Colin NG [Sun, 21 Jan 2018 05:35:30 +0000 (13:35 +0800)]
libavformat/dashdec: Fix for ticket 6658 (Dash demuxer segfault)
1 Add function 'resolve_content_path' to propagate the baseURL from
upper level nodes.
* if no baseURL is available, the path of mpd file will be set as the baseURL.
2 Remove checking for newly established connection.
3 Establish the communication protocol in each connection rather than
applying one protocol to all connection.
Brendan McGrath [Sun, 21 Jan 2018 05:20:02 +0000 (13:20 +0800)]
dashdec: Only free url string if being reused
If no representation bandwidth value is set, the url value returned
by get_content_url is corrupt (as it has been freed).
This change ensures the url string is not freed unless it is about
to be reused
Changes since v1:
1 removed the unneeded 'if' statement (as pointed out by Michael Niedermayer
2 added comment to make it clear why the av_free was required Signed-off-by: Brendan McGrath <redmcg@redmandi.dyndns.org>
Brendan McGrath [Sun, 21 Jan 2018 05:16:42 +0000 (13:16 +0800)]
avformat/hlsenc: Check that data is set
If codecpar->extradata is not set (for example, when the stream goes
through the 'tee' muxer), then a segfault occurs.
This patch ensures the data variable is not null before attempting
to access it
Before the var_stream_map option was available - I was using the tee
muxer to create each resolution as an individual stream.
When running this configuration after the most recent hlsenc change
I hit a segfault
The most simple command which recreates the segfault is:
ffmpeg -i in.ts -map 0:a -map 0:v -c:a aac -c:v h264 -f tee [select=\'a,v\':f=hls]tv_hls_hd.m3u8
avcodec: v4l2_m2m: context: fix raising warning on POLLERR
During the initialization stage, the codec attempts to get free
buffers from the driver before any have been queued (this is to keep
the code simple and generic)
When the kernel driver detects this situation, it returns POLLERR in
revents and ffmpeg therefore raises a warning.
This commit disables the warning since no buffers were queued to the
driver yet.
Qualcomm's db410c/db820 Venus driver currently present in mainline
kernel has a bug which mishandles the CMD_STOP requests causing the
decoder to block while draining [1].
This patch removes the workaround that was used to prevent that
situation.
Encoding/Decoding tested on db820c.
[1] on CMD_STOP, the driver is flushing all buffers and never raising
IPIPE which ends up in blocking on poll.
Mark Thompson [Tue, 9 Jan 2018 22:56:41 +0000 (23:56 +0100)]
avcodec: v4l2_m2m: fix races around freeing data on close
Refcount all of the context information. This also fixes a potential
segmentation fault when accessing freed memory (buffer returned after
the codec has been closed).
avcodec/snowdec: Fix integer overflow before htaps check
Fixes: runtime error: signed integer overflow: -1094995529 * 2 cannot be represented in type 'int' Fixes: 4828/clusterfuzz-testcase-minimized-5100849937252352 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Henrik Gramner [Fri, 4 Aug 2017 22:09:52 +0000 (00:09 +0200)]
x86inc: Enable AVX emulation for floating-point pseudo-instructions
There are 32 pseudo-instructions for each floating-point comparison
instruction, but only 8 of them are actually valid in legacy-encoded mode.
The remaining 24 requires the use of VEX-encoded (v-prefixed) instructions
and can therefore be disregarded for this purpose.
avdevice/decklink: Fix compilation of module on OSX
Clang applies the missing-prototypes warning on C++ files, whereas
gcc only applies it to C. As a result, the decklink_common.cpp file
fails to build because of missing prototypes in DecklinkDispatch.cpp
(which is #included by decklink_common.cpp).
We don't want to change the actual Blackmagic SDK sources, so
suppress the warning just for that one #include.
James Almer [Thu, 18 Jan 2018 22:12:30 +0000 (19:12 -0300)]
fate: update filter-acrossfade test reference file
<jamrial> durandal_1707: 8088b5d69c broke the acrossfade test
<@durandal_1707> jamrial: there was test?
<jamrial> durandal_1707: fate-filter-acrossfade
<@durandal_1707> what broke?
<jamrial> what used to be one frame is now two
<@durandal_1707> ahh, just update test
Gyan Doshi [Sat, 23 Dec 2017 10:44:25 +0000 (16:14 +0530)]
avfilter/vidstab: check bytesPerPixel only for packed formats.
libvidstab introduced this variable only for packed formats but in
vf_vidstab*.c, it's checked for all inputs. So the filter errors out for YUV422/444P streams.
Fixes #6736.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
D3D11 has rather fine grained per format capabilities for different uses
that can be queried at runtime. Since we don't know what the user wants
to do with the formats when av_hwdevice_get_hwframe_constraints() is
called, we simply return all formats that have the most basic support.
wm4 [Tue, 16 Jan 2018 11:57:04 +0000 (12:57 +0100)]
avformat: make avformat_network_init() explicitly optional
It was sort of optional before - if you didn't call it, networking was
initialized on demand, and an ugly warning was logged. Also, the doxygen
comments threatened that it would be made strictly required one day.
Make it explicitly optional. I would prefer to deprecate it fully, but
there might still be legitimate reasons to use this. But the average
user won't need it.
This is needed only for two reasons: to initialize TLS libraries like
OpenSSL and GnuTLS, and winsock.
OpenSSL and GnuTLS were already silently initialized on demand if the
global network init function was not called. They also have various
thread-safety acrobatics, which make concurrent initialization within
libavformat safe. In addition, the libraries are moving towards making
their global init functions safe, which removes all need for central
global init. In particular, GnuTLS 3.5.16 and OpenSSL 1.1.0g have been
found to have safe init functions. In all cases, they use internal
reference counters to avoid that the global uninit functions interfere
with concurrent uses of the library by other API users who called global
init.
winsock should be thread-safe as well, and maintains an internal
reference counter as well.
Since we still support ancient TLS libraries, which do not have this
fixed, and since it's unknown whether winsock and GnuTLS
reinitialization is costly in any way, don't deprecate the libavformat
functions yet.
avcodec/wavpack: Fix integer overflows in wv_unpack_stereo / mono
Fixes: runtime error: signed integer overflow: 2146276249 + 1487583 cannot be represented in type 'int' Fixes: 4823/clusterfuzz-testcase-minimized-4551896611160064 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/ac3dec_fixed: Fix integer overflow in scale_coefs()
Fixes: runtime error: signed integer overflow: 2147483520 + 128 cannot be represented in type 'int' Fixes: 4800/clusterfuzz-testcase-minimized-6110372403609600 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
In addition, this does not allow creating frames contexts with sw_format
for which no known transfer formats exist. In theory, we should check
whether the chroma format (i.e. the sw_format) is supported at all by
the vdpau driver, but checking for transfer formats has the same effect.
Note that the pre-existing code adds 1 to priv->nb_pix_fmts[i] for
unknown reason, and some checks need to account for that to check for
empty lists. They are not off-by-one errors.
sfan5 [Sun, 14 Jan 2018 14:36:02 +0000 (22:36 +0800)]
dashdec: Correct seeking behaviour
dash_read_seek() is called only once to issue a seek
of *all* streams to the specified timestamp. But to
avoid reopening each stream, do a "dry run" for streams
that are in a discarded state.
sfan5 [Sun, 14 Jan 2018 14:33:57 +0000 (22:33 +0800)]
dashdec: Expose bandwidth and representation ID as metadata
The primary goal was making it viable to play YouTube/Vimeo/... videos
using the native demuxer, since mpv currently uses a workaround in form
of Edit Decision Lists (EDL).
Implemented features:
1 Exposing id / bitrate as stream metadata (similar to the HLS demuxer)
2 Support for multiple video and audio streams
3 A few minor parts of the specification that are in use at YouTube
dcaenc: cleanup on init failure and add a threadsafe init codec cap
The encoder didn't clean up if a malloc failed during init.
It also doesn't need any external tables to be initialized on init.
Finally, it didn't need to check for whether avctx->priv_data exists during
uninit.
It tests a useless profile which sounds no better than regular aac and which
takes extremely long to encoder something. Also it has been behind experimental
flag for as long as it has been supported.
Should be removed altogether sometime in the future.
The twoloop coder sounds decent at low bitrates, however at higher bitrates
it sounds worse than the fast coder (which used to be the old twoloop coder
before October 2015) and needs quite a lot more CPU.
Change the default to fast. It has been well tested and has had little changes
over the years so its been confirmed to be quite stable.
Also change its description (not valid for more than a year) and the
documentation.
Marton Balint [Tue, 19 Dec 2017 22:49:41 +0000 (23:49 +0100)]
avfilter/vf_framerate: simplify filter
The framerate filter was quite convoluted with some filter_frame /
request_frame logic bugs. It seemed easier to rewrite the whole filter_frame /
request_frame part and also the frame interpolation ratio calculation part in
one step.
Notable changes:
- The filter now only stores 2 frames instead of 3
- filter_frame outputs all the frames it can to be able to handle consecutive
filter_frame calls which previously caused early drops of buffered frames.
- because of this, request_frame is largely simplified and it only outputs
frames on flush. Previously consecuitve request_frame calls could cause the
filter to think it is in flush mode filling its buffer with the same frames
causing a "ghost" effect on the output.
- PTS discontinuities are handled better
- frames with unknown PTS values are now dropped
Jérôme Martinez [Fri, 5 Jan 2018 10:09:01 +0000 (11:09 +0100)]
avcodec/ffv1enc: mark RGB48 support as non-experimental
Resulting bitstream was tested with a conformance checker
using the last draft of FFV1 specifications.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Also the files are already in the wild, and decoder support is
thus needed. And with decoders widely supporting it, there is no
advantage in not allowing it in the encoder.
The exact bitstream format may change in future versions of the
spec, if improvments are found.
Jörn Heusipp [Thu, 11 Jan 2018 08:11:30 +0000 (09:11 +0100)]
avformat/libopenmpt: Update to libopenmpt 0.3 API
libopenmpt 0.3 deprecates openmpt_module_create_from_memory() and
provides a replacement function openmpt_module_create_from_memory2().
Detecting libopenmpt 0.3 can be done at build time via the API
version macros provided by libopenmpt. libopenmpt 0.2 did not provide
all required macros, however libopenmpt documents the required #define
shims that can be safely added for libopenmpt 0.2.
Using openmpt_module_create_from_memory2() instead of
openmpt_module_create_from_memory() avoids the deprecation warning
when building ffmpeg with libopenmpt 0.3.
openmpt_module_create_from_memory2() provides more fine-grained error
reporting and in particular allows distinguishing out-of-memory from
input file parsing errors. Return appropriate ffmpeg errors
accordingly.
libopenmpt 0.3 is ABI and API compatible with applications built
against libopenmpt 0.2. Building ffmpeg with libopenmpt 0.2 is still
supported.
Signed-off-by: Jörn Heusipp <osmanx@problemloesungsmaschine.de> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
James Almer [Sun, 7 Jan 2018 03:45:37 +0000 (00:45 -0300)]
avcodec: increase AV_INPUT_BUFFER_PADDING_SIZE to 64
AVX-512 support has been introduced, and even if no functions currently
use zmm registers (able to load as much as 64 bytes of consecutive data
per instruction), they will be added eventually.
Reviewed-by: Rostislav Pehlivanov <atomnuker@gmail.com> Tested-by: Michael Niedermayer <michael@niedermayer.cc> Signed-off-by: James Almer <jamrial@gmail.com>
Apple's AIFF protocol clearly states that each chucnk which is odd sized a padding should be added.
In the old version of aiffdec adding of padding was done in `get_meta`. And in case of unknown chunk name it was done in defalut case.
The new version has deleted the padding in default case and added padding adding after the switch.
But the new version didn't removed the padding adding in the `get_meta` function so in some cases padding was added twice which leaded to a bug.
Fixes: sample.aiff Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
James Almer [Sun, 7 Jan 2018 20:58:33 +0000 (17:58 -0300)]
configure: don't use SDL.h in check_func_headers when checking for SDL2
check_func_headers() defines a main() function, which clashes with a
redefinition done by said SDL header. Check for SDL_PollEvent using
SDL_events.h only instead, where the redefinition doesn't happen.
James Almer [Thu, 4 Jan 2018 18:35:59 +0000 (15:35 -0300)]
avresample: remove deprecated attribute from the AVAudioResampleContext struct
Having all the public functions marked as deprecated is enough.
This gets rid of a warning spam when compiling any file including
libavresample/avresample.h even when avresample is not enabled, like
it's the case with fftools/cmdutils.c