Re: [PATCH v4 5/7] perf test: Introduce storing logs for shell tests

[Jakub Brnak <jbrnak@redhat.com>]

On Tue, Sep 30, 2025 at 12:00:19PM -0700, Ian Rogers wrote:
> On Tue, Sep 30, 2025 at 9:09 AM Jakub Brnak <jbrnak@redhat.com> wrote:
> >
> > From: Veronika Molnarova <vmolnaro@redhat.com>
> >
> > Create temporary directories for storing log files for shell tests
> > that could help while debugging. The log files are necessary for
> > perftool testsuite test cases also. If the variable PERFTEST_KEEP_LOGS
> > is set keep the logs, else delete them.
> 
> I think log files for tests are a good idea. We run tests in parallel
> but read the stdout/stderr sequentially. For a failing test we may
> fail to read stdout/stderr in some situations as the process has long
> since died. Rather than add log files specifically for shell tests
> perhaps update start_test to pass the file descriptors to the child
> test processes and then clean up in finish_test?
> https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/tests/builtin-test.c?h=perf-tools-next#n528
> 
> Thanks,
> Ian

Hi Ian, 

Thanks for the feedback.

You've raised a good point about capturing stdout/stderr for parallel runs,
but this patch is solving a different problem.

The logs we're storing here are test-generated data artifacts, not the
stdout/stderr of the test process itself.

Our patch provides a $PERFSUITE_RUN_DIR environment variable so that commands inside the shell script (e.g., perf record)
have a known directory to save these files for debugging.

Thanks,
Jakub