GitLab

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Fix #11930, #11693

See CID #1223114

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Overly large frame sizes are commonly caused by broken streams.

Close #11245
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This patch fixes a problem that occurs at the start of playback of MPEG2
streams.  The symptom is that an initial good frame will display and
then the video appears to skip back a frame or two and some
macroblocking is observed. The stream then plays correctly. This is
really obvious when switching between SAPed multicast streams, where
you're jumping right in to a lot of movement and probably not starting
with an I frame.

The issue appears to be some code introduced way back in 2004 (commit
eaefb850) which resubmits data to the
video codec. The commit notes suggest it was to avoid dropping the first
I frame, but I see no obviously lost I frames after removing the code
with my test samples. Presumably this was a work-around for a libavcodec
issue that is no longer required.

Therefore, this patch basically just reverts the 2004 commit in its
current form.
Signed-off-by: Ilkka Ollakka <ileoo@videolan.org>

This should perhaps be const... but libavcodec seems to require
non-const even when const ought to be enough :-(

This reverts commit d8199d8c.

VLC has status reporting on the input state, but not on the output state.
This makes it impossible to know if playback actually started completely from
checking stream status states.

For instance a failed audio output cannot be determined programatically. This
patch adds an extra field per ES (in src/input/es_out.c) to determine the state
of that ES (DISABLE, ENABLED, ERROR).

If CreateDecoder() cannot instantiate an output, then it set p_dec->b_error = true.
This is used to determine what the return value for es_out_GetEsState() should be.
If p_dec->b_error is true, then it returns an ES_OUT_STATE_ES_ERROR to the caller.
Else it returns ES_OUT_STATE_ES_ENABLED or ES_OUT_STATE_ES_DISABLED.

A function 'es_out_GetEsState()' is available to query an input for the state of
the primary audio, video and SPU elementary streams.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>

Closes: #10331

With the introduction of CODEC_FLAG_OUTPUT_CORRUPT incomplete frames
are silently dropped. Enable it to maintain the previous behaviour.

Close #10140
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

And always pad it while at it.
Codecs always expect extradata to be aligned and padded.

Prevent memory corruption when memalign-hack is enabled.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Fix to actually conform to the range specified in doc.

First introduced in 9d20efbbSigned-off-by: Jean-Baptiste Kempf <jb@videolan.org>

There is no use to CPU memory edges when decoding in hardware (arguably
libavcodec should be "enhanced" to not add them in that case).
This unbreaks VDPAU.

This enables direct rendering for H.264 software decoding.

Such option is more suited to the avcodec-options field and should not be exposed to gui
Signed-off-by: Rafaël Carré <funman@videolan.org>

With the reference counting stuff, the AVFrame structure is gone or
unreachable by the time the release function is called.