Re: [PATCH v1] perf test: Add basic callgraph test to record testing

[Namhyung Kim]

On Thu, Jun 19, 2025 at 09:24:04AM +0100, James Clark wrote:
> 
> 
> On 19/06/2025 1:24 am, Ian Rogers wrote:
> > Give some basic perf record callgraph coverage.
> > 
> > Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
> > ---
> >   tools/perf/tests/shell/record.sh | 20 ++++++++++++++++++++
> >   1 file changed, 20 insertions(+)
> > 
> > diff --git a/tools/perf/tests/shell/record.sh b/tools/perf/tests/shell/record.sh
> > index 2022a4f739be..603fa6b79b5d 100755
> > --- a/tools/perf/tests/shell/record.sh
> > +++ b/tools/perf/tests/shell/record.sh
> > @@ -12,8 +12,10 @@ shelldir=$(dirname "$0")
> >   . "${shelldir}"/lib/perf_has_symbol.sh
> >   testsym="test_loop"
> > +testsym2="brstack"
> >   skip_test_missing_symbol ${testsym}
> > +skip_test_missing_symbol ${testsym2}
> >   err=0
> >   perfdata=$(mktemp /tmp/__perf_test.perf.data.XXXXX)
> > @@ -359,6 +361,23 @@ test_precise_max() {
> >     fi
> >   }
> > +test_callgraph() {
> > +  echo "Callgraph test"
> > +  if ! perf record -o "${perfdata}" -g perf test -w brstack
> 
> Looking at 2dac1f089 ("perf test: Fix 'perf script' tests on s390"), it
> looks like this won't work there. Although I'm not sure why the fix wasn't
> to change the default -g mode to dwarf on s390 if fp doesn't work at all.
> That wouldn't fix -e cpu-clock being required though.

Maybe we can add a new 'perf check callchain' subcommand to check the
relevant information and display and/or verify them properly.

Thanks,
Namhyung