[vlc-devel] [PATCH] Avoid deadlock on stop() using UDP on Android

[Romain Vimont]

Hi Rémi,

Thank you for your comments.

My explanation about pthread_cancel() was inexact and irrelevant for
this bug:
https://mailman.videolan.org/pipermail/vlc-devel/2014-May/098009.html

The problem resides in the UDP access. Therefore, as you suggested, I
propose a new patch (having a more appropriate title and commit message)
which does not touch the FIFO code:
https://mailman.videolan.org/pipermail/vlc-devel/2014-May/098020.html

Thank you for your feedbacks

--
®om

Le dimanche 11 mai 2014 à  4:37 +0300, Rémi Denis-Courmont a écrit :
> Le 2014-05-09 18:58, Romain Vimont a écrit :
> >Android has no pthread_cancel(). Its behavior is emulated in
> >vlc_cancel() using a flag "killed", but it can lead to unavoidable race
> >conditions, typically between a call to vlc_object_alive() and a
> >blocking cond_wait().
> 
> I don't see that. But if there is a bug in the Android port, fix the Android
> port.
> 
> I refuse to hack the UDP access unless there is a bug in the UDP access. If
> there is a bug in the UDP access, then it needs to be described correctly,
> and there is no need to touch the FIFO code.
> 
> Either way, this patch looks wrong.
> 
> -- 
> Rémi Denis-Courmont
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel