Thanks for looking into it. Ed Maste <ema...@freebsd.org> writes:
> On 17 November 2014 at 05:34, Daiki Ueno <u...@gnu.org> wrote: >> >> As I'm getting more reports (perhaps) related to this, I've filed a bug >> at the FreeBSD bugzilla, with a reproducer written in plain pthreads: >> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=195098 > > To clarify, does this issue seem to occur randomly on some runs, > without changing anything?11 Yes. I compiled it with: $ cc -o test-once test-once-pthread.c -lpthread If it ran ok, it took ~13 sec: $ time ./test-once >/dev/null 13.01 real 4.10 user 8.90 sys Otherwise, I had to interrupt the process by ^C: $ time ./test-once >/dev/null ^Ctime: command terminated abnormally 288.40 real 67.36 user 220.85 sys > I've tried your reproduction code on my FreeBSD laptop and have not > been able to reproduce the failure; it runs in ~4 seconds each time. Actually, I haven't tried it on a bare metal, but on Assaf's freebsd10 build-ready VM image: http://www.nongnu.org/pretest/downloads/ running under kvm on GNU/Linux with the following command line: $ qemu-kvm -nographic -m 384 \ -snapshot \ -net user -net nic,model=virtio \ -drive if=virtio,media=disk,index=0,file=freebsd10.build-ready.qcow2 Regards, -- Daiki Ueno
