[cfe-dev] buildbot failure in LLVM on clang-x64-ninja-win7

[Robinson, Paul via cfe-dev]

> -----Original Message-----
> From: cfe-dev [mailto:cfe-dev-bounces at lists.llvm.org] On Behalf Of
> Hahnfeld, Jonas via cfe-dev
> Sent: Wednesday, July 27, 2016 4:18 AM
> To: Renato Golin
> Cc: gkistanova at gmail.com; cfe-dev at lists.llvm.org; mstester.llvm at gmail.com
> Subject: Re: [cfe-dev] buildbot failure in LLVM on clang-x64-ninja-win7
> 
> > -----Original Message-----
> > From: Renato Golin [mailto:renato.golin at linaro.org]
> > Sent: Wednesday, July 27, 2016 12:56 PM
> > To: Hahnfeld, Jonas
> > Cc: cfe-dev at lists.llvm.org; gkistanova at gmail.com;
> mstester.llvm at gmail.com
> > Subject: Re: [cfe-dev] buildbot failure in LLVM on clang-x64-ninja-win7
> >
> > On 27 July 2016 at 11:36, Hahnfeld, Jonas via cfe-dev <cfe-
> > dev at lists.llvm.org> wrote:
> > > Huh, Clang::Rmodule-build.m failed for no obvious reason and is green
> > > again for the next build:
> > > http://lab.llvm.org:8011/builders/clang-x64-ninja-win7/builds/14221
> > >
> > > Has anyone noticed this test in the past?
> >
> > A few pages of waterfall have shown no signs of R tests randomly
> failing...
> > This is weird.
> >
> > I'd let it be, for now. Both Galina and Denis are copied. If it happens
> > again,
> > they can investigate, as it doesn't seem related to your commit in any
> way.
> >
> > cheers,
> > --renato
> 
> [Copying also Reid]
> 
> http://lab.llvm.org:8011/builders/clang-x86-win2008-selfhost/builds/9307
> was
> red,
> http://lab.llvm.org:8011/builders/clang-x86-win2008-selfhost/builds/9308
> will
> probably be green again (has already passed check 1 which failed before)
> 
> The failing tests (which are different) both show:
> remark: building module '...' as '...'
> which triggers a CHECK-NOT and "error: 'remark' diagnostics seen but not
> expected:"
> 
> Thanks,
> Jonas


We put up a new internal Windows bot on Monday, and it has been suffering
a lot of transient failures in Modules and PCH.  On the other hand,
llvm-clang-lld-x86_64-scei-ps4-windows10pro-fast has not been seeing these
(so far as I've noticed) so I wonder if there's some environmental thing
going on.  I had been going on the assumption that there was an internal
problematic patch, but if folks upstream are also seeing flakiness in these
tests then maybe something else is at the root of it.

Our new bot is using 32 threads; are these tests unusually susceptible
to problems with parallel execution?

FTR the list of tests I've seen with this behavior:

Modules/fmodules-validate-once-per-build-session.c	3
Modules/no-stale-modtime.m		13
Modules/Rmodule-build.m			3
Modules/validate-system-headers.m	3
PCH/headersearch.cpp			3
PCH/verify_pch.m			6

--paulr