[vlc-devel] Re: vlc: svn commit r19076 (sigmunau)

[Rémi Denis-Courmont]

Le jeudi 1 mars 2007 19:42, Christophe Massiot a écrit :
> This is not entirely true - if you look closely at the source,
> Sigmund's version has an extra condition before calling
> clock_nanosleep :

Thanks for explaining my own code to me :)

The point is, it is (at least theoretically) incorrect to wait for a 
period of time, since it could be that the threads gets yield *between* 
mdate() and mwait(), in which case the delay value will be too big.

> However, commenting out the if() doesn't reintroduce the problem,

Maybe it does not, but I surely wouldn't risk waiting for a negative 
amount of time, while it is theoretically safe to wait for a past date.
But of course, if mdate() needs to be called, this is a non-issue.

> I am not as familiar as you with the clock_nanosleep behaviour, but I
> can only suspect a kernel or libc bug at this point.

At some point in the past, my delay division was buggy. But that seems 
fine now. Still, I have not been able to build any testcase showing a 
problem in libc/kernel.

-- 
Rémi Denis-Courmont
http://www.remlab.net/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20070301/b8045806/attachment.sig>