https://bugs.kde.org/show_bug.cgi?id=445743
--- Comment #11 from Paul Floyd <pjfl...@wanadoo.fr> ---
I think that I understand the problem, at least on Linux.
PTHREAD_PRIO_INHERIT causes the type of mutex to change so that it will return
ERESTARTNOINTR if interrupted.
I see, with strace
14177 <... futex resumed> ) = ? ERESTARTNOINTR (To be restarted)
Standalone I see
20015 rt_sigreturn({mask=[]}) = 202
20015 futex(0x7fffffffd070, FUTEX_LOCK_PI_PRIVATE, 1, NULL <unfinished ...>
20014 <... nanosleep resumed> NULL) = 0
The problem is that the Valgrind sigreturn mechanism hijacks the return address
so we no longer return to the guest futex syscall. That means that the futex
doesn't get restarted.
Even worse, since the futex was supposed to be restarted the return code to
userland is zero, so we don't even see that it failed.
I don't see any way to fix this at the moment on Linux.
--
You are receiving this mail because:
You are watching all bug changes.
