Comment # 4 on bug 967412 from Michael Matz

(In reply to Takashi Iwai from comment #2)
> (In reply to Michael Matz from comment #0)
> > Whatever the GCC testsuite is doing, it of course shouldn't be able to crash
> > the kernel.  It seems this happened somewhen around 11th February, because 
> > a notification email tells me that that was the last time this package built.
> 
> AFAIK, there is no change in kernel around February 11.  The update of
> kernel was done in Feb. 01 to 4.4.0, then in Feb. 17 to 4.4.1.

It might always be that the build kernel used by obs was bumped only later.
Or that gcc5-testresults wasn't built until much later.  The job history
of that packages ends with:

2016-01-19 06:29:56 CET  gcc5-testresults  new build        succeeded          
 4h  8m 33s   morla1:24       
2016-01-29 08:56:52 CET  gcc5-testresults  rebuild counter  failed             
 2h 10m 11s   lamb13:4        
2016-01-30 19:28:55 CET  gcc5-testresults  new build        succeeded          
 3h  0m 27s   build74:3       
2016-02-01 01:27:31 CET  gcc5-testresults  rebuild counter  succeeded          
 3h 48m  8s   build84:1       
2016-02-04 00:59:59 CET  gcc5-testresults  new build        failed             
 9h 14m 24s   lamb05:5        
2016-02-05 10:12:28 CET  gcc5-testresults  new build        failed             
 9h 31m 21s   lamb04:7        
2016-02-08 19:10:18 CET  gcc5-testresults  source change    succeeded          
 4h 12m  4s   build72:2       
2016-02-09 21:41:47 CET  gcc5-testresults  new build        succeeded          
 3h 36m 11s   build72:3       
2016-02-12 08:18:16 CET  gcc5-testresults  new build        failed             
 2h 54m 40s   lamb14:2

So, it seems it's always failing on the lamb* hosts, but not on the others.
I dimly remember hearing something in the hallway or IRC about the lamb*
being new and somehow special (nvme?) ...

> If this
> started from Feb. 11, it must be something else that triggers the bug.  Was
> there a compiler update, or any other toolchain changes around that time?

... so it's possible that it's not triggered by a kernel update, but rather
the introduction of the lamb* hosts uncovered a latent kernel bug.

> Of course, this is a kernel bug, and the kernel has to be fixed.  But it's
> good to understand what causes the issue...


I don't know much more than the above :-/