Re: buildbot is green!

["Carlos O'Donell" <carlos at redhat dot com>]

On 09/04/2015 06:44 PM, Roland McGrath wrote:
> For the first time, we have fully green buildbot runs for i686 and x86_64.
> (The power8 bot has several math failures.)

Huzzah! :-)

> With the elf/tst-protected1[ab] XFAILs in place, i686 went green.
> For x86_64, posix/tst-getcpu was failing.  That appears to have been
> the combination of the GCE virtualization and the 3.16 kernel making
> the vDSO sched_getcpu return bad results (always 0).  I installed a
> newer kernel, which no longer uses the RDTSCP instruction (which I
> suspect the virtualization is breaking) in the vDSO and now it's happy.
> 
> Please start getting into the habit of looking at the buildbot page.
> If you commit something and then the i686 or x86_64 buildbot goes
> red, there's probably something you need to fix ASAP.
> 
> We still need some more cooks in this soup.  Please check out the
> glibc-buildbot git repository and start poking around for ways to
> make things better.
> 
> One current nuisance I haven't figured out is that some of the runs
> of 'make check' end with:
> 	make[1]: *** [tests] Error 1
> 	Makefile:9: recipe for target 'check' failed
> 	make: *** [check] Error 2
> but no summary output.  Someone could look into why that's
> happening.  It might need a fix in the libc makefiles, or it might
> just need a fix in the shell script in the buildbot repo that runs
> the build.

That's weird. I'll save a log if I see it.

c.