Commit 73cb9586ce renamed linker
section from _k_memory_pool to _k_mem_pool, but the references
in linker scripts are still there. So remove them now.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Running checkpatch in pre-commit hook could be a constraint
for day to day work. Though, running it before pushing to
zephyr repo is highly recommended.
Add a series push script callable from pre-push hook.
Document how to use it in contribute section.
Signed-off-by: Erwan Gouriou <erwan.gouriou@linaro.org>
In order to simplify the usage of `west build`, take a positional
argument with the source directory instead of requiring the `-s,
--source-dir` flag. This makes it easier and quicker to invoke west when
building, as well as being consistent with CMake.
Signed-off-by: Carles Cufi <carles.cufi@nordicsemi.no>
Users often get confused when they change Kconfig sources and then
rebuild only to discover that nothing happens. To fix this we add a
dependency between re-running cmake, and all Kconfig sources, similair
to how touching CMakeLists.txt files cause CMake to re-run.
This fixes#5634
Signed-off-by: Sebastian Bøe <sebastian.boe@nordicsemi.no>
Dictionaries are iterated in a random order by Python 3.5 and before.
This could have caused "Unstable" CI in PR #13921 and maybe others.
Anyway we want builds to be determimistic by default. Explicit
randomness can be added for better coverage but not by default.
1. When running "make kobj_types_h_target" repeatedly one can observe
that the following .h files keep changing in
build/zephyr/include/generated/:
- kobj-types-enum.h
- otype-to-str.h
- otype-to-size.h
Switching kobjects to OrderedDict makes these 3 .h files deterministic.
2. When running this test repeatedly with CONFIG_USERSPACE=y:
rm build/zephyr/*.gperf && make -C build obj_list
... the dict used for --gperf-output seems to be deterministic, probably
because its keys are all integers (memory addresses). However we can't
take that for granted with Python < 3.6 so out of caution also switch
the output of find_objects() in elf_helper.py to a sorted OrderedDict.
PS: I would normally prefer official Python documentation to
StackOverflow however this one is a good summary and has all the
multiple pointers to the... official Python documentation.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
To reproduce the issue this addresses, run "cmake -B96b_nitrogen"
twice with Python 3.5. Observe the order of the flash partition table
changing in:
"build{1,2}/zephyr/include/generated/generated_dts_board.conf" and
"generated_dts_board_unfixed.h"
Dictionaries are iterated in a random order by Python 3.5 and before.
This could have caused "Unstable" CI in PR #13921 and maybe others.
Anyway we want builds to be determimistic by default. Explicit
randomness can be added for better coverage but not by default.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Sigvart pointed out that the error message for missing Kconfig files
(which rambles on a lot about $srctree) can easily kick people off in
the wrong direction.
Update Kconfiglib to upstream revision 9f26eb3ae3 to add the following
commit, which makes the error message point to stuff that's more likely
to be the problem:
Give more helpful error messages when files are missing
The current error message talks a lot about $srctree, but $srctree
is seldom the culprit in practice. More common is 'source
"$(SOME_ENV_VAR)/foo"`, where SOME_ENV_VAR hasn't been set.
Include the complete 'source ...' line for missing Kconfig files,
and mention unset environment variables as a hint. Only mention
$srctree briefly.
Also shorten the message when a .config can't be a found a bit. This
message would usually only be seen when working directly with the
library.
Example error:
Kconfig:7: '/Kconfig' not found (in 'source
"$(SOME_ENV_VAR)/Kconfig"'). Check that environment variables are
set correctly (e.g. $srctree, which is unset or blank). Also note
that unset environment variables expand to the empty string.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
App shared memory defines a bunch of symbols in the
linker script. Namespace them properly as private
zephyr variables.
The variables which indicate the bounds of the entire
partition now end with "_part_start", "_part_size",
and "_part_end" to make them easy for scripts to
distinguish them from other generated symbols for
data/bss sizes.
Finally, the bss size is not rounded up, this was
causing unnecessary memory to be zeroed.
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
When re-running failed tests, do not go through any filtering, just load
the failed tests directly and execute them. The filtering was done based
on default platforms and any tests that were failing on non-default
platforms were ignored.
Fixes#13956
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
This is useful when running the script locally on a set of commits to
reproduce how things are run in CI.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Zephyr Continuous Integration made the choice of rebasing pull requests
on its fast moving master branch before testing them. Unfortunately the
Shippable interface assumes the other choice[*] and highlights the SHA1
of the pull request _before_ rebasing which is very misleading. For now
the only way to realize this disconnect is to scrutinize the build logs
and/or this source code (or tribal knowledge).
To mitigate and reduce a bit this confusion, run a short:
git log -n 5 --oneline --decorate
before and after rebase.
Making it worse, Shippable sub-runs are not pegged to a common base
version. For instance the reason why shippable sub-run 35578.1 passed
while all other 35578.x sub runs failed is very likely because PR #13803
was merged to master in the mean time. This commit would have been able
to prove that.
[*] In an ideal world with infinite time and resources, Continuous
Integration would test pull requests BOTH alone and combined with the
moving tip.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
On (at least) Fedora "pytest" is python2 and python2 only.
Also update remote name in sample invocation from "upstream" to "origin"
because that's how west sets things up by default (and this script
requires west)
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Was apparently an artifact of Python2 to Python3 conversion. Led to
printing of literal '='*110i instead of a line of ='s.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
When building incrementally, filter-known-issues.py reports a varying
and totally different set of "new" issues than when building from
scratch. Warnings for unrelated upstream code disappearing and
re-appearing are especially confusing. Expand the messages a bit to
clarify.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
insert_defs() is the only function that adds keys to 'defs', and they
always get added with an 'aliases' key.
This function could get super simple if the duplicate-removal code could
be removed. It only seems to be needed to prevent duplicates for some
(unfortunately tricky) flash-related stuff now.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
- No deep copying is needed. The deepcopy() is just a hack to work
around that the value might not be a list. That deserves a comment at
least.
- A catch-all try/except will hide stuff like misspelled variables as
well. It's usually a bad idea.
Rewrite things to be more explicit, and add some comments.
reg.py was indirectly using the deepcopy() imported in global.py,
because it does 'from global import *'. Have it import deepcopy() itself
instead. reg.py is the only remaining user of deepcopy().
reg.py relying on the deepcopy() import in global.py was super hard to
discover, due to another catch-all try/except around the deepcopy()
call.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
All nodes have a 'children' key.
Also simplify a bit. The loop is copying a dictionary verbatim.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
'/foo' was turned into ''. Turn it into '/' instead. That's how the root
is represented in 'reduced' too.
Also remove some code from get_addr_size_cells() that was only there to
work around the bug.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
If a binding specifies 'generate: define' for 'interrupt-names' (like
some bindings do), then that ought to generate #defines for it, but the
scripts/dts code currently hardcodes 'interrupt-names' to be ignored
(along with some other properties).
Maybe the 'generate: define' in those bindings is a mistake, but the
code still ought to respect it. That also gets rid of some mystery code.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
This is an integral part of userspace and cannot be used
on its own. Fold into the main userspace configuration.
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Add some short doc comments at the beginning so that people can quickly
get an idea of what they're about.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
It's confusing that "address" is often used within the same function to
refer to both node paths and e.g. address cells.
Make things easier to understand by calling /foo/bar a path instead.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
extract_property() is meant to generate #defines for a single property,
like 'foo = <1 2 3>'. Currently, it also generates node-level #defines
related to parent buses.
That makes the intent of the code hard to understand, and also means
that identical node-level #defines get redundantly added multiple times
(once per property).
Generate the node-level bus #defines before processing properties, in
generate_node_defines(). Use a new generate_bus_defines() helper.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Due to the way the code was structured,
flash.extract_partition(node_path) was called multiple times for the
same 'node_path'.
That must've been a mistake to begin with (but was hard to spot before
the code was cleaned up). Move the generation of per-node #defines out
of the property loop.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Do not load the GENERATED_DTS_BOARD_CONF if in doc mode, since it will
not defined as it doesn't apply. No need to defined it to a dummy value.
Signed-off-by: Carles Cufi <carles.cufi@nordicsemi.no>
generate_node_defines() immediately returns if the node's 'compatible'
is not in the binding.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Derive 'node_compat' and 'prop_val' (the dictionary for the property
from the YAML binding) inside extract_property().
That gives it just two parameters and makes it clearer that it's just
generating #define's for a single device tree property. 'prop_val' was
only used to look up prop_val['type'].
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Moves it closer to where it's used, and will allow other simplifications
in generate_node_defines().
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
After lots of reverse-engineering, I understand why deepcopy() is used:
1) Various functions defined in scripts/dts/extract/* modify the list
of names in-place
2) A plain list.copy() won't work when the *-names property happens to
have a single name in it, because devicetree.py gives a string
instead of a list in that case
Using deepcopy() to solve (2) is very confusing (especially with no
comments), because no deep copying is actually needed.
Get rid of deepcopy(), add a helper function for fetching the names, and
some comments.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
- Remove dead "if 'props' in nodes" check
- Rename the 'nodes' parameter to 'node'. It represents a single node.
- Use a defaultdict(int) for 'last_used_id' (makes unset keys give 0)
- Use a global for 'last_used_id'
- Move some related code so it appears together
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Update Kconfiglib (and menuconfig, just to sync) to upstream revision
99a7af769352b, to add the commit below, for a doc issue reported by
Sebastian Bøe:
Document that kconfig_filenames keeps absolute paths as-is
Came up in https://github.com/ulfalizer/Kconfiglib/issues/67.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
The commit header as added by "git commit -s" is "Signed-off-by:",
and not any other casing. While it perhaps case-insensitive, use
canonical casing to not confuse (attentive) users.
Signed-off-by: Paul Sokolovsky <paul.sokolovsky@linaro.org>
Turn
if a:
if b:
...
into
if a and b:
...
Simplify some error messages with .format() as well, and get rid of some
redundant str()s. '{0} {1} {1}'.format('foo', 'bar') gives
'foo bar bar'.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
- Reduce the indentation and remove the awkward line breaks. This is
possible now that there's no recursion.
- Be consistent with quotes
- Move some initialization closer to where it's used
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
The recursion was used to implement 'properties:' within 'properties:'
in binding files, which seems to be a dead leftover (and undocumented).
Removing it gets rid of code and makes things more transparent.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Use more descriptive naming, get rid of unused 'v' variable, and an
unnecessary None check (None won't appear in get_binding_compats()).
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
...to generate_node_defines().
More explicit. "extract node include info" can be read in many ways:
- Extract a node's "include info"
- Extract node and include info
- etc.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
