On Sat, Mar 10, 2012 at 06:40:48PM +0100, Torbjorn Granlund wrote: > Matthieu Herrb <matth...@herrb.eu> writes: > > Since I couldn't reproduce test errors on my local machine running the > development version of OpenBSD 5.1, I'll probably ugrade gcc64 again > to what is going to be OpenBSD 5.1. (even if not formally released, > the binary sets that are going to make the final release are already > available). > > This is hardly gcc problem. It would be very surprising if it is not a > kernel problem. > > OK, it could be a hardware problem, but the exact same thing happens on > a different OpenBSD 5 system I've tried. It could be a C library > problem, but the intermittence speaks against that. > > Last, it could be bugs in the failing software. I did a lot of code > staring on the relevant GMP code, and feel very confident it is correct. > That particular code has not changed for about 10 years, and has been > used on many systems without problems, including OpenBSD prior to 5.0. > > Since others now report the same type of problems with other software, > the likelihood that this ia an OpenBSD problem is very high.
I upgraded the machine to 5.1, and it seems the problem is still there. I can now also reproduce it on my machine, by loading it while running the tests. Mine has 2 CPUs, so when othewise idle it goes through the tests without failure most of the time. That re-inforces the hypothesis of a register/save restore issue in the OpenBSD/sparc64 kernel. I've reported the issue to OpenBSD developpers. -- Matthieu Herrb _______________________________________________ Gcc-cfarm-users mailing list Gcc-cfarm-users@gna.org https://mail.gna.org/listinfo/gcc-cfarm-users
