[FFmpeg-devel] [PATCH] huffyuv: assign correct per-thread avctx pointer to HYuvContext::avctx.

[Ronald S. Bultje]

Hi,

On Wed, Apr 5, 2017 at 3:29 PM, Michael Niedermayer <michael at niedermayer.cc>
wrote:

> On Mon, Apr 03, 2017 at 10:28:33PM -0400, Ronald S. Bultje wrote:
> > Fixes the following tsan warning when running fate-vsynth_lena-ffvhuff:
> >
> > WARNING: ThreadSanitizer: data race (pid=6484)
> >   Write of size 8 at 0x7d64000154b8 by main thread (mutexes: write
> M1331):
> >     #0 update_context_from_user src/libavcodec/pthread_frame.c:331
> (ffmpeg+0x000000dca887)
> > [..]
> >   Previous read of size 8 at 0x7d64000154b8 by thread T2 (mutexes: write
> M1334):
> >     #0 draw_slice src/libavcodec/huffyuvdec.c:857
> (ffmpeg+0x000000bcc86f)
> > ---
> >  libavcodec/huffyuvdec.c | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/libavcodec/huffyuvdec.c b/libavcodec/huffyuvdec.c
> > index 5572b98..979c4b9 100644
> > --- a/libavcodec/huffyuvdec.c
> > +++ b/libavcodec/huffyuvdec.c
> > @@ -579,6 +579,8 @@ static av_cold int decode_init_thread_copy(AVCodecContext
> *avctx)
> >      HYuvContext *s = avctx->priv_data;
> >      int i, ret;
> >
> > +    s->avctx = avctx;
>
> LGTM
>
> do you think its better in decode_init_thread_copy() or decode_frame() ?
> I think i would have instinctivly put it in decode_frame() but i dont
> see a issue with decode_init_thread_copy()


I see _init() as logically more correct since this data is supposed to be
stable across multiple (all) frames. But in practice I don't think it would
matter.

Ronald