Re: [Valgrind-developers] valgrind CI failure

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

Hi Martin,

Let me CC valgrind-developers, maybe someone else has already looked
into this.

On Thu, Jun 16, 2022 at 07:19:33AM +0200, Martin Liška wrote:
> Do we know what's wrong with:
> https://builder.sourceware.org/buildbot/#/builders/92/builds/28

This (the opensuse tumbleweed builder) and the fedora latest builder
both started failing when they upgraded to GCC 12. All other builders
are green (but none use GCC 12 yet):

https://builder.sourceware.org/buildbot/#/builders?tags=valgrind

The failure is in make auxchecks which is not part of the normal
regtests. The test builds the GNU Scientific Library gsl 1.6 and
compares a testrun under valgrind.

It is seeing the following failures under valgrind:

+FAIL: random32-bsd, 10000 steps (852261210 observed vs 1663114331 expected)
+FAIL: random64-bsd, 10000 steps (210970120 observed vs 864469165 expected)
+FAIL: random32-libc5, 10000 steps (367802360 observed vs 1967452027 expected)
+FAIL: random64-libc5, 10000 steps (221021662 observed vs 2106639801 expected)

I haven't looked into why that is. But with earlier GCC versions these
tests PASSed when run under valgrind.

Cheers,

Mark