Age | Commit message (Collapse) | Author |
|
vmaf 2.x is now on the official repositories.
References
----------
https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=a7c134ebf0db48dbd4701299c98239a389905fe4
|
|
|
|
Upstream ffmpeg finally has added support for libvmaf 2.0. It now
requires vmaf >= 2.0.0 to build. Repository vmaf package is still
at version 1.5.3 because repository ffmpeg does not support vmaf
2.0 at the time of this writing.
References
----------
https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=3d29724c008d8f27fecf85757152789b074e8ef9
|
|
|
|
The cuda11 patch is not needed anymore, since it's fixed[1] by
upstream.
Update svt patches. Note: the svt-hevc docs patch currently fails
to apply.
References
----------
[1] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=c67f354be8610eefb9a2b195f6dc4d6da637b687
|
|
libavresample was removed by upstream[1].
Temporarily disable libtensorflow due to a build error caused by
a recent upstream commit[2].
Also on this commit:
- refresh cuda patch
References
----------
[1] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=420cedd49745b284c35d97b936b71ff79b43bdf7
[2] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=f183d6555e714e00b41aec728feb8a731826cbdc
|
|
|
|
Upstream now supports librist[1].
glslang support was temporarily disabled previously[2] and now it
can be re-enabled since glslang on the official repositories was
updated.
Repository ffmpeg added LTO, so we add it here too. RFC about LTO
on Arch Linux is being discussed and is on the way to be approved.
Also on this commit:
- update svt-vp9 patch
- refresh cuda11 patch
[1] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=4098f809d605fc540870883a9f91880e13ce1d6b
[2] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=bd252d99089862a9c74d43a88837d243c52de38e
[3] https://github.com/archlinux/svntogit-packages/commit/ceff68a52e22bfedf1f8e75228a608b7ddab3d15
|
|
It currently fails to build with the stable version of lensfun
from the official repositories (stable release is from 2018).
Requires lensfun-git.
Also on this commit: update svt-hevc and svt-vp9 patches.
|
|
Missing on previous commit[1].
References
----------
[1] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=8095a6af514cae40f5f539a61aace56d7b20e2c8
|
|
|
|
rockchip-mpp was disabled[1] a while ago because the upstream
repository was temporarily set to private[2] (hidden from the
public). Now the rockchip-mpp repositry is public[3] again and
we can re-enable this feature.
References
----------
[1] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=40ba9358a0bfbe5d9944072621eb9264e30dcce8
[2] https://github.com/HermanChen/mpp/issues/15#issuecomment-690804709
[3] https://github.com/rockchip-linux/mpp/
|
|
At the time of writing, it does not work with the uavs3d stable
version and requires the git master version (uavs3d-git).
References
----------
[1] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=c952db9d68d31d1eca5ac770cc53bf35b885b087
|
|
Upstream svt-vp9 has now fixed their patch to support ffmpeg git
master[1].
Also on this commit: refresh cuda11 patch.
References
----------
[1] https://github.com/OpenVisualCloud/SVT-VP9/commit/7951c3cf6773c5e0ede00e4ce3e3ad2f7e090cfb
|
|
smbclient support was previously removed[1] due to a build failure
with samba 4.9. It looks like that the recently released samba 4.13
fixed[2][3] the issue and now ffmpeg is building fine with it.
References
----------
[1] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=861ed49ab870c6fdfb4cad8a2443b49af78f32a5
[2] https://github.com/samba-team/samba/pull/212/
[3] https://github.com/samba-team/samba/commit/1114b02a72ce0c86a5301816560d270ec47f8be3
|
|
libwavpack was removed by upstream[1]. It causes the svt-vp9
patch to fail, and thus a local patch is now needed while
upstream svt-vp9 does not provide an updated one.
Also on this commit: refresh cuda patch.
References
----------
[1] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=45070eec4c089b06947f07e25cdb1bc8b2102553
|
|
Upstream repository for rockchip mpp is gone.
It was at: https://github.com/rockchip-linux/mpp/
I could not find any upstream word about it. The Rockchip Wiki
still points to this repository, without saying if it was really
dropped or moved to another place.
Due to this, new users now cannot build rockchip-mpp package.
Disabling it so new users can build ffmpeg-full-git.
|
|
Upstream ffmpeg now requires[1] a glslang version that is not
yet on the official repositories. glslang-git from the AUR can
be used, but the package needs a lot of improvements (see
comments on the package page). Temporarily disable it until we
have a supported glslang version on the official repositories
or until glslang-git is improved.
The patch for svt-hevc doc currently fails to apply becasue
there was a split on upstream file doc/general.texi[2]. Now
it's splitted into doc/general_contents.texi. A local modification
of the patch does the job.
Upstream svt-hevc updated[3] their patch to work with the current
ffmpeg git master, so we do not need the custom patch anymore.
Now using the svt-hevc patch directly from upstream.
Also on this commit: refresh the vmaf and cuda11 patches.
References
----------
[1] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=98ea1a662ee172961feaa374fe8d26078838d250
[2] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=6accb7718aa4c9f5ca8f3280a1ed4e09587a627f
[3] https://github.com/OpenVisualCloud/SVT-HEVC/commit/ead6fdf7c9ff84511b42fc1658c1654b84d83e4b
|
|
The tensorflow build issue was fixed[1] upstream by ffmpeg.
References
----------
[1] http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=3c05c8a15f2fa5543213f6c8cbe439ad4376137d
[2] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=b38e834c47562855016e855af82456e3f752eccd
|
|
A recent commit[1] at upstream ffmpeg broke tensorflow support.
Temporarily disabling it until it's fixed upstream.
References
----------
[1] http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=0a51abe8ab71c03eff66200265142fc910bcbfe2
|
|
|
|
ffmpeg has dropped the underscore in svt-av1 library names[1]
(libsvt_av1 -> libsvtav1).
References
----------
[1] http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=30a4bdbc1f4853c1deb2cd551d8c661f7b067503
|
|
|
|
svt-av1 is now directly supported by ffmpeg[1].
The upstream patch from svt-hevc now fails to apply after this
recent ffmpeg change. Created a patch for it.
References
----------
[1] http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=0e20dee5c21a40f23ecfe7af2c360493a169c0f0
|
|
Upstream ffmpeg fixed the build issue with srt 1.4.1[1].
References
----------
[1] http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=7c59e1b0f285cd7c7b35fcd71f49c5fd52cf9315
[2] https://trac.ffmpeg.org/ticket/8760
[3] https://aur.archlinux.org/cgit/aur.git/commit/?h=ffmpeg-full-git&id=8f9bff8de0df502a0a13f2d243c6ae018e67603c
|
|
cuda 11.0 minimum supported architecture is 5.2 (compute_52/sm_52).
svt-vp9 support is better now with the current upstream stable
version.
|
|
|
|
|
|
- Now that svt-av1 0.84 is out, it seems that this stable version
is suitable to be used here.
- Removed support for legacay nvidia-340xx drivers since I do not
use it anymore.
- Added intel-openvino option in disabled mode. This is a new
upstream option[1][2]. It looks like that ffmpeg segfaults
when using it with an unsupported cpu. Users wanting to
experiment with it will need to add -I and -L options with
the corresponding include and library paths.
References
----------
[1] http://git.videolan.org/?p=ffmpeg.git;a=commit;h=ff37ebaf30e675227655d9055069471bb45e5ceb
[2] http://git.videolan.org/?p=ffmpeg.git;a=commit;h=9bcf2aa47756c253238903f3e5ef6c96d927cfe6
|
|
Upstream svt-av1 has provided support for the recent API
changes of ffmpeg git master[1], so it can be added again.
Upstream srt recently introduced a commit[2] that removes
deprecated API and breaks ffmpeg support. Bug reports were
opened upstream at srt[3] and ffmpeg[4], and it's up to
ffmpeg to fix it. Unfortunatelly it was marked as a 'wish'
at the ffmpeg bug tracker, so it can take very long to be
fixed.
References
----------
[1] https://github.com/OpenVisualCloud/SVT-AV1/commit/b255a71cf47038ec73873ae507a56bc9fdef7c8f
[2] https://github.com/Haivision/srt/commit/0e2201aff6b379979cec43fee5e8f162717f6346
[3] https://github.com/Haivision/srt/issues/1382
[4] https://trac.ffmpeg.org/ticket/8760
|
|
|
|
A recent commit[1] on upstream ffmpeg broke svt-av1. The fix
seems to be not so simple, so temporarily removing it until
upstream svt-av1 provides a fix.
References
----------
[1] http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=827d6fe73d2f5472c1c2128eb14fab6a4db29032
|
|
|
|
Some function declarations were moved to packet_internal.h[1].
This fixes the following build error:
libavcodec/libsvt_av1.c: In function ‘eb_receive_packet’:
libavcodec/libsvt_av1.c:396:5: error: implicit declaration of function ‘ff_side_data_set_encoder_stats’ [-Werror=implicit-function-declaration]
396 | ff_side_data_set_encoder_stats(pkt, headerPtr->qp * FF_QP2LAMBDA, NULL, 0, pict_type);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
References
----------
https://github.com/FFmpeg/FFmpeg/commit/6e1903938befc0bce2f64e01770af2f65d625cfc
|
|
|
|
Patch does not apply anymore on git master. And it's buggy at
the current moment, not working properly at all.
|
|
|
|
|
|
The avisynth headers were removed from the ffmpeg source tree[1].
It now requires avisynthplus 3.5 or later for for building with
avisynth support[2][3].
References
----------
[1] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=0c75acb4ce5db623f4c1c2729468e66c3e28ad67
[2] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=6d8cddd1c67758636843f6a08295b3896c2e9ef8
[3] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=6e959ad680288e34eccb4ee2dd205e4ffa495f95
[4] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=56f59246293de417d27ea7e27cb9a7727ee579fb
[5] https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=7b0f7a7f3b6bf09c076c23d0701bf42a2c997ca2
|
|
References
----------
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/a88449ffb2f249e31ecf98f84e7f8c338308e7e4
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/6fca61bbc917678b3e517be4f6594ce52a16a93c
https://git.ffmpeg.org/gitweb/ffmpeg.git/commit/1676d0fb25c6851d0750ed6926b8205ed20973eb
|
|
intel-media-sdk is now installed on '/usr' prefix. No need to
specify the pkg-config search path anymore.
|
|
References
----------
https://git.videolan.org/?p=ffmpeg.git;a=commit;h=01994c93db43e50c01349203a76c0b7111d1d2f3
|
|
References
----------
https://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=d8bf24459b694338de4ceb2a2e6d4d2949d6658d
|
|
|
|
|
|
Upstream ffmpeg recently added support for AMD AMF hardware
accelerated video encoder for Linux[1] via Vulkan (it was
available for Windows only). At the time of this writing,
the git master version of AMF headers are needed for building.
References
----------
[1] https://git.ffmpeg.org/gitweb/ffmpeg.git/commitdiff/f8ad2ddd7a51df5d6e768ba2a6f65ee64b94e29b
|
|
Although this option is specific to the Raspberry Pi, it was
compiling fine so it was ok to leave it there.
But a recent upstream ffmpeg commit[1] broke build with this
option, making it to fail with the following error:
‘OMX_IndexConfigBrcmVideoRequestIFrame’ undeclared.
Now it needs headers[1] that are specific to the Raspeberry Pi
firmware, so it will be removed.
References
----------
[1] https://git.ffmpeg.org/gitweb/ffmpeg.git/commitdiff/76c82843ccad11666603e3f1f6d9d9a36c780a32
[2] https://github.com/raspberrypi/firmware/blob/1.20190819/hardfp/opt/vc/include/IL/OMX_Index.h#L381
|
|
|
|
|
|
Patches for intel svt components will be different at each
component release, so they better be versioned.
This commit updates intel-svt-av1 patch from version 0.5.0
to 0.6.0.
|