GitLab

same as v4l2, can't do win-compile myself, so buildbot will be my judge.

Fix #3497
See http://msdn.microsoft.com/en-us/library/dd407253%28VS.85%29.aspx?ppud=4 for more information

It was duplicated with es_format_t::i_sar_num/den.

Replace argument = realloc( argument, size ); with realloc_or_free() in modules/*, and while at it add assert( argument ) to mark unhandled ENOMEM conditions, also for malloc().

For each CoInitialize we have to call CoUninitialize.

(cherry picked from commit a6110fb1d253146810bbe5f9b5ba0097b72b4e3c)
(cherry picked from commit b24fc9a9680a2e62d1a3f660a7c20f5c9ba5cdb2)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

The codec defined will only be used by encoder to force one of the types.

 There will be some regressions, I used automatic sed replacement with manual
checks but there was a lot to change.
 It's only the first pass.

Except for a few cases:
 - dialog providers -> cannot be selected explicitly,
 - window providers -> cannot be selected explicitly,
 - ffmpeg-deinterlace -> "ffmpeg" does not make much sense here??
 - kate -> already duplicating its own shortcut

Note sure about packetizers: added just in case

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>

Invalid free pointed out by xxcv.

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

-Fix dshow "none" deivice be recognized as none
 Note that another patch concerning var_optionParse string
 dequoting should be applied for this to be fixed
 ( the concept is not to provide quoted strings to dshow options )
 -Account for the case when BOTH devices could not be opened
   (file an error, instead of seeming to play)
 -Cleaned a little
 -Added dbg and error msgs
 -Added coments

Rev 3
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Those multi-lines are still to be done (it's going to be fun...)
See also 5af4cc85.

# HG changeset patch
# User Edouard Gomez <ed.gomez@free.fr>
# Date 1224021637 -7200
# Node ID 0c4727aa17ba532172cae4aded7d16d70ea4ea53
# Parent  8e7c3f94407dc1500438237ac6bf3d484bfba742
dshow: try to improve multithreading code for RAW sample grabbing

This patch tries to address two theoritical problems:
 - The filter capturepin should not lock p_sys->lock. By chance, win32 mt
   locking is recursive by default.
    - Do not lock again, this makes things clearer for poor POSIX coders like
      me.
 - The current code does not try to push samples as fast as it receives them.
   This is caused by the arbitrary msleep call which can differ sample delivery
   with up to 10ms delay. Moreover, only a single sample at a time was processed
   either for audio or for video.
    - Use MT condition instead of sleep so no artificial delay is introduced in
      the delivery chain.
    - Process all available samples at once.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

# HG changeset patch
# User Edouard Gomez <ed.gomez@free.fr>
# Date 1224021633 -7200
# Node ID 9564c1736b76c96f37fb85476d36496b3d7955bb
# Parent  40d5c2186f7fb31e944b5b542b3366bbc420cc0e
dshow: prevent division by zero

In the DShow capture module, division by zero is possible in
multiple modulo statements when dealing with granularity values
returned by capture filters.

Change the code so that:
 - if granularity is zero, then width/height/samplingrate must be
   a perfect match with requested settings.
 - if granularity is non zero, then use the old condition tests.

This fixes real crashes reported by VLC users on different forums.

See:
http://forum.doom9.org/archive/index.php/t-134655.html
http://forum.videolan.org/viewtopic.php?f=2&t=37056&p=114065&hilit=blackmagic

Tested with both a Decklink HD Extreme card and a Decklink HD Pro card.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>