There are several reasons for doing that:
1. It documents the code for the reader and helps find
inconsistencies and bugs.
2. For rej_perms, it guarantees the change will be done
even if the output reference can be created by several
code paths.
3. It can be used to predict cases where a copy will,
or will not happen and optimize buffer allocation
(for example not request a rare direct-rendering buffer
from a device sink if it will be copied anyway).
Note that a filter is still allowed to manage the permissions
on its own without using these fields.
The resolution is in the packets, so decoding must happen.
Since most other formats do not set the dimension, make it
a special case for PGS. If other codecs were to have the
same requirement, using a CODEC_CAP would be preferred.
DVD subtitles packets can only encode a single rectangle:
if there are several, copy them into a big transparent one.
DVD subtitles rely on an external 16-colors palette:
use a reasonable default one, stored in the private context,
and encode it into the extradata, as specified by Matroska.
TODO: allow to change the palette with an option.
Each packet can use four colors out of the global palette.
The old logic was to map transparent colors to the color 0
and all other colors to 3, 2, 1, cyclically in descending
frequency order, completely disregarding the original color.
Select the "best" four colors from the global palette, according
to heuristics based on frequency, opacity and brightness, and
arrange them in standard DVD order: background, foreground,
outline, other.
TODO: select the alpha value more finely; see if CHG_COLCON can
allow more than 4 colors per packet.
Reference:
http://dvd.sourceforge.net/dvdinfo/spu.html
With these changes, dvdsubenc can be used to transcode DVB subtitles
and get a very decent result.
The old code generates a termination packet with the same regions as the
start packet and page_state set to "only what changed"; the result is
that the termination packet is decoded as identical to the start packet.
The new code does as found in some DVB broadcasts: produce a packet with
no regions. This is done by expecting num_rects to be 0 rather than
using a flip-flop. ffmpeg.c is updated accordingly.
With this feature, it becomes possible to perform commonly
requested tasks, such as hardcoding bitmap subtitles.
This will be reverted once libavfilter has proper support
for subtitles. All the changes have the string "sub2video"
in them, it makes it easy to spot the parts.
The code currently use cur_buf as the target of the copy,
but cur_buf can be cleared by the filter if it has given
the reference away or stored it elsewhere as soon as start_frame.
The code still relies on the fact that the reference is not
destroyed until end_frame. All filters currently follow that condition.
An av_assert1() is added to check it; it should at least cause
very visible errors in valgrind.
The boilerplate states that the files are under LGPL,
but refer the user to the GPL at one place.
These files were (re)implemented specifically for FFmpeg.
There is no need to start_frame immediately on the output link
since the rest is only done with the last frame of the tile.
link->out_buf is now automatically dereferenced; since we give
it away it must be cleared.
Fix an assert failure; found by Clément Bœsch.
Commit 2bf0982 introduced an overflow check in loco_decode_plane,
but the error code is never taken into account, leading to
completely idiotic return values.
samples_size and samples_idx are supposed to be multiple of
channels at all time. If they are, the division is exact;
if they are not, something is very wrong in the code.
Add "data" and "extradata" attributes to the schema.
Add "tag" element in "streams".
Print extradata before tags to avoid closing the element.
Fix trac ticket #1588.
The values are listed if setting them fails.
Using "-profile help" or "-profile list" have that effect.
Similar to 3aba391.
Suggested by "rogerdpack" in trac ticket #1529.
The voice register functions return the same voice structure
upon multiple registration. It causes us two problems:
If we delete a voice without deregistering it, it leaves
a dangling pointer inside the library.
If we delete or unregister a voice at uninit, it may still
be in use by another instance of the filter.
The second problem is solved by keeping an usage counter inside
asrc_flite. This is not thread-safe, but neither is flite itself.
Filters can change the timings, so input files must not
necessarily be read at the same rythm. This patch select
the input file to read based on the timestamp at output
instead of input. With complex filter graphs, finding the
input for a given output is done by making a request and
checking to what buffer source it has been forwarded.
It adds the following header to the HTML files:
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
which helps some browsers to display correctly the few non-ASCII
characters in the doc.
Reap buffers stored in the buffer sinks even when
the avfilter_graph_request_oldest() loop only returns EOF.
avfilter_graph_request_oldest() can cause frames to arrive
to the sinks even when it returns EOF.
If skip_samples is set and timestamps are synthesized using durations,
make them start at -skip_samples (rescaled) instead of 0,
so that the timestamp of the first undiscarded sample is 0.
Nicolas George