In 81f2549, output pixel format was changed from rgb24 to planar yuv,
but the initialization was left unchanged. As a result, the chroma
planes were left uninitalized. This was not noticed since the current
ff_get_video_buffer() seems to be initializing the buffer. This won't be
the case anymore after the Evil Plan.
This prevents trying to do some subtitles conversion for each event when
the character encoding is not found. It now aborts early instead of
flooding stderr.
It seems to break a lot on MacOS because of a mixup between multiple
iconv libraries. The issue is that the configure check link with no
particular library path (so it uses /usr/lib/libiconv.dylib where all
the symbols are defined). OTOH, the final build might link with extra
library paths, such as the ports lib directory (typically you get
-L/opt/local/lib because of an extra library such as SDL). Whatever this
option position (before or after -liconv), it will try to link with
/opt/local/lib/libiconv.dylib, and the issue is that this version has
the same symbols as the ones in /usr/lib/libiconv.dylib except that they
all start with 'lib' (libiconv_open, libiconv_close, ...). I don't plan
to try to workaround this mess, so any better solution is welcome.
Sometimes the muxer modifies the packet, like for instance lavf/mp3enc
changing pkt->destruct in order to keep a copy. These changes must be
kept, even though the muxer behaviour is questionable. Regression since
0072116.
Fixes#2124.
This was supposed to be removed in 565e4993, but it was delayed due to
the port of the remaining filters. These buffer ref pointers are not
used anymore, thus dropped.
The function find_things() in configure is confused by component
registration calls as part of multiline macros defining combined
component registration. Coalesce those macros into one line to
work around the issue.
Signed-off-by: Diego Biurrun <diego@biurrun.de>
Signed-off-by: Martin Storsjö <martin@martin.st>
The data does not contain timing or trailing line breaks anymore. In
addition to being less idiotic, it is consistent with other codecs and
thus allows more switches between formats and codecs. It also fixes the
issue of the trailing line returns being simple \n instead of CRLF in
the ASS rectangle dialogue (this is the reason of the FATE update).
"que" sounds like a slang word to me. This commit renames a few
variables, fix the comments and the logging messages (sometimes along
with small other typo fixes).
Current MicroDVD AVPackets contain timing information and trailing line
breaks. The data is now only composed of the markup data. Doing this
consistently between text subtitles decoders allows to use different
codec for various formats. For instance, MicroDVD markup is sometimes
found in some VPlayer files. Also, generally speaking, the subtitles
text decoders have no use of these timings (and they must not use them
since it would break any user timing adjustment).
Technically, this is a major ABI break. In practice, a mismatching
lavf/lavc will now error out for MicroDVD decoding. Supporting both
formats requires unnecessary complex and fragile code.
FATE needs update because line breaks in the ASS file were "\n" (because
that's what is used in the original file). ASS format expect "\r\n" line
breaks; this commit fixes this issue. Also note that this "\r\n"
trailing need to be moved at some point from the decoders to the ASS
muxer.
Note that the linebreaks text codec option (but not the feature) has
been removed; its main goal was to allow demuxers to configure the text
decoder (and not meant to be used by users), but the AVOption are not a
viable solution. This is solved differently in this commit.
This commit also makes sure the extradata and subtitle_header are NUL
terminated, without taking into account the trailing '\0' in account in
the size.
At the same time, it should fix 'warning: dereferencing type-punned
pointer will break strict-aliasing rules' warning for compilers who
don't consider uint8_t** and char** compatibles.
The SRT muxer is reponsible for separating events with two line breaks,
there is no need to add more than necessary. Similarly, other muxers
(such as Matroska) are not supposed to add line breaks at the end of the
payload.
This is consistent with the other ff_ass_split_* functions.
It also fixes a crash when trying to split a dialog with text=NULL
(which seems to happen when the text of the dialog is empty); basically,
this commit fixes crashes when trying to encode an empty text subtitle
dialog (see subrip and mov_text encoders).
Fixes Ticket2048.
Clément BœschandMichael Niedermayer
