* qatar/master: (23 commits)
avconv: Reformat s16 volume adjustment.
ARM: NEON optimised vector_fmac_scalar()
dca: use vector_fmac_scalar from dsputil
dsputil: add vector_fmac_scalar()
latmenc: Fix private options
vf_unsharp: store hsub/vsub in the filter context
vf_unsharp: adopt a more natural order of params in apply_unsharp()
vf_unsharp: rename method "unsharpen" to "apply_unsharp"
vf_scale: apply the same transform to the aspect during init that is applied per frame
vf_pad: fix "vsub" variable value computation
vf_scale: add a "sar" variable
lavfi: fix realloc size computation in avfilter_add_format()
vsrc_color: use internal timebase
lavfi: fix signature for avfilter_graph_parse() and avfilter_graph_config()
graphparser: prefer void * over AVClass * for log contexts
avfiltergraph: use meaningful error codes
avconv: Initialize return value for codec copy path.
fate: use 'run' helper for seek-test
fate: remove seek-mpeg2reuse test
Fix memory (re)allocation in matroskadec.c, related to MSVR-11-0080.
...
Conflicts:
doc/filters.texi
libavfilter/avfilter.h
libavfilter/avfiltergraph.c
libavfilter/avfiltergraph.h
libavfilter/graphparser.c
libavfilter/vf_scale.c
libavfilter/vsrc_color.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Or in other word, do not go to the next line when the text cannot be
rendered within the frame.
The rationale is that wrapping is hard to manage, as it depends on the
position of the text, thus making hard/impossible to compute the size
occupied by the rendered text, and makes the filter behavior
unpredictible with moving text (implemented in a pending patch).
Also the current wrapping code is very limited (for example it splits
words in the middle), and correctly handling wrapping requires much
more complexity (alignment, gravity, word parsing etc.), while
drawtext is meant to be only a thin wrapper around libfreetype, so
such advanced features don't belong here.
The new syntax is more extensible and more user-friendly.
This breaks the previous syntax, should not be an issue as possibly
no-one is already using anullsrc.
Add aconvert filter to perform sample format, channel layout, and
packing format conversion.
The aconvert code depends on audio conversion code in libavcodec, so
this requires a dependency on libavcodec.
Based on previous work by S.N. Hemanth Meenakshisundaram and Mina Nagy
Zaki, performed for the GSoC 2010 and 2011.
This list is mainly useful to be directly used as a parameter by the
audio sink buffer, when there is no need to force a specific channel
layout.
The list of formats is generated programmatically, and avoids channel
layouts not supported by libavfilter (with more than 8 channels).
We preferred to avoid compile-time generation in order to avoid some
unnecessary complexity (build hacks, generating code maintainance
etc.).
A more compact way to express this list through the use of macros
should be possible when libavfilter will support more than 8 channels,
as there will be no need to filter out channel layouts with more than
8 channels.
See thread:
Subject: [FFmpeg-devel] [PATCH] lavfi: add ff_all_channel_layouts internal symbol
Date: Sat, 20 Aug 2011 15:33:16 +0200
A function name should tell what the function does rather than what
the function returns. This also avoids possible conflicts
(e.g. between a function and a public list of all supported formats),
and clarifies the code.
Breaks audio API/ABI, add a backward compatibility layer for video
API/ABI.
These changes were forgotten in commits:
lavfi: simplify signature for avfilter_get_audio_buffer() and friends
(commit 2c0317419b)
and
lavfi: consistently use int for sample_rate in AVFilterLink and AVFilterBufferRefAudioProps
(commit 4381bddc9f)
The additional parameters were never used and are complicating the
function interface. Also, they were inconsistent with the way the
video API works.
So this assumes that a requested samples buffer will have *always* the
format specified in the requested link.
This breaks audio filtering API and ABI.
This is done in order to clarify the non-video-specific nature of the
buffersink code, as the result of the video/audio API unification of
the previous commit, and for improving overall consistency.
The new API is more generic (no distinction between audio/video for
pulling frames), and avoids code duplication.
A backward compatibility layer is kept for avoiding tools ABI breaks
(only for the video binary interface, audio interface was never used
in the tools).
Allow to cache more than one frame (e.g. for filters which return
more than one frame when avfilter_request_frame() is called on them),
and do not discard previously cached frames when a new one is added.
The gradfun filter was already integrated natively in libavfilter.
Both filters issue the same output, and have a comparable performance.
See thread:
Subject: [FFmpeg-devel] [PATCH] lavfi: remove gradfun mp wrapper
Date: Fri, 5 Aug 2011 00:19:04 +0200
Update/extend documentation related to the link fields to set, and fix
a pads->filters braino.
Signed-off-by: Stefano Sabatini <stefano.sabatini-lala@poste.it>
Link properties have to be checked after config_props() is called to
make sure everything is sane, so the default config_props() for output
links was redundant.
Also remove channel_layout check as it is now negotiated.
Signed-off-by: Stefano Sabatini <stefano.sabatini-lala@poste.it>
Previously, "dar" and "a" were documented like "input display aspect
ratio", but their actual value was "in_w/in_h".
In order to avoid to break scripts which rely on the "a" variable, the
patch keeps the same semantics but fixes the corresponding docs, and
fixes the semantics of the recently added "dar" variable, which
now correctly expresses the input Display Aspect Ratio value as
"(inw_w/in_h)*sar".
The licence was changed from GPL to LGPL with explicit approval from
the original author.
See thread:
Subject: [FFmpeg-devel] [PATCH] lavfi: add rgbtestsrc source,
ported from MPlayer libmpcodecs
Date: Mon, 11 Jul 2011 16:32:41 +0200
With the following additions:
* support to gray format
* support to yuva420p format
* parametric luma/chroma/alpha radius
* consistency check on the radius values, avoid crashes with invalid values
Since avfilter_graph_parse() creates the "[in]" inout for the first
unlabelled input pad, it is expected that it will create an "[out]"
inout for last unlabelled output pad, even in the case where it cannot
find any open input pad with that name.
This change removes the check on the existence of an open input pad
named "out", so it simplifies the checked condition while implementing
a more intuitive behavior.
