Re: [Qemu-devel] [PATCH v2 2/2] tests: Disable qht-bench parallel test when using gprof

[Eduardo Habkost]

On Thu, Jan 10, 2019 at 04:25:17PM +0000, Alex Bennée wrote:
> 
> Philippe Mathieu-Daudé <address@hidden> writes:
> 
> > This test is failing on the Travis CI [*] since some time now,
> > disable it until it get fixed.
> >
> > [*] https://travis-ci.org/qemu/qemu/builds/474821674
> >
> > Signed-off-by: Philippe Mathieu-Daudé <address@hidden>
> > ---
> > v2: Use CONFIG_GPROF via config_host_mak (check-unit isn't target)
> > ---
> >  configure              | 1 +
> >  tests/Makefile.include | 3 ++-
> >  2 files changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/configure b/configure
> > index fa5c079f98..69a125814d 100755
> > --- a/configure
> > +++ b/configure
> > @@ -7470,6 +7470,7 @@ alpha)
> >  esac
> >
> >  if test "$gprof" = "yes" ; then
> > +  echo "CONFIG_GPROF=y" >> $config_host_mak
> >    echo "TARGET_GPROF=y" >> $config_target_mak
> >    if test "$target_linux_user" = "yes" ; then
> >      cflags="-p $cflags"
> > diff --git a/tests/Makefile.include b/tests/Makefile.include
> > index 3f5a1d0c30..b8f91fdd97 100644
> > --- a/tests/Makefile.include
> > +++ b/tests/Makefile.include
> > @@ -88,7 +88,8 @@ check-unit-y += tests/test-rcu-simpleq$(EXESUF)
> >  check-unit-y += tests/test-rcu-tailq$(EXESUF)
> >  check-unit-y += tests/test-qdist$(EXESUF)
> >  check-unit-y += tests/test-qht$(EXESUF)
> > -check-unit-y += tests/test-qht-par$(EXESUF)
> > +# FIXME: {test-qht-par + gprof} often break on Travis CI
> > +check-unit-$(call lnot,$(CONFIG_GPROF)) +=
> >  tests/test-qht-par$(EXESUF)
> 
> I got confused between test-qht-par and qht-bench. However we already
> have this in the test case:
> 
>     if (g_test_quick()) {
>         g_test_add_func("/qht/parallel/2threads-0%updates-1s", test_2th0u1s);
>         g_test_add_func("/qht/parallel/2threads-20%updates-1s", 
> test_2th20u1s);
>     } else {
>         g_test_add_func("/qht/parallel/2threads-0%updates-5s", test_2th0u5s);
>         g_test_add_func("/qht/parallel/2threads-20%updates-5s", 
> test_2th20u5s);
>     }
> 
> So maybe we should be asking why it isn't limiting itself to 1s?

We are, but we have no idea why it's happening.  This patch is a
workaround until we find the root cause of the problem.

If nobody objects, I will merge it through my tree.

-- 
Eduardo