For each item that was appended or deleted, the ("Tracks", NULL) pair
was inserted (duplicated over and over) in the tracklist_properties
dictionary. The dictionary was inserting it in the same position and
it assumed it had hash collisions, triggering an expansion at every
insertion of that pair.

Check if the key is present before inserting in the dict again, at
playlist_item_append, playlist_item_deleted events.
Signed-off-by: Ilkka Ollakka <ileoo@videolan.org>
(cherry picked from commit 0d3ac6a7df7e205bdaff6e1493abdf1a62fa45b8)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Ilkka Ollakka <ileoo@videolan.org>
(cherry picked from commit 308259d1ab249c19463c88b382f705ffde1343e5)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

(cherry picked from commit 6fa1935b5a4d1894689fac30c358a5b1887677c9)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

(cherry picked from commit d77948263d850f221ac18606f3e0f6c19295d51b)
Signed-off-by: Sébastien Escudier <sebastien-devel@celeos.eu>

We can only use it at application level
Just use the default libav* callback instead which prints to stderr
We might want to set libav* verbosity with av_log_set_level()
(cherry picked from commit 20155f800ecee139e81babf82b031cc5a8bbb045)
Signed-off-by: Rafaël Carré <funman@videolan.org>

detect errors and abort writing if they happen
(cherry picked from commit acf912ad543fd1a26a186bd5765c59c50955c0bb)
Signed-off-by: Rafaël Carré <funman@videolan.org>

In process-level mode we would call the context-level callback
(cherry picked from commit 58d5803a30527d3299e890c72484acc9b6de48db)
Signed-off-by: Rafaël Carré <funman@videolan.org>

Define it to EFBIG (file too big), it's close enough in this specific use
(cherry picked from commit aa77af7405751bec3f8adfd1c671133261dc9c8d)
Signed-off-by: Rafaël Carré <funman@videolan.org>

(cherry picked from commit 3cb559491e0ade3f98b029ef2e690ea32191a296)
Signed-off-by: Rafaël Carré <funman@videolan.org>

The field was renamed in lavc54 and setting values changed from
enum to bitfield.
We don't need to care about this because VLC doesn't specify values.
(cherry picked from commit b3cd3651868cce57d54e94c6662a383f7d94444e)
Signed-off-by: Rafaël Carré <funman@videolan.org>

avformat mux: convert to new avio API (cherry picked from commit 0011b7043c5c9115bf130068265e90142277d2f6)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

Pass its error output to msg_Err with %m
(cherry picked from commit aca40fdc52ae97e3a188d55af74a0b734b415664)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

avformat: remove cast to void * (cherry picked from commit 91899a03e70024e68c20603e2c243cb9742898b9)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

avformat: prefer avformat_new_stream() over av_new_stream() (cherry picked from commit 5dbdccf2fd7829a8a0fb3ccae65ed17f71c7078a)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

avformat: put_flush_packet was renamed some time ago (cherry picked from commit a6ebc7d7eb2231939ec94dd5a94bf819fb7c2086)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

Allocate avformat context and set its I/O context before using it
This is the correct way to it (and is documented in avformat.h)
(cherry picked from commit c5f388e8e37f38b4526fbca379305a918dd12141)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

- Use AVIOContext instead of ByteIOContext
- Use avformat_open_input instead of av_open_input_stream
- Use avio_alloc_context instead of init_put_byte, allocate context on the heap
- Use AVIOContext.seekable

print avformat_open_input errors

avformat_open_input is the most recent and is documented as available since
libavformat 53.2.0 (other changes happened in 52.105)
avformat_open_input was introduced in 52.94 in FFmpeg
(cherry picked from commit 4c3037ac45b92853d5d85f7bedebdb4b281da0c6)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

Workaround FFmpeg doc/APIchanges lies (cherry picked from commit 5b55c4dbab0ee2700300f9d8b79747a42ef26158)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

print avformat_find_stream_info errors
(cherry picked from commit 49944dcb5ac7ecc39da86bdb300bd7be8c9d2e47)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

Palette data is now transmitted as a side data in AVPacket
(cherry picked from commit 4ca6ef05e16554c71c7480a7b4df59ab3a43f6c4)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

avformat: shut up warning (unsigned char -> uint8_t) (cherry picked from commit 7df0345e7f8917ae534592caea34eba1bef4c26e)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

They were copied verbatim to AVDictionary equivalents some time ago
(cherry picked from commit 55c0636c559e3db6cb354306438c83e278430f48)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

They were only used to hold one opaque pointer
(cherry picked from commit e7e1a0f3bd279ce5fd2bdd4f1994bd7ff5be3deb)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

avcodec: update to libavcodec 54 (cherry picked from commit 8f24e725ff9945975f6d7232477cff7f7a8e2cff)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

- fix bug in ACCESS_GET_PTS_DELAY where it would fall through the next case
- print error messages from libavformat
- remove useless strdup
- check for exclusive access to avio when context open succeeded
- support for multiple input / output at the same time with libavformat > 54 (untested)
(cherry picked from commit cca698a05ea024a380330e1f3144996bba4c20e0)
Signed-off-by: KO Myung-Hun <komh@chollian.net>
Signed-off-by: Rafaël Carré <funman@videolan.org>

it's in libdvbpsi since 2004
(cherry picked from commit 0344812b799fd5aa1db60b91bda8c3d73ced8de1)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

(cherry picked from commit d7df112353124aa073b1b9fcb741ae24a3e1e7d1)

(cherry picked from commit c3ea7f28b638c8f54ad3463d12e1e3eeb1f682f1)

macosx: fixed resizing issue, which was in part fixed by the previous commit by accident (close #6203)
(cherry picked from commit f56aba45d9d053e74ff35c1aa877c4326391753a)

close #6252
Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>
(cherry picked from commit 73741d0175775e35447000b3b30f7e7f799d523b)

Make sure that if we would resize beyond the screen we set the origin values to ensure that the complete window is visible.
Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>
(cherry picked from commit 06f57d2bd7dd633d1ba731107b7cfc7048fc7e03)

(cherry picked from commit 8f813fc43ae83fa0d04d591a1f3c6b02b1d288e5)

(cherry picked from commit 84dc949e138f6231ba0c58572b7949ca4e5724a8)

(cherry picked from commit d843a2fc47bf6b9f4277310736b25caee8309ee7)

Signed-off-by: Felix Paul Kühne <fkuehne@videolan.org>
(cherry picked from commit be40ef554ab4d76d5ff221567940d0d4d5e3aa65)

There is no reliable way to determine the channel configuration of a
certain ALSA device. The ALSA operates at too low level for that.
So add a configuration option with default to stereo (as in previous
VLC versions). However, the user will not need to re-enable 5.1 every
time.

This commit also removes all the fancy device choices. The static
choices are overriden dynamically by the UI anyway. So the only use of
the list was in vlc --help.
(cherry picked from commit 119c3fc9f9cc1c05df23728a6e16ac9c702e93f2)

(cherry picked from commit 75977c876f7169e6e3da5605d86d78706bbfd130)

Do not clobber the output format channels to stereo. This would break
the VLC filtering pipeline.
(cherry picked from commit ea6e934aedda48cc024c4554cd33542ee83ab8e4)