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 7.0 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
<JEEB>
2.8, finally hitting the point where testsrc2 no longer exists :D
<JEEB>
fun, either 3.0 -> 2.8 or testsrc2 -> testsrc bumped memory consumption of the test case from 4 gigs to 8 gigs
<JEEB>
ok, so that is just the diff between testsrc and testsrc2
<JEEB>
(re-tested with 3.0 and testsrc and it's also closer to 8 gigs)
<JEEB>
2.7 also affected
<JEEB>
2.6 also checks out :D
<JEEB>
with the amount of recompiles I'm not sure if I should have just plain utilized git bisect with the whole history... was hoping I'd get it between two releases at least ^^;
* galad
buys JEEB a working time machine
<JEEB>
2.5 and still 8 gigs of memory usage vs around 60 that was until 2023-12 possible to achieve with the input being separate from the filter chain
<JEEB>
*vs around 60 megs
<JEEB>
2.4 and 2.3 too :)
Traneptora_ has joined #ffmpeg-devel
Traneptora has quit [Read error: Connection reset by peer]
cone-691 has joined #ffmpeg-devel
<cone-691>
ffmpeg Lynne master:f55b58782084: aacdec: move aacdec_common to aac/aacdec_tab
<cone-691>
ffmpeg Lynne master:3600f757db3f: aacsbr_template: include mem.h
<cone-691>
ffmpeg Andreas Rheinhardt master:6975d965fcb6: avcodec/aacsbr: Make ff_aac_sbr_* funcs accept ChannelElement*
<cone-691>
ffmpeg Andreas Rheinhardt master:5bd7b8d9995b: avcodec/aacdec: Split SBR context from ChannelElement
<cone-691>
ffmpeg Lynne master:ed009bfd3d2d: aacdec: add a decoder DSP structure
<cone-691>
ffmpeg Andreas Rheinhardt master:fc3c2ea8dce6: avcodec/aacdec: PredictorState array out of SingleChannelElement
<cone-691>
ffmpeg Andreas Rheinhardt master:7117e0c7be6a: avcodec/aacdec: Use same AACDecContext for fixed and float
<cone-691>
ffmpeg Andreas Rheinhardt master:e3693c6f564d: avcodec/aacdec_template: Deduplicate aac_decode_close()
<cone-691>
ffmpeg Andreas Rheinhardt master:8b2261e573b9: avcodec/aacdec: Always include AV(Fixed|Float)DSPContext in AACDecContext
<cone-691>
ffmpeg Andreas Rheinhardt master:980a55fb4677: avcodec/aacdec_template: Deduplicate AVClass+AVOptions
<cone-691>
ffmpeg Andreas Rheinhardt master:9de66fd449eb: avcodec/aacdec_template: Deduplicate common part of aac_decode_init()
<JEEB>
release/1.2 and finally something different, although unfortunately that something is signal 11 :D
<Lynne>
complex filtergraphs are that old?
<JEEB>
yea
<JEEB>
I mean on the other hand 1.2 is just 2013
<JEEB>
meanwhile, I did not expect we'd be getting this amount of snow now
<JEEB>
I think it was around 5cm yesterday it sure ain't stopping this morning
<JEEB>
yea, quickly tested 1.1 as well and it also just signal 11s, so that's no longer as testable. should check tomorrow morning when that new sync logic was added to lavfi
<JEEB>
since that is one of the things that might be relevant
<JEEB>
unless I just hit that with 1.2
<JEEB>
which is why it now segfaults instead
Workl has quit [Read error: Connection reset by peer]
emersion has quit [Changing host]
emersion has joined #ffmpeg-devel
<cone-691>
ffmpeg Zhao Zhili master:309cff378cdb: configure: Lower Android API level requirement of mediandk
<cone-691>
ffmpeg Zhao Zhili master:98b5e80fd698: avcodec/mediacodec_wrapper: link to NDK mediacodec API directly
Krowl has quit [Read error: Connection reset by peer]
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 252 seconds]
psykose has quit [Remote host closed the connection]
psykose has joined #ffmpeg-devel
rvalue- is now known as rvalue
cone-691 has quit [Quit: transmission timeout]
Krowl has joined #ffmpeg-devel
novaphoenix has quit [Quit: i quit]
novaphoenix has joined #ffmpeg-devel
<BBB>
I don't know who's in charge of the linkedin ffmpeg account
<BBB>
but asking netint to contribute a wrapper is fine, only if the sdk-being-wrapped is actually compatibly opensourced
<BBB>
(I understand it's a hw encoder)
<BBB>
if it's a proprietary encoder, I'd rather not upstream it
mkver has joined #ffmpeg-devel
<BBB>
(encoder=sdk)
<BtbN>
Don't those usually fall under the system lib exception, cause driver?
<BBB>
I don't know where to draw that line exactly. but we had capture card sdks and support for that was removed
<BBB>
we also had real (realvideo) encoder contributions and that was rejected as well
jamrial has joined #ffmpeg-devel
<JEEB>
yea I think the SDK needs to be OSS, and then the device should be something actually purchasable or so?
<JEEB>
so real => this is software and not OSS, so nope. random amazon or so specific HW encoder => not available to common folk so nope
<BBB>
something like that, yes
<BBB>
also remember how dll loaders like what mplayer had back in the day were discouraged and rejected
<BBB>
we said we preferred RE'ing opensource implementations
<BBB>
which we did, and I believe it's one of the (many) reasons ffmpeg became so popular
<BBB>
svq1, svq3, that sort of stuff. later mpeg4/h264 decoders
<BBB>
etc.
<BBB>
(mpeg4/h264 was not RE'ed, but the divx3 compatibility of msmpeg4 was helpful
<BBB>
and then divx4/5 compat of mpeg4 also, later
<BBB>
etc.
<BBB>
fundamentally it's not good if we start making closedsourceware easy to use and take over their maintenance burden, or worse, as if we're endorsing it
<JEEB>
yea
jkhsjdhjs has quit [Ping timeout: 256 seconds]
jkhsjdhjs has joined #ffmpeg-devel
microchip__ has joined #ffmpeg-devel
microchip_ has quit [Ping timeout: 240 seconds]
Krowl has quit [Read error: Connection reset by peer]
<Lynne>
alpha fate passes my aac patches, so I'll guess that loongson will pass too
Krowl has joined #ffmpeg-devel
kurosu has joined #ffmpeg-devel
microchip__ has quit [Quit: There is no spoon!]
microchip__ has joined #ffmpeg-devel
qeed has quit [Quit: qeed]
qeed has joined #ffmpeg-devel
microchip__ is now known as microchip_
<jamrial>
Lynne: the aac patches broke msvc
cone-221 has joined #ffmpeg-devel
<cone-221>
ffmpeg James Almer master:63702d5f9c92: avcodec/aac/aacdec: remove double colon
<Lynne>
jamrial: thanks for fixing that
<jamrial>
i have no idea if that's enough. i don't have msvc at hand
<jamrial>
nevcairiel: can you confirm that fixed the failures, or is something else needed?
<nevcairiel>
i'll try to build
<nevcairiel>
seems to have compiled fine
<nevcairiel>
and fate-aac worked, so i suppose its fine
<jamrial>
it really doesn't like double colons then. it printed a billion errors because of it :p
Workl has joined #ffmpeg-devel
Krowl has quit [Ping timeout: 245 seconds]
MikhailAMD has joined #ffmpeg-devel
MikhailAMD has quit [Client Quit]
tufei_ has quit [Remote host closed the connection]
tufei_ has joined #ffmpeg-devel
Workl has quit [Read error: Connection reset by peer]
SystemError has quit [Remote host closed the connection]
SystemError has joined #ffmpeg-devel
<thardin>
got two clients asking about scte 35. I see there's a patchset on the ml, but kieran didn't like it due to lack of PCR extrapolation as called for by the spec
<thardin>
so I may take a stab at a better solution
<JEEB>
scte35 should also be possible to pass into mp4 through emsg packets
j45_ has joined #ffmpeg-devel
j45 has quit [Ping timeout: 268 seconds]
j45_ is now known as j45
j45 has quit [Changing host]
j45 has joined #ffmpeg-devel
kurosu has quit [Quit: Connection closed for inactivity]
j45_ has joined #ffmpeg-devel
j45 has quit [Ping timeout: 245 seconds]
j45_ is now known as j45
j45 has quit [Changing host]
j45 has joined #ffmpeg-devel
j45_ has joined #ffmpeg-devel
j45 has quit [Ping timeout: 260 seconds]
j45_ is now known as j45
j45 has joined #ffmpeg-devel
j45 has quit [Changing host]
j45_ has joined #ffmpeg-devel
j45 has quit [Ping timeout: 260 seconds]
j45 has joined #ffmpeg-devel
j45_ is now known as j45
j45 has quit [Changing host]
j45_ has joined #ffmpeg-devel
j45 has quit [Ping timeout: 268 seconds]
j45_ is now known as j45
j45 has joined #ffmpeg-devel
j45 has quit [Changing host]
Krowl has joined #ffmpeg-devel
j45 has quit [Ping timeout: 245 seconds]
j45 has joined #ffmpeg-devel
j45 has quit [Changing host]
j45 has joined #ffmpeg-devel
<cone-221>
ffmpeg James Almer master:ce33a5f16d7a: avformat/movenc: remove one level of indentation
<JEEB>
(3GPP raised hell after ISO singlehandedly changed the requirements for a spec to be free, as they had provided their stuff under the understanding that DASH would be free)
<JEEB>
oofff, xscte35 emsg example at the end of the doc
<JEEB>
SCTE35... in XML
<JEEB>
right, and DASH-IF-IOP pdf then has an example of a binary SCTE35 emsg message
<ePirat>
also I do not understand what the "non drop frame" and "drop frame" are supposed to mean?
Coinflipper has joined #ffmpeg-devel
<JEEB>
drop frame I think has to do with the whole /1.001 mess?
<JEEB>
in timecodes. which are not the same as timecodes (although matroska has during the years brought up many a people who think timecodes = timestamps)
<JEEB>
*which are not the same as timestamps
<Lynne>
remember when facebook tried to end non-integer framerates?
cone-221 has quit [Quit: transmission timeout]
Krowl has quit [Read error: Connection reset by peer]
<frankplow>
You'd think Microsoft employees might refrain from making demands while proclaiming they are Microsoft employees for a little while, apparently not
<frankplow>
Well to be fair to the guy, "demand" is probably a bit harsh this is more politely worded than that trac issue.
<ePirat>
yeah and also they not using git.ffmpeg.org for vcpkg is probably for the better, load-wise…
<another|>
not sure what they meant by sha512 though
<JEEB>
?
<jamrial>
JEEB: see Javier Matos' email
<JEEB>
ah
<Lynne>
I hadn't even heard of rtmpdump, let alone known we host a repo for it
<JEEB>
it's the classic tool that used to get usage, was related to mplayer IIRC?
<jkqxz>
Lynne: Do you know where to find the Vulkan CTS input samples?
<BtbN>
well, it does
<nevcairiel>
that looks easy enough to fix
<BtbN>
I worked around this now by just manually cloning it to that path
<BtbN>
It's a chicken/egg problem anyway, fate.sh needs to exist first :D
<nevcairiel>
thats why my own scripts around it do the pull as well
<beastd>
BtbN, nevcairiel: If I read the history correctly the switch-case git: differentiation are a leftover from the time we needed to support both svn and git
<beastd>
could be changed to support more prefixes or probably even better make it a catch all if i'm not missing sth
<BtbN>
yeah, assuming it's a git url by default is probably sane
<nevcairiel>
just adding http* to it sounds like a safe change
<nevcairiel>
it tells you to do that, listen to the instructions? =P
<BtbN>
How would I know we maintain an entire wrapper somewhere
<BtbN>
from that message, I had assumed I need to install something via apt
<nevcairiel>
i would expect a dev to just throw a keyword into google, but you found it so whatever :P
<BtbN>
Well, google finds _everything_ for that keyword, but nothing helpful
<BtbN>
An all it finds talks about old Apple stuff
<beastd>
BtbN: hey google, find some gas for my ffmpeg build on windows on arm :P
<beastd>
we should also invent some more tools like leg and doors :)
<Lynne>
jkqxz: yeah... the files are renamed, b8 -> basic8, etc.
<Lynne>
but if you grep for the middle part (e.g. sizeup, or intrabc-extreme), there's a unique match for them in the samples
<beastd>
another|, ePirat, Lynne: Regarding ffmpeg.git stuff. It should be like this: a) primary ffmpeg.git on git.videolan.org b) source.ffmpeg.org is an alias for git.videolan.org c) git.ffmpeg.org a mirror of that repo at a
<jkqxz>
Um, all argon streams files have names like test34637864_43789743.obu.
<jkqxz>
Which are nicely unique to find them, but don't match names with text in them.
<BtbN>
That'd defeat the point of setting up a fate runner for native MSVC
<BtbN>
Though I do intend to set one up for that toolchain as well
<wbs>
yeah I didn't say you shouldn't set up MSVC, just as an addition
<BtbN>
I should probably set out to fix configure in a WSL setting at some point
<wbs>
but any WSL path oddities from trying to set up native windows building with make/shell in WSL probably is the same there
<BtbN>
that issue comes up quite a bit, and shouldn't be that hard to fix
<Lynne>
jkqxz: with valgrind, yeah, nothing
<BtbN>
For now I'll just mod fate.sh to build in-tree
<Lynne>
valgrind does warn there's a single uninitialized jump... in nvidia-egl.so
cone-872 has joined #ffmpeg-devel
<cone-872>
ffmpeg Martin Storsjö master:cdf71f789b2f: Remove .travis.yml
<jkqxz>
I do mean asan. Valgrind would miss overruns on arrays which go into other valid data, which is what would be needed to get a bad value in loop_filter_levels.
<BtbN>
One more curious thing I'm noticing right now is that makeops from the fate config are somehow completely ignored, and I see no good reason as to why
<wbs>
anyway, while it's probably good to improve the use of WSL, it may be worth testing building with the x64 msys as well - it's not a whole lot of code that needs to be emulated, so it can very well be that it's faster than with WSL. or not, who knows
<wbs>
did you set up WSL1 or 2?
<jkqxz>
I can't see anything nearby which looks like it could overrun, though, so it's a bit of a wild guess.
<BtbN>
WSL1 is dead since like 5 years or so
<BtbN>
I doubt it was ever implemented for aarch64
<wbs>
it was
<BtbN>
It at the very least can't run modern distros anymore
<wbs>
none of the windows/arm64 devices I have are capable of WSL2 (prototype HW and such where virtualization isn't enabled, and an early first gen commercial device where there was no virtualization available)
<nevcairiel>
"dead" is such an overused word, WSL1 still works perfectly fine
<BtbN>
modern ones crash due to glibc using syscalls they never implemented
<nevcairiel>
personally i really prefer how WSL1 works, i dont need a hypervisor sitting on my system constantly, too bad they gave up and used the easier method
<wbs>
yep
<beastd>
+1 for WSL1
<nevcairiel>
anyway i dont need a "modern" distribution for the things i do, so i'll stick with whatever LTS I have installed in there =P
<BtbN>
well, even Ubuntu 22.04 and Debian 12 are too modern