[vlc-devel] [RFC PATCH] decoder: do not depend on allocated block for flush

[Steve Lhomme]

On Wed, Nov 18, 2015 at 7:22 PM, Rémi Denis-Courmont <remi at remlab.net> wrote:
> On Wednesday 18 November 2015 18:24:42 Thomas Guillem wrote:
>> With this patch, a flush won't fail because of a block_Alloc failure.
>
> I imagined a constant static block before, but gave up due to concerns about
> p_next or other block_t fields being potentially modified. Storing the flush
> block per decoder is an innovative trick to partly alleviates the problem. I
> am not certain that there are no corner cases where the block would end up
> being used more than once concurrently still.
>
> I think we would be better off with explicit flush callbacks.

With my pf_flush patch, the block is never actually pushed to the
decoder. There's one pushed for legacy compatibility though, ie
decoders/packetizers that don't provide their own pf_flush. It doesn't
have to be this internal flush block_t though. It's allocated
separately, and shouldn't have BLOCK_FLAG_CORE_FLUSH in the first
place.

> --
> Rémi Denis-Courmont
> http://www.remlab.net/
>
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel