michaelni changed the topic of #ffmpeg-devel to: Welcome to the FFmpeg development channel | Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg | This channel is publicly logged | FFmpeg 7.1.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
cone-140 has joined #ffmpeg-devel
<cone-140>
ffmpeg Andreas Rheinhardt master:c914fd2977e9: avformat/av1dec: Remove redundant avio_tell()
<cone-140>
ffmpeg Andreas Rheinhardt master:70fa44dfa84a: avcodec/vulkan_encode_h264: Fix memleak on error
<cone-140>
ffmpeg Andreas Rheinhardt master:85887703670e: avformat/av1dec: Avoid branch for setting position
<cone-140>
ffmpeg Andreas Rheinhardt master:b18aaf209f00: fftools/ffmpeg_dec: Always receive frames from decoder
<cone-140>
ffmpeg Andreas Rheinhardt master:f71d0f0559c2: avcodec/apv_parser: Mark close as av_cold
<cone-140>
ffmpeg Andreas Rheinhardt master:d794ecd9d2d2: avcodec/vulkan_encode_hevc: Fix memleak on error
<cone-140>
ffmpeg Andreas Rheinhardt master:0d7172a9ffd1: avcodec/cbs: Avoid branch
arch1t3cht has quit [Read error: Connection reset by peer]
arch1t3cht1 has joined #ffmpeg-devel
pross has joined #ffmpeg-devel
j45_ has joined #ffmpeg-devel
<fflogger>
[newticket] uttam_32472: Ticket #11589 ([avcodec] null check for AVSideDataDescriptor* in ffmpeg_enc.c) created https://trac.ffmpeg.org/ticket/11589
j45 has quit [Ping timeout: 276 seconds]
j45_ is now known as j45
j45 has quit [Changing host]
j45 has joined #ffmpeg-devel
ChitraMicrosoft has quit [Quit: Client closed]
ChitraMicrosoft has joined #ffmpeg-devel
ChitraMsft has joined #ffmpeg-devel
kasper93 has quit [Read error: Connection reset by peer]
\\Mr_C\\ has joined #ffmpeg-devel
kasper93 has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
kasper93_ has joined #ffmpeg-devel
kasper93 is now known as Guest4931
kasper93_ is now known as kasper93
kasper93_ has joined #ffmpeg-devel
kasper93 is now known as Guest357
Guest357 has quit [Killed (calcium.libera.chat (Nickname regained by services))]
kasper93_ is now known as kasper93
kasper93_ has joined #ffmpeg-devel
kasper93 is now known as Guest4185
kasper93_ is now known as kasper93
Guest4931 has quit [Ping timeout: 260 seconds]
Guest4185 has quit [Ping timeout: 248 seconds]
cone-197 has joined #ffmpeg-devel
<cone-197>
ffmpeg Michael Niedermayer master:12b853530a68: Add tools/merge-all-source-plugins
ChitraMicrosoft has quit [Quit: Client closed]
microlappy has joined #ffmpeg-devel
ChitraMicrosoft has joined #ffmpeg-devel
ChitraMicrosoft7 has joined #ffmpeg-devel
ChitraMicrosoft has quit [Client Quit]
ChitraMsft has quit [Quit: Client closed]
<Lynne>
av1+dovi encodes are on the pirate bay already
<Lynne>
they move fast
ChitraMsft has joined #ffmpeg-devel
<Lynne>
its not always going to be xvid+aac in avi
minimal has joined #ffmpeg-devel
Guest93 has joined #ffmpeg-devel
ChitraMsft has quit [Quit: Client closed]
ChitraMsft has joined #ffmpeg-devel
\\Mr_C\\ has quit [Remote host closed the connection]
microlappy has quit [Quit: Konversation terminated!]
<ePirat>
I get an assertion in ffmpeg_filter.c in fftools when I use for a file with video and audio: ffmpeg -loglevel verbose -i input.mp4 -filter_complex 'noop' -f null -
<ePirat>
Assertion best_input >= 0 failed at fftools/ffmpeg_filter.c:2194
<ePirat>
It seems related to the issue I hit yesterday with missing frames
<Lynne>
ChitraMicrosoft: is this about the j2k code?
<ChitraMicrosoft>
Its a patch I submitted that is being reviewed by Michael in which there is feedback on the extra copying being done. I am trying to figure out a solution to reduce the number of copies . I implemented the first set of feedback comments
<ChitraMicrosoft>
use perf, if loads are hot then you've got too much memory usage
<ChitraMicrosoft>
- Can you elaborate? Is that a preprocessor directive?
<Lynne>
linux-perf
<ChitraMicrosoft>
Got it
<Lynne>
view the disassembly
<ChitraMicrosoft>
view the disassembly
<ChitraMicrosoft>
- Will do .Do you guys have any feedback on the proposed patch?
<Lynne>
you can only speed up j2k on the CPU so much, since the main bottleneck is still the incredibly slow arithmetic decoder
<Lynne>
a GPU decoder would be much faster
<ChitraMicrosoft>
Agree
<Lynne>
if there's interest, I can write one in vulkan
<ChitraMicrosoft>
if there's interest, I can write one in vulkan
<ChitraMicrosoft>
- Do you mean refactoring dwt_decode97_float? This is the bottleneck function for us that is impacting FPS performance . Happy to take any help you can offer :)
<ChitraMicrosoft>
In my suggested change michaelni had some feedback on the extra copying I am doing . If there is a better solution we would be very interested
ChitraMicrosoft has quit [Quit: Client closed]
ChitraMicrosoft has joined #ffmpeg-devel
halloy5771 has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
<ChitraMicrosoft>
Is Vulkan natively supported by Applu GPUs
halloy5771 has quit [Read error: Connection reset by peer]
halloy5771 has joined #ffmpeg-devel
halloy5771 has quit [Read error: Connection reset by peer]
<ramiro>
wbs: in libavutil/aarch64/asm.S, the function macro has a default alignment of 2. shouldn't this be at least 4 for 16-byte alignment? (or better yet 6 for a cache-friendly 64-byte alignment)
halloy5771 has joined #ffmpeg-devel
<llyyr>
ch
<llyyr>
ChitraMicrosoft: no
System_Error has joined #ffmpeg-devel
<wbs>
ramiro: I guess it could
halloy5771 has quit [Read error: Connection reset by peer]
halloy5771 has joined #ffmpeg-devel
<BtbN>
Just got an interesting real-life bug report about the rtsp muxer... Some video wall is refusing to play an rtsp stream sent to it by ffmpeg, while ffplay/vlc play it just fine. Looking at the stream in wireshark, Wireshark does indeed claim that what ffmpeg is muxing is malformed, pointing at the sprop-parameter-sets= entry.
<cone-800>
ffmpeg James Almer release/7.1:7eaa8c110e83: avcodec/h264_slice: insert LCEVC side data before get_buffer() call
<BtbN>
The condition is if(sps && sps_end - sps >= 4 && p - psets <= MAX_PSET_SIZE - strlen(profile_string) - 7). The last part should be fine, given there is not even nearly 1024 bytes following it, and there's also definitely two datasets with more than 4 bytes each there. Sooo, why ain't it there.
<BtbN>
oh nvm, it's _before_ it
<BtbN>
that stream then can't be coming from ffmpeg, unless I'm missing something, huh.
halloy5771 has quit [Read error: Connection reset by peer]
<Lynne>
ChitraMicrosoft: it works
<ChitraMicrosoft4>
Is refactoring that part of the code using Vuklan an option Lynne ?
<Lynne>
not refactoring, writing a new decoder entirely in vulkan
<ChitraMicrosoft4>
I'm new to this area , is Vulkan better than CUDA?
<ChitraMicrosoft4>
Also Is this a lengthy work
ChitraMicrosoft4 has quit [Quit: Client closed]
ChitraMicrosoft8 has joined #ffmpeg-devel
<Lynne>
well it runs on all vendors
<Lynne>
and yes, it would be lengthy work
<ChitraMicrosoft8>
This would be very interesting to us at Microsoft as a lot of tools like HandBrake , Blender heavily rely on FFMPEG but I am not sure how your development process works
<ChitraMicrosoft8>
Do you need approval to do this , and are you open to taking feature requests like these
<ChitraMicrosoft8>
Please let me know if you guys would like to contnue engaging with us
<Lynne>
yes, I'd like to do it, though I'd prefer to get compensated for it
<ChitraMicrosoft8>
Checking internally
<ChitraMicrosoft8>
Give me some time
halloy5771 has joined #ffmpeg-devel
ChitraMicrosoft has quit [Quit: Client closed]
ChitraMicrosoft has joined #ffmpeg-devel
ChitraMicrosoft8 has quit [Quit: Client closed]
ChitraMicrosoft has quit [Quit: Client closed]
JEEB has joined #ffmpeg-devel
appaji has joined #ffmpeg-devel
halloy5771 has quit [Read error: Connection reset by peer]
<appaji>
hello, if i received a reply on my patch, can i reply to it through gmail?
<appaji>
I replied through gmail, but it's not visible on the mailing list page. Did I mess it up?
<BtbN>
It's a mailing list. So you can reply _only_ via mail
<appaji>
I mean, do I have to use any different client
<appaji>
I dont know why my reply is not visible on the mailing list page.
<JEEB>
nope, as long as your from is the email that you registered it should be fine
<appaji>
okay, thanks!
<appaji>
I see it now, it took a bit of time hehe
<JEEB>
:)
<ramiro>
wbs: what's your latest branch with github actions to run all tests?
<wbs>
(there are native arm runners these days; I have experimented with taking that into use as well, but those test jobs there don't use them quite yet)
appaji has quit [Quit: Konversation terminated!]
<ramiro>
wbs: thanks!
JEEB has quit [Ping timeout: 240 seconds]
halloy5771 has joined #ffmpeg-devel
ChitraMicrosoft has joined #ffmpeg-devel
halloy5771 has quit [Read error: Connection reset by peer]
halloy5771 has joined #ffmpeg-devel
ChitraMicrosoft has quit [Quit: Client closed]
halloy5771 has quit [Read error: Connection reset by peer]
<cone-800>
ffmpeg softworkz master:8f42d90413f6: fftools/textformat: Apply formatting and whitespace changes
<cone-800>
ffmpeg softworkz master:1f2b8d7238ef: fftools/graphprint: Now, make it a Killer-Feature!
halloy5771 has joined #ffmpeg-devel
<ramiro>
oh, wow. are we really calling system(cmd)?
<mkver>
I always wanted to look at this patchset; always tomorrow.
<mkver>
I can't believe that he really used "Now, make it a Killer-Feature!" as a commit message.
<jamrial>
was this reviewed at all?
<BtbN>
Not properly I guess, it is/was a huge patchset, and reviewing it is annoying with that Github-Forwarder
<BtbN>
Why on earth does it even need to open a browser?
<BtbN>
Just print the bloody url
<BtbN>
And yeah, that commit message is super odd. On the ML I see a bunch of unresolved comments as well.
<jamrial>
someone was a bit too eager to push i guess
<jamrial>
michaelni: maybe don't give push access to people so quickly...
<ramiro>
on my first commit I broke pcm for big-endian CPUs :P
<jamrial>
breaking big endian is a rite of passage
<ramiro>
lol
<jkqxz>
"SUMMARY: AddressSanitizer: 33052 byte(s) leaked in 39 allocation(s)." Seems to be all under graphprint.c.
halloy5771 has quit [Read error: Connection reset by peer]
<jamrial>
please point that out in the thread. lets give him a chance to tidy things up before we go the nuclear route
BradleyS has quit [Quit: quit]
mkver has quit [Ping timeout: 248 seconds]
<jkqxz>
At least if you pass the sg option then it accidentally overwrites the user-supplied file name so it can't be used directly for shell injection.
BradleyS has joined #ffmpeg-devel
<ramiro>
yes, let's not go nuclear. I had to resort to ChatGPT to sugarcoat it. (I believe) we can do it.
minimal has quit [Quit: Leaving]
<BtbN>
The last commit specifically imo needs at least partially reverted. system() calls are just a huge no
<BtbN>
Replace that whole function with just a printf of the URL, and it's fine enough
<ramiro>
jkqxz: bobby tables :)
ChitraMicrosoft has joined #ffmpeg-devel
halloy5771 has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
<BBB>
hm bobby tables :)
AtleoS has joined #ffmpeg-devel
<cone-800>
ffmpeg James Almer release/4.4:22994d2ceebb: avcodec/libtheora: fix setting keyframe_mask
<cone-800>
ffmpeg Michael Niedermayer release/4.4:e954f5271314: swscale/output: Fix integer overflow in yuv2gbrp_full_X_c()
<cone-800>
ffmpeg Michael Niedermayer release/4.4:576353a90098: configure: Clearer documentation for "disable-safe-bitstream-reader"
<cone-800>
ffmpeg Michael Niedermayer release/4.4:bc6cf25378a8: avformat/hls: Add cmfv and cmfa to allowed_extensions
<cone-800>
ffmpeg Michael Niedermayer release/4.4:08d68379a622: avformat/hls: Add ec3 to allowed_extensions
<cone-800>
ffmpeg Michael Niedermayer release/4.4:5046edebecd8: avformat/hls: add fmp4 to allowed_extensions
<cone-800>
ffmpeg Michael Niedermayer release/4.4:ddaf2ff45892: avformat/hls: Fix Youtube AAC
<cone-800>
ffmpeg Michael Niedermayer release/4.4:3ef6173e0bb4: avformat/hls: Split allowed_segment_extensions off allowed_extensions
<cone-800>
ffmpeg Michael Niedermayer release/4.4:d2d823205655: avformat/hls: Fix flash1.bogulus.cfd support
<cone-800>
ffmpeg Michael Niedermayer release/4.4:f34dd12844e4: libpostproc: check minimum size
<cone-800>
ffmpeg Michael Niedermayer release/4.4:a3fb72c6901f: avcodec/vorbisdec: Dont treat overread as error
<cone-800>
ffmpeg Michael Niedermayer release/4.4:eb74fe6908ed: avcodec/h264_mb: Fix tmp_cr for arm
<cone-800>
ffmpeg Michael Niedermayer release/4.4:32527e141b73: avformat/avidec: Ignore duplicate GAB2
<cone-800>
ffmpeg Michael Niedermayer release/4.4:04b03f7cf388: avformat/mov: reject negative ELST durations
<cone-800>
ffmpeg Michael Niedermayer release/4.4:29561e3758d4: avcodec/sonic: Check num_taps
<cone-800>
ffmpeg Michael Niedermayer release/4.4:c284ef000a52: avcodec/svq3: Check there are bits left before decompression
<cone-800>
ffmpeg Michael Niedermayer release/4.4:d6e02ba2013f: avcodec/takdec: Check remaining space for first predictors
<cone-800>
ffmpeg Michael Niedermayer release/4.4:44b04492bfc8: Changelog: Update