[PATCH i-g-t 1/3] igt_hook: Add feature

[Kamil Konieczny]

Hi Gustavo,
On 2024-05-16 at 09:19:26 -0300, Gustavo Sousa wrote:
> Quoting Kamil Konieczny (2024-05-16 07:40:58-03:00)
> >Hi Gustavo,
> >On 2024-05-15 at 14:35:55 -0300, Gustavo Sousa wrote:
> >> Quoting Kamil Konieczny (2024-05-15 14:10:55-03:00)
> >> >Hi Gustavo,
> >> >On 2024-05-09 at 12:24:29 -0300, Gustavo Sousa wrote:
> >> >> For development purposes, sometimes it is useful to have a way of
> >> >> running custom scripts at certain points of test executions. A
> >> >> real-world example I bumped into recently is to collect information from
> >> >> sysfs before and after running each entry of a testlist.
> >> >> 
> >> >> While it is possible for the user to handcraft a script that calls each
> >> >> test with the correct actions before and after execution, we can provide
> >> >> a better experience by adding built-in support for running hooks during
> >> >> test execution.
> >> >> 
> >> >> That would be even better when adding the same kind of support for
> >> >> igt_runner (which is done in an upcoming change), since the user can
> >> >> also nicely resume with igt_resume with the hook already setup in case a
> >> >> crash happens during execution of the test list.
> >> >> 
> >> >> As such provide implement support for hooks, integrate it into
> >> >> igt_core and expose the functionality via --hook CLI option on test
> >> >> executables.
> >> >
> >> >Hmm, why not just a pre-hook@ and post-hook@ in testlist itself?
> >> >It will be easier to handle - just more parsing.
> >> 
> >> How would that work with respect to filters? The current proposal allows
> >> something filtering the events to be tracked. For example, one can use
> >> `--hook "pre-test,pre-dyn-subtest:echo hello"` to run the command only
> >> before test binary starts and before each dynamic subtest.
> >> 
> >> Also, there are cases where a testlist is not really used. Examples are
> >> calling a test binary directly or calling igt_runner without
> >> --test-list. So, while I believe we could consider support for
> >> describing hooks in testlist, I do not think that would be a substitute
> >> for the --hook option.
> >> 
> >> On a personal note, my current use case for hooks is more towards
> >> debugging, so for me it is more convenient to have a --hook option than
> >> having to make a copy of a testlist only to add the hook instructions
> >> there.
> >> 
> >> --
> >> Gustavo Sousa
> >> 
> >> >
> >> >Added Petri to cc.
> >> >
> >> >Regards,
> >> >Kamil
> >> >
> >
> >Ok, that makes sense, I will look into your patches later (maybe next week).
> 
> Thanks!
> 
> >In meantime please look into GitLab failure here:
> >https://gitlab.freedesktop.org/gfx-ci/igt-ci-tags/-/jobs/58540606
> 
> Yeah, already tried to take a look at it a few days ago [1], but as I
> mentioned there, I'm not sure how I can get more info (e.g. logs) on the CI
> failure. The test works fine for me locally. I wonder if it is possible
> to setup the container locally so that I get the same test envionment.
> 
> [1] https://lore.kernel.org/all/171527425512.5772.1654105508452706087@gjsousa-mobl2/
> 
> --
> Gustavo Sousa
> 

Command used is in line 29:
meson test -C build --num-processes ${FDO_CI_CONCURRENT:-4}

and fail is:
13/30 lib igt_hook_integration FAIL 0.72 s (killed by signal 11 SIGSEGV)

I tested this locally after building (so binaries are already there)
using 'meson test -C build' and it fails:
killed by signal 7 SIGBUS

Regards,
Kamil

> >
> >Regards,
> >Kamil
> >