On Mon, Sep 05, 2016 at 05:26:18PM +0200, Sean McGovern wrote: > > After committing the patch to bump RELEASE and the Changelog, it was > discovered that a some of the FATE nodes on https://fate.libav.org/v9/ had > not reported since late 2014 where they were all green. > > When they were re-enabled (thanks, Diego), some of the tests from vsynth* > and lavf* are now failing with different checksums. See > https://fate.libav.org/x86_32-linux-inline-asm-disabled-v9/20160904132041 > and https://fate.libav.org/x86_32-linux-mmx-disabled-v9/20160904134406 > > The common element between them is they are 32-bit builds (the 64-bit > equivalents are succeeding), but also note that the compiler has changed on > those nodes since when they ran in 2014. I tried building 9.17 and 9.16 > with the same compiler and they are also failing those tests, so possibly > adds some weight to it. > > My theory is that it is experiencing either some form of UB, or maybe this > is a sign-extension issue somewhere. > > Does anyone else have any thoughts on this?
The two failing configs did not run before as in 2014 I had only x86_64 instances on FATE. I'm just going to drop the two instances and blissfully pretend to be unaware of the failures. It is an unsupported branch after all and it's not like these would be the only known bugs... Diego _______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel
