Age | Commit message (Collapse) | Author |
|
|
|
|
|
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/56c14f231159d9a0dfa0a46661886c9baad870a6
|
|
cuda is no longer needed at run time after disabling libnpp in the
previous commit[1], but now only at build time.
[1] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=4b2bf561fc40b47b985e56e7835b148456c8af0d
|
|
npp filters (like scale_npp, from libnpp) are not compatible with
cuda 13.
A comment[1] from a ffmpeg developer indicates that libnpp is to
be considered deprecated and that npp filters will not receive any
compatibility fixes. He also said that libnpp eventually may be
disabled in future ffmpeg commits by the configure script if it
detects a newer libnpp version.
[1] https://code.ffmpeg.org/FFmpeg/FFmpeg/issues/20191#issuecomment-6514
|
|
Upstream splitted the C and C++ flags. Now, for correctly building
with decklink-sdk (which is C++ in FFmpeg), the include flag needs
to be also in CXXFLAGS, otherwise, a build error will happen at
compile time. The flag is still needed in CFLAGS for detecting the
decklink-sdk, otherwise, a build error will happen at configure
time.
Also changed the priority of the custom pkg-config search path in
this commit.
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/0ce413af9cc8040095a6b714600d6a8dceaca514
|
|
ffmpeg currently have no support for decklink sdk 14.3 and later.
It currently fails to build with it, building only with version
14.2 and bellow.
Using external sources to a lower version. This will avoid making
a dedicated package for decklink 14.2 during the time that ffmpeg
remains without support for 14.3 and later.
|
|
|
|
whisper.cpp AUR package is broken at the time of writing,
building it locally as a static library for the time being.
References
----------
[1] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/13ce36fef98a3f4e6d8360c24d6b8434cbb8869b
|
|
The 'Cache::~Cache()' symbol conflict between avisynthplus and
vmaf was fixed upstream by avisynthplus.
References
----------
[1] https://gitlab.archlinux.org/archlinux/packaging/packages/ffmpeg/-/issues/19
[2] https://github.com/AviSynth/AviSynthPlus/commit/d1a32d5229602f181574c95fa0db68bcf1762606
[3] https://github.com/AviSynth/AviSynthPlus/pull/452
|
|
With the release of openapv 0.2.0.0, ffmpeg now builds fine with
this stable version.
Also in this commit
- add ohcodec in disabled mode (OpenHarmony is an OS)
- refresh patches
|
|
openapv git master is currently required for building.
Also in this commit: refresh the other patches.
|
|
Now that libplacebo was updated in the official repositories
to a version greater or equal than 7.350.0, we can use the
stable version again.
References
----------
https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=981efa86c8d70b054043ea18aa0ca8b965d876a8
|
|
Also:
- update dependencies
- refresh patches
- cosmetic changes
References
----------
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/8c920c4c396163e3b9a0b428dd550d3c986236aa
|
|
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/fab691edaf53bbf10429ef3448f1f274e5078395
|
|
|
|
This fixes a build error due to a recent change[1] in upstream
FFmpeg, which removed 'av_stream_get_side_data()'. It is fixed[2]
by upstream libplacebo, and the fix is present only in the git
master branch at the time of writing. Using libplacebo-git while
there is no stable libplacebo with this fix.
The original and now unmaintained svt-hevc and svt-vp9 patches
no longer applies on top of FFmpeg git master. Added fixes for
them.
[1] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/ec8e796b42bccf999fdf31c110385e2cc119b7f2
[2] https://github.com/haasn/libplacebo/commit/563ea4a1a23237c0acd9f0ba29f5166ebcd5cf60
|
|
Upstream ffmpeg recently added a change[1] in amfenc that uses
code present[2] only in amf git master at the current time.
[1] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/95217872adff358af6ed4d06777e4625648904cd
[2] https://github.com/GPUOpen-LibrariesAndSDKs/AMF/commit/8f5a645e89380549368eec68935b151b238aa17b
|
|
It looks like the vulkan decoding segmentation fault with nvidia
is fixed[1][2] upstream, so the patch can now be removed.
[1] https://trac.ffmpeg.org/ticket/10847#comment:4
[2] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/ca591e6b50ee616fc36711cc66d594e51209dddb
|
|
References
----------
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/dc11c12b6466795f5f9eb057b1aebd76bf129785
|
|
Fix the avisynthplus patch applying.
|
|
|
|
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/c75940db290478df657c09089605d221dc47118e
|
|
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/a1304272c3a0ef7c3d77c702dd466b3f11bf300f
|
|
Move vpl-runtime from depends to optdepends to easy the dependency
burden on non-Intel users.
|
|
References
----------
[1] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/240fd04db231efa4abad3abbd808140b4112324a
|
|
|
|
Using '-isystem' will let the installed headers to be treated as
system headers[1] as it should be expected, potentially reducing
warnings that may come from them.
[1] https://gcc.gnu.org/onlinedocs/cpp/System-Headers.html
|
|
The flite audio filter is giving a segmentation fault, so we
better use only a compiler flag to fix the build with gcc 14.
When using just the compiler flag, the flite filter also gives
the segmentation fault, but in this circunstance it looks better
to leave the upstream source unthouched.
|
|
This fixes the following error when compiling with gcc 14:
libavfilter/asrc_flite.c: In function ‘init’:
libavfilter/asrc_flite.c:230:21: error: assignment to ‘cst_audio_stream_callback’ {aka ‘int (*)(const struct cst_wave_struct *, int, int, int, void *)’} from incompatible pointer type ‘int (*)(const cst_wave *, int, int, int, cst_audio_streaming_info *)’ {aka ‘int (*)(const struct cst_wave_struct *, int, int, int, struct cst_audio_streaming_info_struct *)’} [-Wincompatible-pointer-types]
230 | flite->asi->asc = audio_stream_chunk_by_word;
| ^
|
|
We have just added tensorflow again, as the repository package was
fixed, allowing to properly use it here again. Unfortunately, there
is a tensorflow upstream issue[1] that prevents applications to
startup when they are linked to both ffmpeg and protobuf. We better
remove tensorflow again to not break client applications, until
this is fixed by upstream.
References
----------
[1] https://github.com/tensorflow/tensorflow/issues/59081#issuecomment-2093988476
|
|
This will prevent the compiler flags '-I' and '-L' to be shown in
the ffmpeg version string. This is likely an unneeded information
to be present there.
|
|
|
|
Original patch for fixing nvidia vulkan decoding:
https://github.com/streetpea/chiaki4deck/issues/239#issuecomment-1929365254
References
----------
[1] https://trac.ffmpeg.org/ticket/10847
[2] https://trac.ffmpeg.org/ticket/10970
|
|
References
----------
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/030e1401451200566a5303f35cbe1456e31dd81e
|
|
|
|
Tensorflow is now fixed[1] and can be added again[2].
The new upstream option --enable-libtorch[3] can be added only
when the repository ffmpeg sonames (library versions) catch up
with ffmpeg git master, as python-pytorch depends on ffmpeg and
the current ffmpeg git master bumped[4] the soname of all
libraries. Otherwise, the ffmpeg libraries will conflict.
References
----------
[1] https://gitlab.archlinux.org/archlinux/packaging/packages/tensorflow/-/issues/2
[2] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=5b34804510e96f108909ef41963a34fdb25df903
[3] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/f4e0664fd1bc94ee1a1afbd20c5a32ac33d93a2d
[4] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/783d00b203b2a6c67ec289856ff080c7a36ddb6e
|
|
Also:
- explicitly disable libshaderc, since it is mutually exclusive
with libglslang
- prioritize openvino pkgconfig search path
|
|
[1] https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/899302bb5f6ac0484fedc2865ee3beca021eba85
|
|
opencv support was disabled[1] a long time ago because ffmpeg failed
to build with it. opencv 3.x does not support C compilation mode[2]
anymore, and ffmpeg uses only C. There is no ffmpeg interest[3] in
compiling opencv in C++ mode, so opencv2 must be used here.
References
----------
[1] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=f9d589fa30dddc8318fff742a87b7927566b2a9e
[2] https://github.com/opencv/opencv/issues/10963
[3] https://trac.ffmpeg.org/ticket/7059
|
|
References
----------
[1] https://git.videolan.org/?p=ffmpeg.git;a=commit;h=b6138633cdf2ecae22c4c02029e91245c430c4fe
[2] https://git.videolan.org/?p=ffmpeg.git;a=commit;h=c59a96fd08620bd8239c218f2e0dfb8429c81c3c
[3] https://git.videolan.org/?p=ffmpeg.git;a=commit;h=cfe29478873e5000416a2da58a92102b7ed7255c
|
|
|
|
|
|
|
|
HTML pages fails to build with texinfo 7.1. Temporarily disabling
it until there is an upstream fix.
nvenc now builds fine the the stable ffnvcodec-headers package.
Let's give another chance for staying with the stable version.
|
|
ffmpeg git master currently requires ffnvcodec-headers git master.
This is not the first time it happens, so maybe it's better to keep
it for the long run to avoid breakages in building.
|
|
This was missing when switched libfmx to onevpl.
|
|
The stable version of ffnvcodec-headers in the official
repositories was updated and now gives support for
av1_nvenc.
|
|
|
|
A workaround was added to the flite1 package to fix its build.
The newly released pocketsphinx 5.0.0 currently fails to build.
Temporarily disable it until there is support by upstream ffmpeg.
|