[FFmpeg-devel] [PATCH] lavf/matroskadec: fix is_keyframe for early Blocks
Chris Cunningham
chcunningham at chromium.org
Tue Jan 31 22:02:01 EET 2017
On Tue, Jan 31, 2017 at 1:07 AM, wm4 <nfxjfg at googlemail.com> wrote:
> On Tue, 31 Jan 2017 09:57:24 +0100
> wm4 <nfxjfg at googlemail.com> wrote:
>
> > On Mon, 30 Jan 2017 17:05:49 -0800
> > Chris Cunningham <chcunningham at chromium.org> wrote:
> >
> > > Blocks are marked as key frames whenever the "reference" field is
> > > zero. This is incorrect for non-keyframe Blocks that take a refernce
> > > on a keyframe at time zero.
> > >
> > > Now using -1 to denote "no reference".
> > >
> > > Reported to chromium at http://crbug.com/497889 (contains sample)
> > > ---
> > > libavformat/matroskadec.c | 9 ++++++---
> > > 1 file changed, 6 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/libavformat/matroskadec.c b/libavformat/matroskadec.c
> > > index e6737a70b2..0d033b574c 100644
> > > --- a/libavformat/matroskadec.c
> > > +++ b/libavformat/matroskadec.c
> > > @@ -89,6 +89,7 @@ typedef const struct EbmlSyntax {
> > > int list_elem_size;
> > > int data_offset;
> > > union {
> > > + int64_t i;
> > > uint64_t u;
> > > double f;
> > > const char *s;
> > > @@ -696,7 +697,7 @@ static const EbmlSyntax matroska_blockgroup[] = {
> > > { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0,
> offsetof(MatroskaBlock, bin) },
> > > { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0,
> offsetof(MatroskaBlock, duration) },
> > > { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0,
> offsetof(MatroskaBlock, discard_padding) },
> > > - { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0,
> offsetof(MatroskaBlock, reference) },
> > > + { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0,
> offsetof(MatroskaBlock, reference), { .i = -1 } },
> > > { MATROSKA_ID_CODECSTATE, EBML_NONE },
> > > { 1, EBML_UINT, 0,
> offsetof(MatroskaBlock, non_simple), { .u = 1 } },
> > > { 0 }
> > > @@ -1071,6 +1072,8 @@ static int ebml_parse_nest(MatroskaDemuxContext
> *matroska, EbmlSyntax *syntax,
> > >
> > > for (i = 0; syntax[i].id; i++)
> > > switch (syntax[i].type) {
> > > + case EBML_SINT:
> > > + *(int64_t *) ((char *) data + syntax[i].data_offset) =
> syntax[i].def.i;
> > > case EBML_UINT:
> >
> > Isn't there a break missing?
> >
> > > *(uint64_t *) ((char *) data + syntax[i].data_offset) =
> syntax[i].def.u;
> > > break;
> > > @@ -3361,7 +3364,7 @@ static int matroska_parse_cluster_incremental(MatroskaDemuxContext
> *matroska)
> > > matroska->current_cluster_num_blocks = blocks_list->nb_elem;
> > > i = blocks_list->nb_elem -
> 1;
> > > if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
> > > - int is_keyframe = blocks[i].non_simple ?
> !blocks[i].reference : -1;
> > > + int is_keyframe = blocks[i].non_simple ?
> blocks[i].reference == -1 : -1;
> > > uint8_t* additional = blocks[i].additional.size > 0 ?
> > > blocks[i].additional.data : NULL;
> > > if (!blocks[i].non_simple)
> > > @@ -3399,7 +3402,7 @@ static int matroska_parse_cluster(MatroskaDemuxContext
> *matroska)
> > > blocks = blocks_list->elem;
> > > for (i = 0; i < blocks_list->nb_elem; i++)
> > > if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
> > > - int is_keyframe = blocks[i].non_simple ?
> !blocks[i].reference : -1;
> > > + int is_keyframe = blocks[i].non_simple ?
> blocks[i].reference == -1 : -1;
> > > res = matroska_parse_block(matroska, blocks[i].bin.data,
> > > blocks[i].bin.size,
> blocks[i].bin.pos,
> > > cluster.timecode,
> blocks[i].duration,
> >
> > I don't quite trust this. The file has negative block references too
> > (what do they even mean?). E.g. one block uses "-123". This doesn't
> > make much sense to me, and at the very least it means -1 is not a safe
> > dummy value (because negative values don't mean non-keyframe according
> > to your patch, while -1 as exception does).
> >
> > The oldest/most used (until recently at least) mkv demuxer, Haali
> > actually does every block reference element as a non-keyframe:
> >
> > http://git.1f0.de/gitweb?p=ffmpeg.git;a=blob;f=
> libavformat/MatroskaParser.c;h=173c2e1c20da59d4cf0b501639c470
> 331cd4515f;hb=HEAD#l2354
> >
> > This seems much safer.
> >
> > Do you have any insight why the file contains such erratic seeming
> > reference values? I'm sure I'm missing something. Or is it a broken
> > muxer/broken file?
>
> Oh, nevermind. The values in the reference elements are
> supposed to be _relative_ timestamps. This means -1 is still not a safe
> dummy value. But then, what is a value of "0" supposed to mean?
>
> Going after Haali seems the safest fix, as it most likely won't break
> anything.
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
Thanks for taking a look.
Definitely missing a "break;" - will fix in subsequent patch.
Agree timestamps should be relative (didn't realize this). Vignesh points
out that "0" in the test file is due to a bug in ffmpeg (and probably other
muxers) where this value is not written as a relative timestamp, but
instead as the timestamp of the previous frame. https://github.com/FFmp
eg/FFmpeg/blob/master/libavformat/matroskaenc.c#L2053
<https://www.google.com/url?q=https%3A%2F%2Fgithub.com%2FFFmpeg%2FFFmpeg%2Fblob%2Fmaster%2Flibavformat%2Fmatroskaenc.c%23L2053&sa=D&sntz=1&usg=AFQjCNGs8m6GsWbhTvCZl0Q_juGAldQblA>
.
Anyway, I'm all for following Haali. Its not obvious how best to do this. I
don't think there's any great default value to indicate "not-set" and the
generic embl parsing code that reads the reference timestamp doesn't really
lend itself to setting an additional field like
MatroskaBlock.has_reference. I can sort this out, but I'll pause in case
you have a recommendation in-mind.
More information about the ffmpeg-devel
mailing list