Re: [Valgrind-developers] 3.10.1 release proposed for Tuesday 25 Nov

[Mark W. <mj...@re...>]

On Mon, 2014-11-24 at 15:41 +0100, Mark Wielaard wrote:
> On Mon, 2014-11-17 at 14:45 +0100, Julian Seward wrote:
> > There have been a number of bug fixes on trunk which it would be nice
> > to make available in the 3.10 line.  I therefore propose to release
> > 3.10.1 on Tuesday 25 Nov (8 days from now).
> 
> I incorporated valgrind-3.10.1-TEST1 into the fedora package and did
> some builds. Most look pretty good, same or better than 3.10.0, except
> on armv7hl.
> [...]
> armv7hl:
> http://koji.fedoraproject.org/koji/getfile?taskID=8218319&name=build.log
> == 558 tests, 117 stderr failures, 23 stdout failures, 2 stderrB
> failures, 5 stdoutB failures, 0 post failures ==
> (Too many to list, see build.log for full logs/diffs)
> 
> Somewhat worse than 3.10.0, but this is on rawhide with newer/beta
> glibc. All the new failures seem to be cause by:
> 
> Process terminating with default action of signal 11 (SIGSEGV)
>  Access not within mapped region at address 0xBD97FFF8
>    at 0x401BC7C: strcmp
> 
> I don't know what changed there (we should intercept strcmp on arm7hl).

Just for the record, I have been unable to replicate these bad results.
And the final armv7hl fedora rawhide results of 3.10.1 final are the
same as the original 3.10.0 fedora 21 results. So it is a mystery why
that one build showed that many strange failures.

Cheers,

Mark