CPython is sometimes described as "single threaded" due to the GIL,
but the interpreter will still "preemptively" switch between threads
(the details seem poorly documented).
So the time between checking whether acquire is 1 and decrementing the
count could result in more than one thread seeing an "available"
device, and more than one test being run (simultaneously, on the same
physical device!). We have a big herd of threads all polling for
this, so in a large test run this would happen maybe one time out of
20-30 tries.
Use a lock. Also remove the very similar looking
DUT.get_available_device() method, which had the same bug but appears
to be dead code.
Fixes#32679
Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
Use EXTRA_GEN_DEFINES_ARGS to error out on deprecated devicetree
properties when warnings are treated as errors.
Signed-off-by: Martí Bolívar <marti.bolivar@nordicsemi.no>
sometimes a test instance hasn't been boot up to execute successfully,
so the results is an empty dictionary, it will not be added into final
twister_report.xml file.
in order to include it, fill the results as BLOCK.
Signed-off-by: peng1 chen <peng1.chen@intel.com>
Currently duplicated testcases are only determined by the classname.
This results in testcases not beeing added to the twister.xml if the
testcases are from the same class (e.g. sample.yaml), but with different
names.
The changed check if the combination of classname and testname
already exists.
Signed-off-by: Sebastian Wezel <sebastian.wezel@nordicsemi.no>
Include coverage support for device testing. It works by switching
capture_coverage to True (as it would meet coverage data start
pattern). Then we continue tot read coverage data until we get
coverage data end pattern. Otherwise, after receiving test result
pattern, we close serial console and do not get time to find start
coverage data pattern.
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
Fix twister output. "Test suite" used previously did not correspond
to what was counted (test scenarios). A test scenario is an entry from
testcase/sample.yaml under "tests:", which specify configuration used to
build a test. Changes in docs will follow later on. Also replace the
amount of preselectd platforms with the amount of platforms that
were not filtered out and actually tested.
Signed-off-by: Maciej Perkowski <Maciej.Perkowski@nordicsemi.no>
Followup commit on comment in #32003.
Using pathlib to avoid mixing of `/` and `\` if twister someday will be
able to run on Windows.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
Fix possible race when determining available devices. Find available
device and return it in the same operation to avoid the race.
Fixes#31769
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Fixes: #30713
Now using the common script `verify-toolchain.cmake` for verifying
toolchain settings.
Signed-off-by: Torsten Rasmussen <Torsten.Rasmussen@nordicsemi.no>
As of today we unconditionally call 'stty sane' that leads to
lots of error messages
'stty: standard input: Inappropriate ioctl for device'
when output isn't connected to tty (for example when we launch
twister in Jenkins)
Fix that by check if output is really connected to tty before
'stty sane' call.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
This is always the same, so we might as well just remove it for now and
save some space since it is the default anyways. Type can be added later
when it provides more value.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Logic for scope of platforms to be checked was changed and this options
now only applies to limited scope, fix this and reset scope when a
testcase declares it needs to build on available platforms.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
The commit
531fe89e80 (sanitycheck: use multiprcoessing instead of threads)
introduce regression for ARC nsim_hs_smp platform.
The regression is that verification on nsim_hs_smp hangs. That
happens because now we don't call try_kill_process_by_pid()
in the 'with' scope of the 'subprocess.Popen' when we spawning
BinaryHandler thread. Previously it was called via `terminate`
method of `BinaryHandler` but it was changed in 531fe89e80.
So if we can't terminate the simulator in a normal way
(which is expected for 'mdb' which is used for nsim_hs_smp
simulation) we will hang forever - we will never return from
'with' scope of the 'subprocess.Popen' as we are waiting for
process termination but the try_kill_process_by_pid() is located
latter and we never reach it.
Fix that.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
With the earlier version of Git, git describe command will
use the default 7 hexadecimal digits as the abbreviated
object name. But currently the Linux kernel project needs
11 to 12 hexdigits to stay unique.So it's necessary to add
--abbrev=12 to ensure twister always generate the same tags.
In https://testing.zephyrproject.org/daily_tests/versions.json,
all tags have 10 hexadecimal digits.
Signed-off-by: Jingru Wang <jingru@synopsys.com>
Do not store shared memory variables into yaml and check for validity of
imported hardware map before attempting to modify it.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Add exception check for psutil when running CI with nSIM
Sometimes nSIM random failures may turn nsimdrv process into zombie
process, parent process can't terminate it by calling proc.wait(0.1)
and kill command doesn't apply to zombie process
Signed-off-by: Jingru Wang <jingru@synopsys.com>
Now we create plaform reports for each enabled platform, this is time
consuming, especially if you are not using those reports for anything.
Make this an option and enable only if the generated reports are going
to be used.
Individual reports can also be generated from the JSON file if needed.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Move the "module" classes used by sanitycheck into a new location under
pylib, where we can have other python libraries reside instead of having
them all over the place.
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
