2) >;
lt2 = < (2 < 2) >;
lt3 = < (3 < 2) >;
lteq1 = < (1 <= 2) >;
lteq2 = < (2 <= 2) >;
lteq3 = < (3 <= 2) >;
gt1 = < (1 > 2) >;
gt2 = < (2 > 2) >;
gt3 = < (3 > 2) >;
gteq1 = < (1 >= 2) >;
gteq2 = < (2 >= 2) >;
gteq3 = < (3 >= 2) >;
lshift = < (2 << 3) >;
rshift = < (16 >> 3) >;
add = < (3 + 4) >;
sub = < (7 - 4) >;
mul = < (3 * 4) >;
div = < (11 / 3) >;
mod = < (11 % 3) >;
unary_minus = < (-3) >;
bitnot = < (~1) >;
not0 = < (!-1) >;
not1 = < (!0) >;
not2 = < (!1) >;
not3 = < (!2) >;
nest = < (((--3) + (-2)) * (--(-2))) >;
char_lits = < ('a' + 'b') >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""",
"""
/dts-v1/;
/ {
ter1 = < 0x3 >;
ter2 = < 0x2 >;
ter3 = < 0x1 >;
ter4 = < 0x1 >;
or1 = < 0x0 >;
or2 = < 0x1 >;
or3 = < 0x1 >;
or4 = < 0x1 >;
and1 = < 0x0 >;
and2 = < 0x0 >;
and3 = < 0x0 >;
and4 = < 0x1 >;
bitor = < 0x3 >;
bitxor = < 0x5 >;
bitand = < 0x2 >;
eq1 = < 0x0 >;
eq2 = < 0x1 >;
neq1 = < 0x1 >;
neq2 = < 0x0 >;
lt1 = < 0x1 >;
lt2 = < 0x0 >;
lt3 = < 0x0 >;
lteq1 = < 0x1 >;
lteq2 = < 0x1 >;
lteq3 = < 0x0 >;
gt1 = < 0x0 >;
gt2 = < 0x0 >;
gt3 = < 0x1 >;
gteq1 = < 0x0 >;
gteq2 = < 0x1 >;
gteq3 = < 0x1 >;
lshift = < 0x10 >;
rshift = < 0x2 >;
add = < 0x7 >;
sub = < 0x3 >;
mul = < 0xc >;
div = < 0x3 >;
mod = < 0x2 >;
unary_minus = < 0xfffffffd >;
bitnot = < 0xfffffffe >;
not0 = < 0x0 >;
not1 = < 0x1 >;
not2 = < 0x0 >;
not3 = < 0x0 >;
nest = < 0xfffffffe >;
char_lits = < 0xc3 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""")
verify_error("""
/dts-v1/;
/ {
a = < (1/(-1 + 1)) >;
};
""",
".tmp.dts:4 (column 18): parse error: division by zero")
verify_error("""
/dts-v1/;
/ {
a = < (1%0) >;
};
""",
".tmp.dts:4 (column 11): parse error: division by zero")
#
# Test comment removal
#
verify_parse("""
/**//dts-v1//**/;//
//
// foo
/ /**/{// foo
x/**/=/*
foo
*/</**/1/***/>/****/;/**/}/*/**/;
""",
"""
/dts-v1/;
/ {
x = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""")
#
# Test get_node()
#
def verify_path_is(path, node_name):
try:
node = dt.get_node(path)
if node.name != node_name:
fail("expected {} to lead to {}, lead to {}"
.format(path, node_name, node.name))
except dtlib.DTError:
fail("no node found for path " + path)
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
def verify_path_error(path, msg):
prefix = "expected looking up '{}' to generate the error '{}', " \
"generated ".format(path, msg)
try:
dt.get_node(path)
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
def verify_path_exists(path):
if not dt.has_node(path):
fail("expected path '{}' to exist, didn't".format(path))
def verify_path_missing(path):
if dt.has_node(path):
fail("expected path '{}' to not exist, did".format(path))
dt = parse("""
/dts-v1/;
/ {
foo {
bar {
};
};
baz {
};
};
""")
verify_path_is("/", "/")
verify_path_is("//", "/")
verify_path_is("///", "/")
verify_path_is("/foo", "foo")
verify_path_is("//foo", "foo")
verify_path_is("///foo", "foo")
verify_path_is("/foo/bar", "bar")
verify_path_is("//foo//bar", "bar")
verify_path_is("///foo///bar", "bar")
verify_path_is("/baz", "baz")
verify_path_error("", "no alias '' found -- did you forget the leading '/' in the node path?")
verify_path_error("missing", "no alias 'missing' found -- did you forget the leading '/' in the node path?")
verify_path_error("/missing", "component 'missing' in path '/missing' does not exist")
verify_path_error("/foo/missing", "component 'missing' in path '/foo/missing' does not exist")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_path_exists("/")
verify_path_exists("/foo")
verify_path_exists("/foo/bar")
verify_path_missing("/missing")
verify_path_missing("/foo/missing")
#
# Test /aliases
#
def verify_alias_target(alias, node_name):
verify_path_is(alias, node_name)
if alias not in dt.alias2node:
fail("expected {} to be in alias2node".format(alias))
if dt.alias2node[alias].name != node_name:
fail("expected alias {} to lead to {}, lead to {}"
.format(alias, node_name, dt.alias2node[alias].name))
dt = parse("""
/dts-v1/;
/ {
aliases {
alias1 = &l1;
alias2 = &l2;
alias3 = &{/sub/node3};
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
alias4 = &{/node4};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
l1: node1 {
};
l2: node2 {
};
sub {
node3 {
};
};
node4 {
node5 {
};
};
};
""")
verify_alias_target("alias1", "node1")
verify_alias_target("alias2", "node2")
verify_alias_target("alias3", "node3")
verify_path_is("alias4/node5", "node5")
verify_path_error("alias4/node5/node6",
"component 'node6' in path 'alias4/node5/node6' does not exist")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_error("""
/dts-v1/;
/ {
aliases {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
a = [ 00 ];
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""",
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
"expected property 'a' on /aliases in .tmp.dts to be assigned with either 'a = &foo' or 'a = \"/path/to/node\"', not 'a = [ 00 ];'")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_error(r"""
/dts-v1/;
/ {
aliases {
a = "\xFF";
};
};
""",
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
r"value of property 'a' (b'\xff\x00') on /aliases in .tmp.dts is not valid UTF-8")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_error("""
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
/dts-v1/;
/ {
aliases {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
A = "/aliases";
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""",
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
"/aliases: alias property name 'A' should include only characters from [0-9a-z-]")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_error(r"""
/dts-v1/;
/ {
aliases {
a = "/missing";
};
};
""",
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
"property 'a' on /aliases in .tmp.dts points to the non-existent node \"/missing\"")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
# Test Property.type
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
def verify_type(prop, expected):
actual = dt.root.props[prop].type
if actual != expected:
fail("expected {} to have type {}, had type {}"
.format(prop, expected, actual))
dt = parse("""
/dts-v1/;
/ {
empty;
bytes1 = [ ];
bytes2 = [ 01 ];
bytes3 = [ 01 02 ];
bytes4 = foo: [ 01 bar: 02 ];
bytes5 = /bits/ 8 < 1 2 3 >;
num = < 1 >;
nums1 = < >;
nums2 = < >, < >;
nums3 = < 1 2 >;
nums4 = < 1 2 >, < 3 >, < 4 >;
string = "foo";
strings = "foo", "bar";
path1 = &node;
path2 = &{/node};
phandle1 = < &node >;
phandle2 = < &{/node} >;
phandles1 = < &node &node >;
phandles2 = < &node >, < &node >;
phandle-and-nums-1 = < &node 1 >;
phandle-and-nums-2 = < &node 1 2 &node 3 4 >;
phandle-and-nums-3 = < &node 1 2 >, < &node 3 4 >;
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
compound1 = < 1 >, [ 02 ];
compound2 = "foo", < >;
node: node {
};
};
""")
verify_type("empty", dtlib.TYPE_EMPTY)
verify_type("bytes1", dtlib.TYPE_BYTES)
verify_type("bytes2", dtlib.TYPE_BYTES)
verify_type("bytes3", dtlib.TYPE_BYTES)
verify_type("bytes4", dtlib.TYPE_BYTES)
verify_type("bytes5", dtlib.TYPE_BYTES)
verify_type("num", dtlib.TYPE_NUM)
verify_type("nums1", dtlib.TYPE_NUMS)
verify_type("nums2", dtlib.TYPE_NUMS)
verify_type("nums3", dtlib.TYPE_NUMS)
verify_type("nums4", dtlib.TYPE_NUMS)
verify_type("string", dtlib.TYPE_STRING)
verify_type("strings", dtlib.TYPE_STRINGS)
verify_type("phandle1", dtlib.TYPE_PHANDLE)
verify_type("phandle2", dtlib.TYPE_PHANDLE)
verify_type("phandles1", dtlib.TYPE_PHANDLES)
verify_type("phandles2", dtlib.TYPE_PHANDLES)
verify_type("phandle-and-nums-1", dtlib.TYPE_PHANDLES_AND_NUMS)
verify_type("phandle-and-nums-2", dtlib.TYPE_PHANDLES_AND_NUMS)
verify_type("phandle-and-nums-3", dtlib.TYPE_PHANDLES_AND_NUMS)
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_type("path1", dtlib.TYPE_PATH)
verify_type("path2", dtlib.TYPE_PATH)
verify_type("compound1", dtlib.TYPE_COMPOUND)
verify_type("compound2", dtlib.TYPE_COMPOUND)
#
# Test Property.to_{num,nums,string,strings,node}()
#
dt = parse(r"""
/dts-v1/;
/ {
u = < 1 >;
s = < 0xFFFFFFFF >;
u8 = /bits/ 8 < 1 >;
u16 = /bits/ 16 < 1 2 >;
u64 = /bits/ 64 < 1 >;
bytes = [ 01 02 03 ];
empty;
zero = < >;
two_u = < 1 2 >;
two_s = < 0xFFFFFFFF 0xFFFFFFFE >;
three_u = < 1 2 3 >;
three_u_split = < 1 >, < 2 >, < 3 >;
empty_string = "";
string = "foo\tbar baz";
invalid_string = "\xff";
strings = "foo", "bar", "baz";
invalid_strings = "foo", "\xff", "bar";
ref = <&{/target}>;
refs = <&{/target} &{/target2}>;
refs2 = <&{/target}>, <&{/target2}>;
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
path = &{/target};
manualpath = "/target";
missingpath = "/missing";
target {
phandle = < 100 >;
};
target2 {
};
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
};
""")
# Test Property.to_num()
def verify_to_num(prop, signed, expected):
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
try:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
actual = dt.root.props[prop].to_num(signed)
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
except dtlib.DTError as e:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
fail("failed to convert '{}' to {} number: {}"
.format(prop, "a signed" if signed else "an unsigned", e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
if actual != expected:
fail("expected {} to have the {} numeric value {:#x}, had the "
"value {:#x}".format(prop, "signed" if signed else "unsigned",
expected, actual))
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
def verify_to_num_error(prop, msg):
prefix = "expected fetching '{}' as a number to generate the error " \
"'{}', generated".format(prop, msg)
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
try:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
dt.root.props[prop].to_num()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_num("u", False, 1)
verify_to_num("u", True, 1)
verify_to_num("s", False, 0xFFFFFFFF)
verify_to_num("s", True, -1)
verify_to_num_error("two_u", "expected property 'two_u' on / in .tmp.dts to be assigned with 'two_u = < (number) >;', not 'two_u = < 0x1 0x2 >;'")
verify_to_num_error("u8", "expected property 'u8' on / in .tmp.dts to be assigned with 'u8 = < (number) >;', not 'u8 = [ 01 ];'")
verify_to_num_error("u16", "expected property 'u16' on / in .tmp.dts to be assigned with 'u16 = < (number) >;', not 'u16 = /bits/ 16 < 0x1 0x2 >;'")
verify_to_num_error("u64", "expected property 'u64' on / in .tmp.dts to be assigned with 'u64 = < (number) >;', not 'u64 = /bits/ 64 < 0x1 >;'")
verify_to_num_error("string", "expected property 'string' on / in .tmp.dts to be assigned with 'string = < (number) >;', not 'string = \"foo\\tbar baz\";'")
# Test Property.to_nums()
def verify_to_nums(prop, signed, expected):
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
try:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
actual = dt.root.props[prop].to_nums(signed)
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
except dtlib.DTError as e:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
fail("failed to convert '{}' to {} numbers: {}"
.format(prop, "signed" if signed else "unsigned", e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
if actual != expected:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
fail("expected {} to give the {} numbers {}, gave {}"
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
.format(prop, "signed" if signed else "unsigned", expected,
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
actual))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
def verify_to_nums_error(prop, msg):
prefix = "expected converting '{}' to numbers to generate the error " \
"'{}', generated".format(prop, msg)
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
try:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
dt.root.props[prop].to_nums()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_nums("zero", False, [])
verify_to_nums("u", False, [1])
verify_to_nums("two_u", False, [1, 2])
verify_to_nums("two_u", True, [1, 2])
verify_to_nums("two_s", False, [0xFFFFFFFF, 0xFFFFFFFE])
verify_to_nums("two_s", True, [-1, -2])
verify_to_nums("three_u", False, [1, 2, 3])
verify_to_nums("three_u_split", False, [1, 2, 3])
verify_to_nums_error("empty", "expected property 'empty' on / in .tmp.dts to be assigned with 'empty = < (number) (number) ... >;', not 'empty;'")
verify_to_nums_error("string", "expected property 'string' on / in .tmp.dts to be assigned with 'string = < (number) (number) ... >;', not 'string = \"foo\\tbar baz\";'")
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
# Test Property.to_bytes()
def verify_to_bytes(prop, expected):
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
try:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
actual = dt.root.props[prop].to_bytes()
except dtlib.DTError as e:
fail("failed to convert '{}' to bytes: {}".format(prop, e))
if actual != expected:
fail("expected {} to give the bytes {}, gave {}"
.format(prop, expected, actual))
def verify_to_bytes_error(prop, msg):
prefix = "expected converting '{}' to bytes to generate the error " \
"'{}', generated".format(prop, msg)
try:
dt.root.props[prop].to_bytes()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_bytes("u8", b"\x01")
verify_to_bytes("bytes", b"\x01\x02\x03")
verify_to_bytes_error("u16", "expected property 'u16' on / in .tmp.dts to be assigned with 'u16 = [ (byte) (byte) ... ];', not 'u16 = /bits/ 16 < 0x1 0x2 >;'")
verify_to_bytes_error("empty", "expected property 'empty' on / in .tmp.dts to be assigned with 'empty = [ (byte) (byte) ... ];', not 'empty;'")
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
# Test Property.to_string()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
def verify_to_string(prop, expected):
try:
actual = dt.root.props[prop].to_string()
except dtlib.DTError as e:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
fail("failed to convert '{}' to string: {}".format(prop, e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
if actual != expected:
fail("expected {} to have the value '{}', had the value '{}'"
.format(prop, expected, actual))
def verify_to_string_error(prop, msg):
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
prefix = "expected converting '{}' to string to generate the error " \
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
"'{}', generated".format(prop, msg)
try:
dt.root.props[prop].to_string()
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_string("empty_string", "")
verify_to_string("string", "foo\tbar baz")
verify_to_string_error("u", "expected property 'u' on / in .tmp.dts to be assigned with 'u = \"string\";', not 'u = < 0x1 >;'")
verify_to_string_error("strings", "expected property 'strings' on / in .tmp.dts to be assigned with 'strings = \"string\";', not 'strings = \"foo\", \"bar\", \"baz\";'")
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_string_error("invalid_string", r"value of property 'invalid_string' (b'\xff\x00') on / in .tmp.dts is not valid UTF-8")
# Test Property.to_strings()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
def verify_to_strings(prop, expected):
try:
actual = dt.root.props[prop].to_strings()
except dtlib.DTError as e:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
fail("failed to convert '{}' to strings: {}".format(prop, e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
if actual != expected:
fail("expected {} to have the value '{}', had the value '{}'"
.format(prop, expected, actual))
def verify_to_strings_error(prop, msg):
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
prefix = "expected converting '{}' to strings to generate the error " \
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
"'{}', generated".format(prop, msg)
try:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
dt.root.props[prop].to_strings()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_strings("empty_string", [""])
verify_to_strings("string", ["foo\tbar baz"])
verify_to_strings("strings", ["foo", "bar", "baz"])
verify_to_strings_error("u", "expected property 'u' on / in .tmp.dts to be assigned with 'u = \"string\", \"string\", ... ;', not 'u = < 0x1 >;'")
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_strings_error("invalid_strings", r"value of property 'invalid_strings' (b'foo\x00\xff\x00bar\x00') on / in .tmp.dts is not valid UTF-8")
# Test Property.to_node()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
def verify_to_node(prop, path):
try:
actual = dt.root.props[prop].to_node().path
except dtlib.DTError as e:
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
fail("failed to convert '{}' to node: {}".format(prop, e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
if actual != path:
fail("expected {} to point to {}, pointed to {}"
.format(prop, path, actual))
def verify_to_node_error(prop, msg):
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
prefix = "expected converting '{}' to a node to generate the error " \
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
"'{}', generated".format(prop, msg)
try:
dt.root.props[prop].to_node()
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_node("ref", "/target")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
verify_to_node_error("u", "expected property 'u' on / in .tmp.dts to be assigned with 'u = < &foo >;', not 'u = < 0x1 >;'")
verify_to_node_error("string", "expected property 'string' on / in .tmp.dts to be assigned with 'string = < &foo >;', not 'string = \"foo\\tbar baz\";'")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
# Test Property.to_nodes()
def verify_to_nodes(prop, paths):
try:
actual = [node.path for node in dt.root.props[prop].to_nodes()]
except dtlib.DTError as e:
fail("failed to convert '{}' to nodes: {}".format(prop, e))
if actual != paths:
fail("expected {} to point to the paths {}, pointed to {}"
.format(prop, paths, actual))
def verify_to_nodes_error(prop, msg):
prefix = "expected converting '{}' to a nodes to generate the error " \
"'{}', generated".format(prop, msg)
try:
dt.root.props[prop].to_nodes()
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
verify_to_nodes("zero", [])
verify_to_nodes("ref", ["/target"])
verify_to_nodes("refs", ["/target", "/target2"])
verify_to_nodes("refs2", ["/target", "/target2"])
verify_to_nodes_error("u", "expected property 'u' on / in .tmp.dts to be assigned with 'u = < &foo &bar ... >;', not 'u = < 0x1 >;'")
verify_to_nodes_error("string", "expected property 'string' on / in .tmp.dts to be assigned with 'string = < &foo &bar ... >;', not 'string = \"foo\\tbar baz\";'")
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
# Test Property.to_path()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
def verify_to_path(prop, path):
try:
actual = dt.root.props[prop].to_path().path
except dtlib.DTError as e:
fail("failed to convert '{}' to path: {}".format(prop, e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
if actual != path:
fail("expected {} to contain the path {}, contained {}"
.format(prop, path, actual))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
def verify_to_path_error(prop, msg):
prefix = "expected converting '{}' to a path to generate the error " \
"'{}', generated".format(prop, msg)
try:
dt.root.props[prop].to_path()
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_path("path", "/target")
verify_to_path("manualpath", "/target")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_to_path_error("u", "expected property 'u' on / in .tmp.dts to be assigned with either 'u = &foo' or 'u = \"/path/to/node\"', not 'u = < 0x1 >;'")
verify_to_path_error("missingpath", "property 'missingpath' on / in .tmp.dts points to the non-existent node \"/missing\"")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
# Test top-level to_num() and to_nums()
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
def verify_raw_to_num(fn, prop, length, signed, expected):
try:
actual = fn(dt.root.props[prop].value, length, signed)
except dtlib.DTError as e:
fail("failed to convert '{}' to {} number(s) with {}: {}"
.format(prop, "signed" if signed else "unsigned",
fn.__name__, e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
if actual != expected:
fail("expected {}(<{}>, {}, {}) to be {}, was {}"
.format(fn.__name__, prop, length, signed, expected, actual))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
def verify_raw_to_num_error(fn, data, length, msg):
prefix = "expected {}() called with data='{}', length='{}' to " \
"generate the error '{}', generated" \
.format(fn.__name__, data, length, msg)
try:
fn(data, length)
fail(prefix + " no error")
except dtlib.DTError as e:
if str(e) != msg:
fail("{} the error '{}'".format(prefix, e))
except Exception as e:
fail("{} the non-DTError '{}'".format(prefix, e))
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_raw_to_num(dtlib.to_num, "u", None, False, 1)
verify_raw_to_num(dtlib.to_num, "u", 4, False, 1)
verify_raw_to_num(dtlib.to_num, "s", None, False, 0xFFFFFFFF)
verify_raw_to_num(dtlib.to_num, "s", None, True, -1)
verify_raw_to_num(dtlib.to_nums, "empty", 4, False, [])
verify_raw_to_num(dtlib.to_nums, "u16", 2, False, [1, 2])
verify_raw_to_num(dtlib.to_nums, "two_s", 4, False, [0xFFFFFFFF, 0xFFFFFFFE])
verify_raw_to_num(dtlib.to_nums, "two_s", 4, True, [-1, -2])
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_raw_to_num_error(dtlib.to_num, 0, 0, "'0' has type 'int', expected 'bytes'")
verify_raw_to_num_error(dtlib.to_num, b"", 0, "'length' must be greater than zero, was 0")
verify_raw_to_num_error(dtlib.to_num, b"foo", 2, "b'foo' is 3 bytes long, expected 2")
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
verify_raw_to_num_error(dtlib.to_nums, 0, 0, "'0' has type 'int', expected 'bytes'")
verify_raw_to_num_error(dtlib.to_nums, b"", 0, "'length' must be greater than zero, was 0")
verify_raw_to_num_error(dtlib.to_nums, b"foooo", 2, "b'foooo' is 5 bytes long, expected a length that's a a multiple of 2")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# Test duplicate label error
#
verify_error("""
/dts-v1/;
/ {
sub1 {
label: foo {
};
};
sub2 {
label: bar {
};
};
};
""",
"Label 'label' appears on /sub1/foo and on /sub2/bar")
verify_error("""
/dts-v1/;
/ {
sub {
label: foo {
};
};
};
/ {
sub {
label: bar {
};
};
};
""",
"Label 'label' appears on /sub/bar and on /sub/foo")
verify_error("""
/dts-v1/;
/ {
foo: a = < 0 >;
foo: node {
};
};
""",
"Label 'foo' appears on /node and on property 'a' of node /")
verify_error("""
/dts-v1/;
/ {
foo: a = < 0 >;
node {
foo: b = < 0 >;
};
};
""",
"Label 'foo' appears on property 'a' of node / and on property 'b' of node /node")
verify_error("""
/dts-v1/;
/ {
foo: a = foo: < 0 >;
};
""",
"Label 'foo' appears in the value of property 'a' of node / and on property 'a' of node /")
# Giving the same label twice for the same node is fine
verify_parse("""
/dts-v1/;
/ {
sub {
label: foo {
};
};
};
/ {
sub {
label: foo {
};
};
};
""",
"""
/dts-v1/;
/ {
sub {
label: foo {
};
};
};
""")
# Duplicate labels are fine if one of the nodes is deleted
verify_parse("""
/dts-v1/;
/ {
label: foo {
};
label: bar {
};
};
/delete-node/ &{/bar};
""",
"""
/dts-v1/;
/ {
label: foo {
};
};
""")
#
# Test overriding/deleting a property with references
#
verify_parse("""
/dts-v1/;
/ {
x = &foo, < &foo >;
y = &foo, < &foo >;
foo: foo {
};
};
/ {
x = < 1 >;
/delete-property/ y;
};
""",
"""
/dts-v1/;
/ {
x = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
foo: foo {
};
};
""")
#
# Test self-referential node
#
verify_parse("""
/dts-v1/;
/ {
label: foo {
x = &{/foo}, &label, < &label >;
};
};
""",
"""
/dts-v1/;
/ {
label: foo {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
x = &{/foo}, &label, < &label >;
phandle = < 0x1 >;
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
""")
#
# Test /memreserve/
#
dt = verify_parse("""
/dts-v1/;
l1: l2: /memreserve/ (1 + 1) (2 * 2);
/memreserve/ 0x100 0x200;
/ {
};
""",
"""
/dts-v1/;
l1: l2: /memreserve/ 0x0000000000000002 0x0000000000000004;
/memreserve/ 0x0000000000000100 0x0000000000000200;
/ {
};
""")
expected = [(["l1", "l2"], 2, 4), ([], 0x100, 0x200)]
if dt.memreserves != expected:
fail("expected {} for dt.memreserve, got {}"
.format(expected, dt.memreserves))
verify_error("""
/dts-v1/;
foo: / {
};
""",
".tmp.dts:3 (column 6): parse error: expected /memreserve/ after labels at beginning of file")
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
#
# Test __repr__() functions
#
def verify_repr(obj, expected):
if repr(obj) != expected:
fail("expected repr() to be '{}', was '{}'"
.format(expected, repr(obj)))
dt = parse("""
/dts-v1/;
/ {
x = < 0 >;
sub {
y = < 1 >;
};
};
""",
include_path=("foo", "bar"))
verify_repr(dt, "DT(filename='.tmp.dts', include_path=('foo', 'bar'))")
verify_repr(dt.root.props["x"], "<Property 'x' at '/' in '.tmp.dts'>")
verify_repr(dt.root.nodes["sub"], "<Node /sub in '.tmp.dts'>")
#
# Test names
#
# The C tools disallow '@' in property names, but otherwise accept the same
# characters in node and property names. Emulate that instead of the DT spec
# (v0.2), which gives different characters for nodes and properties.
verify_parse(r"""
/dts-v1/;
/ {
// A leading \ is accepted but ignored in node/propert names
\aA0,._+*#?- = &_, &{/aA0,._+*#?@-};
// Names that overlap with operators and integer literals
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
+ = [ 00 ];
* = [ 02 ];
- = [ 01 ];
? = [ 03 ];
0 = [ 04 ];
0x123 = [ 05 ];
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
_: \aA0,._+*#?@- {
};
0 {
};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""",
"""
/dts-v1/;
/ {
dts: dtlib/edtlib: Add a syntax-based type-checking system Property type-checking has been pretty rudimentary until now, only checking things like the length being divisible by 4 for 'type: array', and strings being null-terminated. In particular, no checking was done for 'type: uint8-array', letting jedec-id = < 0xc8 0x28 0x17 >; slip through when jedec-id = [ 0xc8 0x28 0x17 ]; was intended. Fix it by adding a syntax-based type checker: 1. Add Property.type, which gives a high-level type for the property, derived from the markers added in the previous commit. This includes types like TYPE_EMPTY ('foo;'), TYPE_NUM ('foo = < 3 >;'), TYPE_BYTES ('foo = [ 01 02 ];'), TYPE_STRINGS ('foo = "bar", "baz"'), TYPE_PHANDLE ('foo = < &bar >;'), and TYPE_COMPOUND (everything not recognized). See the Property.type docstring in dtlib for more info. 2. Use the high-level type in Property.to_num()/to_string()/to_node()/etc. to verify that the property was assigned in an expected way for the type. If the assignment looks bad, give a helpful error: expected property 'nums' on /foo/bar in some.dts to be assigned with 'nums = < (number) (number) ... >', not 'nums = "oops";' Some other related changes are included as well: - There's a new Property.to_bytes() function that works like accessing Property.bytes, except with an added check for the value being assigned like 'foo = [ ... ]'. This function solves problems like the jedec-id one. - There's a new Property.to_path() function for fetching the referenced node for assignments like 'foo = &node;', with type checking. (Strings are accepted too, as long as they give the path to an existing node.) This function is used for /chosen and /aliases. - A new 'type: phandle' type can now be given in bindings, for properties that are assigned like 'foo = < &node >;'. - Property.__str__() now displays phandles and path references as they were written (e.g. '< &foo >' instead of '< 0x1 >', if the allocated phandle happened to be 1). - Property.to_num() and Property.to_nums() no longer take a 'length' parameter, because it makes no sense with the type checking. - The global dtlib.to_string() and dtlib.to_strings() functions were removed, because they're not that useful. - More tests were added, along with misc. minor cleanup in various places. - Probably other stuff I forgot. The more strict type checking in dtlib indirectly makes some parts of edtlib more strict as well (wherever Property.to_*() is used). Fixes: #18131 Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2019-08-09 20:38:17 +02:00
aA0,._+*#?- = &_, &{/aA0,._+*#?@-};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
+ = [ 00 ];
* = [ 02 ];
- = [ 01 ];
? = [ 03 ];
0 = [ 04 ];
0x123 = [ 05 ];
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
_: aA0,._+*#?@- {
};
0 {
};
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
""")
verify_error(r"""
/dts-v1/;
/ {
foo@3;
};
""",
".tmp.dts:4 (column 7): parse error: '@' is only allowed in node names")
verify_error(r"""
/dts-v1/;
/ {
foo@3 = < 0 >;
};
""",
".tmp.dts:4 (column 8): parse error: '@' is only allowed in node names")
verify_error(r"""
/dts-v1/;
/ {
foo@2@3 {
};
};
""",
".tmp.dts:4 (column 10): parse error: multiple '@' in node name")
#
# Test bad formatting
#
verify_parse("""
/dts-v1/;/{l1:l2:foo{l3:l4:bar{l5:x=l6:/bits/8<l7:1 l8:2>l9:,[03],"a";};};};
""",
"""
/dts-v1/;
/ {
l1: l2: foo {
l3: l4: bar {
l5: x = l6: [ l7: 01 l8: 02 l9: ], [ 03 ], "a";
dts: Add new DTS/binding parser Add a new DTS/binding parser to scripts/dts/ for generating generated_dts_board.conf and generated_dts_board_unfixed.h. The old code is kept to generate some deprecated defines, using the --deprecated-only flag. It will be removed later. The new parser is implemented in three files in scripts/dts/: dtlib.py: A low-level .dts parsing library. This is similar to devicetree.py in the old code, but is a general robust DTS parser that doesn't rely on preprocessing. edtlib.py (e for extended): A library built on top of dtlib.py that brings together data from DTS files and bindings and creates Device instances with all the data for a device. gen_defines.py: A script that uses edtlib.py to generate generated_dts_board.conf and generated_dts_board_unfixed.h. Corresponds to extract_dts_includes.py and the files in extract/ in the old code. testdtlib.py: Test suite for dtlib.py. Can be run directly as a script. testedtlib.py (uses test.dts and test-bindings/): Test suite for edtlib.py. Can be run directly as a script. The test suites will be run automatically in CI. The new code turns some things that were warnings (or not checked) in the old code into errors, like missing properties that are specified with 'category: required' in the binding for the node. The code includes lots of documentation and tries to give helpful error messages instead of Python errors. Co-authored-by: Kumar Gala <kumar.gala@linaro.org> Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
2018-12-17 20:09:47 +01:00
};
};
};
""")
#
# Test misc. errors and non-errors
#
verify_error("", ".tmp.dts:1 (column 1): parse error: expected '/dts-v1/;' at start of file")
verify_error("""
/dts-v1/;
""",
".tmp.dts:2 (column 1): parse error: no root node defined")
verify_error("""
/dts-v1/; /plugin/;
""",
".tmp.dts:1 (column 11): parse error: /plugin/ is not supported")
verify_error("""
/dts-v1/;
/ {
foo: foo {
};
};
// Only one label supported before label references at the top level
l1: l2: &foo {
};
""",
".tmp.dts:9 (column 5): parse error: expected label reference (&foo)")
verify_error("""
/dts-v1/;
/ {
foo: {};
};
""",
".tmp.dts:4 (column 14): parse error: expected node or property name")
# Multiple /dts-v1/ at the start of a file is fine
verify_parse("""
/dts-v1/;
/dts-v1/;
/ {
};
""",
"""
/dts-v1/;
/ {
};
""")
print("all tests passed")
# Only remove these if tests succeed. They're handy to have around otherwise.
shutil.rmtree(".tmp")
shutil.rmtree(".tmp2")
os.remove(".tmp.dts")
os.remove(".tmp2.dts")
os.remove(".tmp3.dts")
os.remove(".tmp.bin")
if __name__ == "__main__":
run()