Hi,
On Wed, Apr 20, 2011 at 7:45 PM, Anton Khirnov <[email protected]> wrote:
> From: Michael Niedermayer <[email protected]>
>
> Fixes issue2632 if interl=1 is used or the automatic interlace detection is
> enabled
> and works. This has the advantage compared to the patch in issue2632 that it
> causes
> no speed loss and it also works when scaling is used. The disadvantage is that
> interlacing autodetection does not yet work very well it seems.
> This is the same method mplayer uses
>
> Signed-off-by: Michael Niedermayer <[email protected]>
> Signed-off-by: Anton Khirnov <[email protected]>
> ---
> libavfilter/vf_scale.c | 59 ++++++++++++++++++++++++++++++++++++++---------
> 1 files changed, 47 insertions(+), 12 deletions(-)
Unfortunately, this patch is not fine. Poor Stefano's pet project is
indeed being uglified by an idiot that refuses patch review.
> struct SwsContext *sws; ///< software scaler context
> + struct SwsContext *isws[2]; ///< software scaler context for interlaced
> material
Why 2? Why doesn't non-interlaced sws not just use isws[0]?
> + if(strstr(args,"interl=1")){
> + scale->interlaced=1;
> + }else if(strstr(args,"interl=-1"))
> + scale->interlaced=-1;
These checks are insufficient and underdocumented. The -1 is
supposedly autodetect but is unimplemented.
> sws_freeContext(scale->sws);
> + sws_freeContext(scale->isws[0]);
> + sws_freeContext(scale->isws[1]);
> scale->sws = NULL;
Missing =NULLs.
> scale->sws = sws_getContext(inlink ->w, inlink ->h, inlink ->format,
> outlink->w, outlink->h, outlink->format,
> scale->flags, NULL, NULL, NULL);
> + scale->isws[0] = sws_getContext(inlink ->w, inlink ->h/2, inlink
> ->format,
> + outlink->w, outlink->h/2,
> outlink->format,
> + scale->flags, NULL, NULL, NULL);
> + scale->isws[1] = sws_getContext(inlink ->w, inlink ->h/2, inlink
> ->format,
> + outlink->w, outlink->h/2,
> outlink->format,
> + scale->flags, NULL, NULL, NULL);
> if (!scale->sws)
> return AVERROR(EINVAL);
|| !isws[0/1]
> + if(scale->interlaced>0 || (scale->interlaced<0 &&
> link->cur_buf->video->interlaced)){
> + av_assert0(y%4 == 0);
> + out_h = scale_slice(link, scale->isws[0], y, (h+1)/2, 2, 0);
> + out_h+= scale_slice(link, scale->isws[1], y, h /2, 2, 1);
> + }else{
This assert is wrong and allows crashing libavfilter.
Please drop the patch, hopefully Stefano will fix it up so we can
integrate a squashed version into libav, or just leave it out.
Ronald
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
