Re: [Valgrind-developers] Valgrind-3.24.0.RC1 is available for testing

[Mark W. <ma...@kl...>]

Hi Carl, Hi Paul,

On Tue, Oct 29, 2024 at 08:07:52AM -0700, Carl Love via Valgrind-developers wrote:
> On 10/29/24 12:38 AM, Paul Floyd via Valgrind-developers wrote:
> >On 28-10-24 19:52, Carl Love via Valgrind-developers wrote:
> >>PowerPC testing on Power 8, I am seeing 4 additional stderr
> >>failures versus Valgrind 3.23.0.RC2
> >
> >>none/tests/fdleak_cmsg_supp (stderr)          New for
> >>Valgrind-3.24.0.RC1
> >>none/tests/fdleak_cmsg_xml (stderr)             New for
> >>Valgrind-3.24.0.RC1
> >>none/tests/fdleak_ipv4_xml (stderr)               New for
> >>Valgrind-3.24.0.RC1
> >>none/tests/socket_close_xml              (stderr)            New
> >>for Valgrind-3.24.0.RC1
> >
> >It's probably not too serious. The fdleak feature tends to output
> >messages that are highly system dependent and need plenty of
> >filtering in order to have stable regtest results.
> 
> Thanks for the feedback.  The tests were run on various lab machines
> at IBM.  I will note that the nightly regression runs on the GCC
> pool machines gcc110 and gcc112 do not have the same failures.

I could finally replicate these issues on a Power9, Fedora 40 setup.
The issue is that in some cases the "main" frame is missing. I haven't
been able to figure out why though. We did see this before with some
other (fdleak) tests. Which were worked around in commit 04d30049bf9b
"Filter away "main" differences in filter_fdleak"

So for now I have added a few more new exp files (and two new supps).

commit e6960c2e41b103ab8d393cbe13dc6473fb89bffc
Author: Mark Wielaard <ma...@kl...>
Date:   Sat Nov 23 21:28:13 2024 +0100

    Add exp and supp patterns for missing main frame for ppc64le
    
    In some cases on ppc64le we are missing the main frame.
    Add alternative .exp-ppc64le variants for socket_close_xml,
    fdleak_cmsg_xml and fdleak_ipv4_xml. And extra suppressions
    without a main frame for fdleak_cmsg_supp.
    
    See also commit 04d30049b "Filter away "main" differences in filter_fdleak"

With that commit I get a clean make regtest:

== 768 tests, 0 stderr failures, 0 stdout failures, 0 stderrB failures, 0 stdoutB failures, 0 post failures ==

But I also filed:
https://bugs.kde.org/show_bug.cgi?id=496620
Main frame missing in various (fdleak) tests (ppc64le)

Cheers,

Mark