Hi Hirofumi, Thank you for the bug report!
On Sat, Mar 19, 2016 at 11:06 PM, OGAWA Hirofumi <hirof...@mail.parknet.co.jp> wrote: > > When closing wish, process die at > > pthread_cond_wait(¬ifierCV, ¬ifierMutex); > > if cpu is supporting HLE/RTM. (xend instruction doesn't allow to use without > xbegin) > > Tcl_FinalizeNotifier() is using following combination > > pthread_mutex_lock(¬ifierInitMutex); > pthread_cond_wait(¬ifierCV, ¬ifierMutex); > pthread_mutex_unlock(¬ifierInitMutex); > > From history, this is clearly conversion mistake. notifierInitMutex should be > notifierMutex (no Init). I'm not sure that the correct way to fix this is to replace notifierInitMutex by notifierMutex. The notifierInitMutex was introduced recently, and I think it was intentional. So, maybe to fix this we should go the other way: replace notifierMutex by notifierInitMutex in pthread_cond_wait? Could you try and comment on this as well? I don't have access to hardware with HLE/RTM extensions, so can't test myself. Cheers! -- Sergei Golovan