lcov 2.0 added support for processing coverage data in parallel, which
provides a large speedup when processing many files, at the cost of some
additional overhead. When running the Chrome EC tests with coverage,
parallel reporting on a 36C72T machine reduces the time spent generating
coverage reports by 40 minutes (from approximately 1 hour to 20 minutes
total runtime), at the cost of about 3x greater CPU time overall
(assumed to be overhead for parallel processing, likely from spawning
much larger numbers of subprocesses).
The level of lcov parallelism is taken from the --jobs option passed to
twister, allowing lcov to choose if unspecified.
Signed-off-by: Peter Marheine <pmarheine@chromium.org>
This moves handling of changed lcov arguments in lcov 2.0 into helper
functions, significantly simplifying Lcov._generate() to reduce the
visual noise of extra arguments that are required but unimportant to the
task at hand.
Signed-off-by: Peter Marheine <pmarheine@chromium.org>
Make sure we set the gcov tool in a consistent way and avoid issues
where path is set as Path instead of a string.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
BinaryHandler.hanlde() method was trying to call GCOV and fails
silently because of incorrect call parameters. Moreover, even
being fixed, this call is not needed here as it attempts to
find and process .gcno and .gcna files (using wrong paths)
to create .gcov coverage text reports.
Currently the CoverageTool class does all data processing
and report generation using GCOVR or LCOV tools instead of
direct calls to GCOV.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
Verify we have the coverage tool we want to use, otherwise we will end
up with many warnings and errors during coverage data collection.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Additional checks for Twister command line options `--coverage-tool`
and `--coverage-formats`.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
Twister now uses GCOVR by default as the more reliable code
coverage reporting tool instead of LCOV.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
Add two new keys: tags, tests.
tags for aligning with what we use in tests and samples and tests to
associate areas and components with tests.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Documentation pages around Kconfig have moved around on a couple of
occasions in the past [1] [2] and current redirects were dysfonctional
(leading /) and incomplete, while an entire set of redirects was also
missing. This notably fixes the "browse latest development version of
this page" link on
https://docs.zephyrproject.org/2.7.5/guides/build/kconfig/tips.html or a
link to the "Kconfig tips" page as found in Kconfiglib's current README
(https://pypi.org/project/kconfiglib/)
[1] commit 5342bc64dd [2] commit
5c88418428Fixes#66701
Signed-off-by: Benjamin Cabé <benjamin@zephyrproject.org>
When scanning ports the manufacture field is not always
filled. It must be checked before using.
Signed-off-by: Grzegorz Chwierut <grzegorz.chwierut@nordicsemi.no>
The HEAP_MEM_POOL_ADD_SIZE_ is used as a prefix for matching specific
Kconfig option names, i.e. it's not a real option in itself.
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
In some cases genhtml incorrectly built reports from LCOV
coverage data using full path for some of the source files
and relative paths for other files.
This fix adds `--prefix` parameter to shorten paths explicitly
relative to the ZEPHYR_BASE directory.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
This gives us the flexibility to exclude certain platforms'
simulator that's known to fail from running in Twister, but
allows real hardware to be tested.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
NXP boards with CMSID-DAP are not detected by twister
--generate-hardware-map, because serial device name 'mbed' is compared
with upper case 'MBED' in a list of supported manufacturers.
Fix it by making the comparison case-insensitive.
Tested using mimxrt1020_evk.
Fixes#63765
Signed-off-by: Michal Smola <michal.smola@nxp.com>
Rename the bitmask variables from `*_LVL_INTERRUPTS` to
`INTERRUPT_LVL_BITMASK[]` array to be consistent with
`INTERRUPT_BITS`, making it easier to loop over the bitmasks.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
The calculation of `THIRD_LVL_INTERRUPTS` bitmask in the
`update_masks()` function is wrong, the number of bits to shift
should be the sum of the first two levels.
Signed-off-by: Yong Cong Sin <ycsin@meta.com>
Allow using the C pre-processor to generate a
`rimage/config/platform.toml` file from a "source"
`rimage/config/platform.toml.h` file.
This is optional and fully backwards compatible.
To use, do not use `-c` and point west sign at a configuration directory
instead or let it use the default `rimage/config/` directory and change
the files there.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
rimage is very verbose by default and has no -q(uiet) option, so saving
one line out of more than 100 lines is pointless.
RimageSigner.sign() was already very complex and suffering from
combinatorial explosion of parameters. With .toml
pre-processing (#65411) it's getting worse, so we really need all build
logs to show the complete rimage command.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
In the following command, the first argument `for_rimage` is passed to
`rimage` whereas `--for west` goes to west.
```
west sign for_rimage --for west
```
This is somehow valid but we really don't want anyone to do that.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
Zero functional change, preparation for the .toml modularization.
RimageSigner.sign() is also way too long and too complex and this helps.
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
`west sign` has been invoked by `west build` (through CMake) since
commit fad2da39aa, almost one year ago. During that time, this new
workflow has been refined and successfully used by at least two vendors,
multiple CIs across both SOF and Zephyr and many developers.
At the time, the ability to sign from `west flash` was preserved for
backwards compatibility. This means rimage parameters can come from many
different places at once and that rimage can be invoked twice during a
single `west flash` invocation!
Now that Zephyr 3.5 has been released, we need to reduce the number of
rimage use cases and the corresponding validation complexity and
maintenance workload to simplify and accelerate new features like
splitting rimage configuration files (#65411)
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
This removes escape from generate XML_CATALOG_EACH string, because
the string between CDATA tag need to be treated as normal text.
Signed-off-by: Fang Huang <fang.huang@intel.com>
Fixes#65477
Platform key checking seemed to be erroneous;
now the variable names, comments and code seem in line with each other.
Signed-off-by: Lukasz Mrugala <lukaszx.mrugala@intel.com>
Support lcov > 2.0 tool which has strict error checking and some new
configuration options deprecating syntax used in 1.4 versions.
Fixes#62202
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Improve documentation about pytest integration with Twister. Add
examples of usage, improve description of available options and
introduce automatic doc generation of two plugin classes (DeviceAdapter
and Shell) basing on their docstrings from source code.
Signed-off-by: Piotr Golyzniak <piotr.golyzniak@nordicsemi.no>
Introduce Kconfig option in zephyr build system that reflects the TF-M
cmake config variable with the same default value for dummy provisioning
and have it satisfy the IAK present requirement.
This configuration is not suitable for production, and by having this
in zephyr configuration we can have this as part of the hardened
configuration check.
Signed-off-by: Joakim Andersson <joakim.andersson@nordicsemi.no>
as in windows platfrom the format by default is not utf-8,
and we will see below error
Traceback (most recent call last):
File "...zephyr\scripts\twister", line 211, in <module>
ret = main(options)
^^^^^^^^^^^^^
File "...scripts/pylib/twister\twisterlib\twister_main.py",
tplan.load()
File "...scripts/pylib/twister\twisterlib\testplan.py",
self.load_from_file(last_run, filter_platform=connected_list)
File "...scripts/pylib/twister\twisterlib\testplan.py",
instance.create_overlay(platform, self.options.enable_asan,
self.options.enable_ubsan, self.options.enable_coverage,
self.options.coverage_platform)
File "...scripts/pylib/twister\twisterlib\testinstance.py"
f.write(content)
UnicodeEncodeError: 'gbk' codec can't encode character '\xf8'
in position 64: illegal multibyte sequence
Signed-off-by: Hake Huang <hake.huang@oss.nxp.com>
It's not good to see struct_tags.json change from one build to the next
when nothing changes.
Python's sets are not deterministic, see long story for older commit
f896fc2306 ("scripts: gen_handles: Sort the device handles")
Simply convert multiple_directories to a sorted list before using it.
Fixes commit 80e78208e6 ("kernel: syscalls: no need to include all
syscalls in binary")
Signed-off-by: Marc Herbert <marc.herbert@intel.com>
The Console Harness is able to parse its log with patterns to compose
extracted fields into records in 'recording.csv' file in the test's build
directory. This feature allows to extract custom test results like
performance counters.
With this change the extracted records are also written into 'twister.json'
as a part of each test suite object. This makes easier to store
all the data collected by the test for its further processing.
Other improvements:
- compile parsing pattern only once instead of at each input line;
- quote fields in '.csv' to avoid unexpected field separators;
- make 'regex' a required schema field of 'harness_config';
- Twister documentation update.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
This changes the logic in __init__() so that if a path to
MAINTAINERS.yml is passed in, it uses the passed-in value
instead of blindly running git to find the top level of
Zephyr tree.
Signed-off-by: Daniel Leung <daniel.leung@intel.com>
Change the keep-sorted check to handle blocks of code uniformly
indented.
Suggested-by: Benjamin Cabé <benjamin@zephyrproject.org>
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
A race condition was possible at QEMUHandler which didn't wait for
its harness to complete causing false negatives and test suite
retries, for example when the console harness has got rather
long output from a test application to check for patterns.
Signed-off-by: Dmitrii Golovanov <dmitrii.golovanov@intel.com>
