Re: nptl 0.30

["Alexander Terekhov" <terekhov web de>]

"dada1" <dada1 cosmosbay com> schrieb am 18.03.03 13:55:32:
> 
> I beg to differ Ulrich .
> 
> May be Alexander was refering to the known futex_wake() with locked mutex
> symdrom :

I was referring to the pthread_cond_signal/broadcast(), mostly.

> 
> File sysdeps/pthread/pthread_cond_wait.c, line 61
> 
> lll_futex_wake (futex, INT_MAX);
> /* We are done.  */
> lll_mutex_unlock (cbuffer->cond->__data.__lock);
> 
> 
> That means that the awaken thread(s) will hit the locked mutex.
> 
> Just invert those lines to :
> 
> lll_mutex_unlock (cbuffer->cond->__data.__lock); /* much better to call
> futex_wake() with the mutex unlocked */
> lll_futex_wake (futex, INT_MAX);

I think that it's sufficient to wake just one single thread 
here (in cleanup() for async.cancel region "wrapping" futex 
wait(); I guess).

regards,
alexander.

______________________________________________________________________________
Werden Sie kreativ! Jetzt HTML-Mails nicht nur schreiben - nein -
GESTALTEN, bei WEB.DE FreeMail! http://freemail.web.de/features/?mc=021141