[LLVMdev] [llvm-lab-wg] FNT testers reporting success even though they failed

Galina Kistanova gkistanova at accesssoftek.com
Tue Dec 11 11:40:13 PST 2012

Hello everyone,

It might make sense to start from rebooting the buildmaster, just to make sure everything is all right on this end.
Yesterday I have tried to apply the latest changes from zorg and some of them are broken.
Theoretically, checkconfig shouldn't affect the working instance, but the reality could be different...
I planned to rollback to the last known-to-be-good revision and restart the master in the evening, but could do this sooner if you feel like this is urgent.

Please let me know.



-----Original Message-----
From: llvm-lab-wg-bounces at lists.minormatter.com [mailto:llvm-lab-wg-bounces at lists.minormatter.com] On Behalf Of David Blaikie
Sent: Tuesday, December 11, 2012 9:12 AM
To: Duncan Sands
Cc: llvm-lab-wg at lists.minormatter.com; Galina Kistanova; llvmdev at cs.uiuc.edu
Subject: Re: [llvm-lab-wg] FNT testers reporting success even though they failed

On Tue, Dec 11, 2012 at 1:27 AM, Duncan Sands <baldrick at free.fr> wrote:
> All the LLVM lab FNT builders were reporting failures in the same set of
> tests
> (scimark2, LivermooreLoops, some others).  Now they've all turned green but
> the
> same tests are still failing.  I don't see any commits to deliberately make
> the
> FNT pass, so it looks like someone broke failure detection...  Note that the
> FNT "report" text became empty at the same time.  Here's what "report" was
> like
> before:
> http://lab.llvm.org:8011/builders/clang-x86_64-debian-fnt/builds/12575/steps/make.test-suite/logs/report
> and after:
> http://lab.llvm.org:8011/builders/clang-x86_64-debian-fnt/builds/12576/steps/make.test-suite/logs/report
> The first green builds are:
>   http://lab.llvm.org:8011/builders/clang-x86_64-debian-fnt/builds/12576
>   http://lab.llvm.org:8011/builders/clang-x86_64-linux-fnt/builds/57
> http://lab.llvm.org:8011/builders/dragonegg-x86_64-linux-gcc-4.6-fnt/builds/1983
> Does anyone know what might have caused this?

[Thought it might've been my change (r169769) but that doesn't look
like it intersects with this particular test task]

I think you're on the right track with the logs. If the log that
reaches the buildmaster is empty then we won't find any failures &
we'll report success.

Have you checked any of the slaves to see if the log files are empty
in the builds themselves?

Have you run a local execution of the test suite uisng roughly the
same commands & observed whether the reports are empty coming out of

- David
llvm-lab-wg mailing list
llvm-lab-wg at lists.minormatter.com

More information about the llvm-dev mailing list