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 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
Mirarora has quit [Quit: Mirarora encountered a fatal error and needs to close]
<BtbN>
Yeah, it's a 50/50 chance for a busy loop deadlock when hitting q
<BtbN>
Also, trace_headers seems to suggest that "Three Dimensional Reference Displays Information" is present, but largely all 0.
<BtbN>
It's also not part of the extradata
<BtbN>
unless it's just not logged there
<BtbN>
The ff_isom_write_hvcc stuff is able to write those things fine, right?
iive has quit [Quit: They came for me...]
Mirarora has joined #ffmpeg-devel
<BtbN>
jamrial: I'm looking at the Apple-Spec, and it reads like it forces you to write a generic Ref Disp Info SEI into the config record/extradata? Like, it says "write THESE values", which is a bit odd. Would definitely make it possible to write it at init time, if you assume fixed values for the view IDs, which nvenc might be able to do, given it controls them.
<jamrial>
BtbN: it's not going to be in extradata because nvenc will export it with the first frame/packet
<BtbN>
The spec calls for it to be in the extradata tho
<jamrial>
the only way the muxer would attempt to create extradata from a packet's contents is if there's no extradata at all
<BtbN>
And nvidia in a mail said they write it according to the spec
<BtbN>
What I did notice as well is that the spec calls for an lhvc tag to be written, which the mov muxer looks at AV_DISPOSITION_MULTILAYER for. Which is not set anywhere except via parameters or from the mov demuxer
<jamrial>
yeah, that needs to be set with a muxer option
<BtbN>
so there is no way to signal that from an encoder?
<jamrial>
or rather, stream option
<jamrial>
the user (so, the cli) could do it
<BtbN>
That's just not very intuitive, that you have to think about that
<jamrial>
well, avcodeccontext and avstream are different things
<BtbN>
Yeah, it'd be something ffmpeg.c would need to negotiate, based on info gotten from the encoder
<jamrial>
numOfArrays: 0 as you can see in that paste
<BtbN>
It's missing the other PPSs is what I gathered when comparing the trace vs. the spec
<jamrial>
yeah
<jamrial>
no wait, wtf, they are in the second packet
<jamrial>
god, what a mess
<BtbN>
oh god
<BtbN>
I'm starting to wonder if we need to start considering to never use nvencs global headers
<jamrial>
we can't work with this. all the parameter sets should be in the first packet/AU, and of course in the output of nvEncGetSequenceParams() that goes to extradata
<BtbN>
But if it's not even in the first packet, all bets are off
<jamrial>
BtbN: it wouldn't change the fact the second layer param sets are in the second packet
<jamrial>
yeah
<BtbN>
looking at that issue with the busy-looping, nvenc might consider two frames to be one packet?
<jamrial>
could be
<BtbN>
i.e. in MV mode, you need to feed in two frames, and then lock two bitstream buffers, and combine them?
<BtbN>
That's not mentioned anywhere though, and their engineers writing the patch clearly didn't realize that caveat either
cone-001 has quit [Quit: transmission timeout]
<BtbN>
regardless, the lhvc writing (or parsing?) should probably be fixed to not write unparseable files when the multilayer disposition is set but no multiple layers appear
<jamrial>
BtbN: there
<BtbN>
ah, so it is actually a parser issue
<jamrial>
empty lhvc boxes are valid if you use the hev1 brand (default), but not with hvc1, which afaik is required by the apple spec
thilo has quit [Ping timeout: 260 seconds]
thilo has joined #ffmpeg-devel
thilo has quit [Changing host]
thilo has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
Marth64 has joined #ffmpeg-devel
<Marth64>
hstack and vstack are broken starting somewhere within HEAD~50
<Marth64>
when combined with scale. i will make ticket
<Marth64>
e61b9d4094d76f5fede0b65139d17add281e6e45 fftools/ffmpeg: propagate decoded_side_data from decoded streams to the filterchain
<fflogger>
[newticket] Marth64: Ticket #11442 ([undetermined] Scale followed by vstack stopped working) created https://trac.ffmpeg.org/ticket/11442
tufei_ has quit [Quit: Leaving]
tufei has quit [Remote host closed the connection]
Mirarora has quit [Quit: Mirarora encountered a fatal error and needs to close]
paulk has quit [Ping timeout: 260 seconds]
paulk has joined #ffmpeg-devel
tufei has joined #ffmpeg-devel
secondcreek has quit [Remote host closed the connection]
mkver has quit [Ping timeout: 246 seconds]
^Neo has quit [Ping timeout: 248 seconds]
omegatron has quit [Quit: Power is a curious thing. It can be contained, hidden, locked away, and yet it always breaks free.]
Mirarora has joined #ffmpeg-devel
cone-100 has joined #ffmpeg-devel
<cone-100>
ffmpeg James Almer master:c6194b50b1b4: avdevice/alsa_dec: add a ch_layout option
<klaxa>
hi, i happen to be at fosdem this year. i read about an ffmpeg community meetup on the mailinglist but no dates and times yet?
* another|
waves at klaxa
<another|>
AFAIK there are no times, yet
DauntlessOne4 has quit [Ping timeout: 246 seconds]
<fflogger>
[editedticket] fildens: Ticket #11430 ([avformat] [Regression] Data stream in output may glitch "-stats" display since FFmpeg 7?) updated https://trac.ffmpeg.org/ticket/11430#comment:9
SincereLeader has joined #ffmpeg-devel
<SincereLeader>
talk is cheap, send patches
^Neo has joined #ffmpeg-devel
^Neo has quit [Changing host]
^Neo has joined #ffmpeg-devel
^Neo has quit [Ping timeout: 248 seconds]
mkver has joined #ffmpeg-devel
elChippo has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg-devel
microchip_ has quit [Client Quit]
microchip_ has joined #ffmpeg-devel
microchip_ has quit [Remote host closed the connection]
microchip_ has joined #ffmpeg-devel
paulk has quit [Ping timeout: 244 seconds]
paulk has joined #ffmpeg-devel
microchip_ has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg-devel
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 245 seconds]
rvalue- is now known as rvalue
jamrial has joined #ffmpeg-devel
^Neo has joined #ffmpeg-devel
^Neo has quit [Changing host]
^Neo has joined #ffmpeg-devel
ccawley2011 has joined #ffmpeg-devel
<thardin>
hum. BestLeader left. but yeah the index table segments contain frame flags, including whether they're forward/back references
<SincereLeader>
thardin: the ffv1 in mxf need to be marked with AV_PKT_FLAG_KEY otherwise -c:v copy will not work
<thardin>
do you mean muxer or demuxer?
<jamrial>
huh, AV_PKT_FLAG_KEY is not set at all in mxfdec? all this time it relied on avparsers setting it?
<jamrial>
thardin: demuxer. if no packet is flagged as key, stream copy will not work
<jamrial>
there's no avparser for ffv1 i see,
<jamrial>
one could be writte, that said
<thardin>
hm. well the file should have keyframe flag set in its index table segments
<SincereLeader>
demuxer needs to set AV_PKT_FLAG_KEY for keyframes, and MXF already have support for this
<thardin>
but I guess if the file has no index or the index is wrong then yeah
<jamrial>
if there's no index then that's ok, don't set anything
<SincereLeader>
stop allocations/frees all the time
<kierank>
SincereLeader: that's not the issue
<SincereLeader>
it cause delays with pools
<kierank>
not a delay of 20ms or 40ms
<kierank>
issue is pipeline buffering frames
ccawley2011 has joined #ffmpeg-devel
<SincereLeader>
meeting was ultra short with two leaders
<kierank>
While Gogs was open-source, its repository was controlled by a single maintainer, which limited community input and development speed. In response to these limitations, the Gitea developers decided to fork Gogs in November 2016, creating a community-driven development model.
<kierank>
lol, the irony of using a fork of gogs because of an overbearing maintainer
System_Error has quit [Remote host closed the connection]
<kierank>
jamrial: what is the purpose of an ffv1 parser exactly?
<kierank>
is ffv1 stored fragmented in any containers?
<SincereLeader>
it just parse key frames
<jamrial>
kierank: setting the key frame flag on packets. it doesn't packetize
<SincereLeader>
so codec copy works with ffmpeg
<SincereLeader>
and with mxf
<SincereLeader>
but not limited to
<kierank>
strange, surely the source container sets that
<kierank>
but ok
<Marth64>
open thought: Can pattern of demuxer that concatenates segments with subdemuxer be abstracted and made generic? thinking of possibility to consolidate boilerplate in hls, dash, dvd, concat. such boilerplate includes whitelist checking, timestamp correction, etc.
<jamrial>
kierank: mxf demuxer currently doesn't flag keyframes
<fflogger>
[editedticket] Marth64: Ticket #11430 ([avformat] [Regression] Data stream in output may glitch "-stats" display since FFmpeg 7?) updated https://trac.ffmpeg.org/ticket/11430#comment:10
mkver has joined #ffmpeg-devel
Tanay has joined #ffmpeg-devel
quietvoid has quit [Ping timeout: 245 seconds]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
tufei_ has joined #ffmpeg-devel
Tanay has quit [Remote host closed the connection]
SincereLeader has quit [Quit: Client closed]
ccawley2011 has joined #ffmpeg-devel
tufei_ has quit [Quit: Leaving]
tufei has quit [Remote host closed the connection]