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
minimal has quit [Quit: Leaving]
hpkn has quit [Ping timeout: 244 seconds]
hpkn has joined #ffmpeg-devel
<another|> l
<BtbN> Turns out the openssl.exe I was using for testing plain doesn't work...? It was working fine since a while when using WSL openssl or plain ffmpeg itself as dtls client.
<BtbN> that's very strange
<BtbN> I wonder if openssl 3.5 (which is the windows exe) dropped DTLS 1.2 in favour of 1.3, which SChannel does not implement
<kasper93> sounds like fun adventure
<BtbN> The next adventure will be figuring out if I want to implement that "download pem file from URL and use it"
<BtbN> for the sake of compatibility, it'd be nice, cause it matches how every other TLS implementation handles certs
<BtbN> But schannel really wants to get certificates from the users cert store by name
aaabbb has quit [Changing host]
aaabbb has joined #ffmpeg-devel
secondcreek has quit [Remote host closed the connection]
secondcreek has joined #ffmpeg-devel
user23 has quit [Read error: Connection reset by peer]
secondcreek has quit [Remote host closed the connection]
secondcreek has joined #ffmpeg-devel
witchymary has quit [Remote host closed the connection]
witchymary has joined #ffmpeg-devel
khrbtxyz has joined #ffmpeg-devel
averne has quit [Ping timeout: 272 seconds]
averne has joined #ffmpeg-devel
hpkn has quit [Remote host closed the connection]
hpkn has joined #ffmpeg-devel
<compnn> just call it linux only feature and move on :D
Martchus_ has joined #ffmpeg-devel
jamrial has quit []
Martchus has quit [Ping timeout: 276 seconds]
bbbccc has quit [Quit: ZNC 1.9.0 - https://znc.in]
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
TheVibeCoder has joined #ffmpeg-devel
bbbccc has joined #ffmpeg-devel
cone-092 has joined #ffmpeg-devel
<cone-092> ffmpeg Peter Ross master:3b9dbda49bb1: tools/pktdumper: dump extradata buffers
mkver has joined #ffmpeg-devel
MisterMinister has quit [Ping timeout: 248 seconds]
ngaullier has joined #ffmpeg-devel
Anthony_ZO has joined #ffmpeg-devel
<ngaullier> Hi, patchwork raised an error in the middle of my patch serie, but I really think this is a fake alarm since the failed test is not related to the patch of concern. Any feeling about this ?
<JEEB> that seems to be executed on github CI, I wonder how simple it would be to re-run it
mateo` has quit [Quit: WeeChat 4.6.3]
mateo` has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
cone-092 has quit [Quit: transmission timeout]
Anthony_ZO has quit [Remote host closed the connection]
<TheVibeCoder> please allow vibe coders to contribute back to FFmpeg!
HarshK23 has joined #ffmpeg-devel
<kasper93> it's so easy to spot, because LLMs like to add bullshit comments, in completelly obvious code that don't need them. (above others standard patterns)
<kasper93> I'm not against AI use, but it needs to be mostly human work, with help of AI, not drop of random LLMs output.
<ePirat> tbh I dont find the comments in that patchset too unreasonable...
<ePirat> but I really like to comment code quite a bit so...
<ePirat> whats a bit more weird is the different style of comments
<kasper93> I don't mean comments are bad, they are important for code readability. But `allocating buffers` comments to av_malloc calls is adding more noise than information, imho
<TheVibeCoder> patch uses bilinear sws rescaling, into the trahs it goes
hpkn has quit [Ping timeout: 244 seconds]
hpkn has joined #ffmpeg-devel
jamrial has joined #ffmpeg-devel
<ePirat> kasper93, yeah
<beastd> comments aren't necesarrily bad. they have mostly gone out of fashion because of the silly examples (`i = i + 1 // add 1 to i`) and talking them down in general.
<beastd> comments about non-obvious or not-so-obvious-for-most-readers details are pretty important and unfortunately often only make it into the commit message. or somtimes not event there :(
<ePirat> it also makes reviews a lot easier
<llyyr> the point wasnt that comments are bad, just that the comments in that code aren't helpful. humans won't add add a comment saying "Rec.709 to Rec.2020 matrix" above a variable called bt709_to_bt2020
<llyyr> LLM generated code often adds comments like that, while leaving the parts that actually needs comments without them
hpkn_ has joined #ffmpeg-devel
hpkn has quit [Ping timeout: 244 seconds]
hpkn_ is now known as hpkn
<beastd> llyyr,kasper93: yes i understood that point. and while there is some overlap, i agree it's a strong indicator "AI" might have been involved
IndecisiveTurtle has joined #ffmpeg-devel
Xe has quit [Ping timeout: 276 seconds]
Xe has joined #ffmpeg-devel
<fflogger> [newticket] StevenAllanSpielber: Ticket #11653 ([undetermined] -c copy does not copy track names) created https://trac.ffmpeg.org/ticket/11653
<TheVibeCoder> what is OpenHarmony ?
<kasper93> Chinese operating system
<JEEB> user space APIs where it's either under linux kernel or their own microkernel thing
<JEEB> (first being android-based)
<JEEB> and then the latter has its own kernel and no more android compat out of the box
<kasper93> They got far in few years they are alvie
<kasper93> alive*
<fflogger> [editedticket] JackLau: Ticket #11653 ([undetermined] -c copy does not copy track names) updated https://trac.ffmpeg.org/ticket/11653#comment:1
Thulinma has quit [Ping timeout: 260 seconds]
<fflogger> [editedticket] StevenAllanSpielber: Ticket #11653 ([undetermined] -c copy does not copy track names) updated https://trac.ffmpeg.org/ticket/11653#comment:2
ramiro has quit [Ping timeout: 260 seconds]
ramiro has joined #ffmpeg-devel
<BtbN> Can anyone figure out what the "fingerprint" option added by whip to the shared tls options does? I can't see it being used anywhere
bsFFFFFF has joined #ffmpeg-devel
hpkn has quit [Ping timeout: 244 seconds]
bsFFFFFF has quit [Ping timeout: 260 seconds]
bsFFFFFF has joined #ffmpeg-devel
bsFFFFFF has quit [Ping timeout: 276 seconds]
bsFFFFFF has joined #ffmpeg-devel
hpkn has joined #ffmpeg-devel
<BtbN> Also, wtf is setup_srtp() in whip.c? It queries some "materials" from the TLS layer(read: openssl), assumes their exact size and layout, and then does its own crypto with it...?
<kierank> lol
<jamrial> maybe we should revert that until it gets a proper review...
<BtbN> At least partially
<BtbN> though all the changes it did to the generic TLS code are also quite wild
<BtbN> and are largely whip internas leaking out
<BtbN> Though I have no idea how srtp works? The code for that was pre-existing it seems. And really does want that key material?
<BtbN> And apparently that magic keying material follows an RFC, that both openssl quotes for its SSL_export_keying_material() and https://learn.microsoft.com/en-us/windows/win32/secauthn/querycontextattributes--schannel reference: RFC5705
<kierank> 17:19:15 <•jamrial> maybe we should revert that until it gets a proper review...
* kierank said that
<BtbN> It violates so many basic things that just based on that a it probably shouldn't have been merged
<kierank> it's experimental at least
<kierank> so I think it's fine
<BtbN> I can't tell yet if it's fundamentally fine or flawed, but there's a lot of strange things everywhere
<BtbN> I have accumulates over 20 commits yet that just clean stuff up with no functional changes
<beastd> BtbN: In that case we need to weigh if your clean ups will make it right or if revert is the better route
<BtbN> I need to first find out how to actually use the thing
<kasper93> I see 2 years ago there was whip/whep patchset on ML, maybe that was better quality? (no I didn't look into details)
<BtbN> You seem to need DTLS to do it
<BtbN> Wonder how if solved that
<BtbN> *it
<JEEB> wasn't it the set using some 3rd party lib?
<BtbN> that would side-step the issue
<BtbN> though I'm a fan of adding DTLS support to avformat
<BtbN> it's a useful thing to have, and the implementation done in tls_openssl looks fine, outside of all the random whip specific code that got added to it
<kasper93> JEEB: you are right, it was just an adapter for https://github.com/paullouisageneau/libdatachannel
<kasper93> (not that using 3rdparty libs is wrong)
<JEEB> yea, not saying it's wrong
<kasper93> no support for schannel though :^)
averne has quit [Ping timeout: 260 seconds]
averne has joined #ffmpeg-devel
mkver has quit [Ping timeout: 268 seconds]
mkver has joined #ffmpeg-devel
averne_ has joined #ffmpeg-devel
averne has quit [Ping timeout: 244 seconds]
averne_ is now known as averne
<TheVibeCoder> revert all the things
<kierank> TheVibeCoder: revert all paul mahol code
<kierank> he needs auditor
<kierank> not real person
<another|> Now Paul needs to come to VDD to verify his identity /s
<TheVibeCoder> sure
<TheVibeCoder> how one develop/test ARM SIMD for FFmpeg?
<kierank> mac mini access
<kierank> I think apple locked me out
<kierank> need to reflash
<kierank> or ask el presidente and me for access to ampere
Thulinma has joined #ffmpeg-devel
ngaullier has quit [Remote host closed the connection]
<BtbN> You can get a super cheap ARM VM from Hetzner as well. Not sure how modern the architecture is, but stuff like NEON is there on those Ampere CPUs
MisterMinister has joined #ffmpeg-devel
bsFFFFFF has quit [Ping timeout: 276 seconds]
System_Error has quit [Remote host closed the connection]
<Lynne> get a cheap SoC with in+out of order cores
bsFFFFFF has joined #ffmpeg-devel
<TheVibeCoder> no coins left
<TheVibeCoder> only emulators
<TheVibeCoder> are way
System_Error has joined #ffmpeg-devel
<Lynne> not enough suffering in NEON SIMD, write GPU GLSL code instead
<BtbN> hm, is it acceptable if the tls parameters are implementation specific?
<BtbN> SChannel really does not want to load pem files, like the current API wants to
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
<fflogger> [editedticket] carinobra: Ticket #5419 ([avformat] HLS EXT-X-DISCONTINUITY tag is not supported) updated https://trac.ffmpeg.org/ticket/5419#comment:34
___nick___ has joined #ffmpeg-devel
___nick___ has quit [Client Quit]
___nick___ has joined #ffmpeg-devel
<kasper93> TheVibeCoder: add FUNDING.yml easy coin
<TheVibeCoder> seriously: what to use for donations? paypal, something better?
bsFFFFFF has quit [Ping timeout: 248 seconds]
bsFFFFFF has joined #ffmpeg-devel
<compnn> TheVibeCoder, depends what country you are in
<compnn> and how useful your banks are
<compnn> and how easy it is for you to convert crypto in your country
<compnn> TheVibeCoder, so the best option is have a mailbox and list address and ask people to send cash.
<compnn> but that depends on how honest your mailman is...
<TheVibeCoder> lol
<TheVibeCoder> so i get special mails with poisonous content from sane followers
<compnn> just post incoming mails to pastebin
<compnn> enough summarizing them
<compnn> or pay someone to open them
<compnn> :D
<TheVibeCoder> and put such person at risk of content of some crazy lovers
<TheVibeCoder> compnn: still using mplayer?
<compnn> yes but not master because its currently broken
<TheVibeCoder> post a fix
<compnn> also using old version of mpv, from when wm4 was leader
<TheVibeCoder> lol
Everything has joined #ffmpeg-devel
<TheVibeCoder> compnn: so you trust and use only software that have strong leaders as maintainers?
<compnn> TheVibeCoder, what do you use ?
<TheVibeCoder> compnn: i use librempeg
<TheVibeCoder> you should too
<compnn> as a player
<TheVibeCoder> yes
<compnn> should call it librempig
<TheVibeCoder> and also lavfi-preview
___nick___ has quit [Ping timeout: 268 seconds]
<TheVibeCoder> compnn: feel free to fork it
<compnn> TheVibeCoder, i generally dont use forks. because the people who fork large projects ...
<TheVibeCoder> yes?
<compnn> lets just say i havent seen any stable forks
<TheVibeCoder> this on is stable, very very stable
<compnn> is the maintainer stable , in his head ?
<compnn> be aware, i've talked to you for 10+ years...
<TheVibeCoder> no, because he is currently wasting time
<compnn> yes.
<compnn> anyways its that , that i like to avoid
<compnn> kostya also had his own fork somewhere
<compnn> "NihAV"
<TheVibeCoder> why is stable fork, mplayer. currently broken?
<compnn> i'm not aware of mplayer being a fork, nor stable
<compnn> it is broken :)
<ePirat> TheVibeCoder, lavfi-preview looks neat
<compnn> lavfi-preview is great
HarshK23 has quit [Quit: Connection closed for inactivity]
<compnn> TheVibeCoder, mplayer is broken due to changes in ffmpeg master , i think. mplayer should still compile if done against one of the stable versions of ffmpeg. although i'm not remembering which version atm. just regular changes mplayer has to do when things are added to ffmpeg.
<compnn> i havent been updating the codecs.conf either :V
Chagall has quit [Ping timeout: 260 seconds]
<BtbN> Well, in theory my schannel DTLS (and whip-extras) implementation is now complete.
<JEEB> 'grats
<BtbN> Now to test if it works
<TheVibeCoder> how you did what corpos cant with vibe coding?
<kasper93> what's the usecase for whip in ffmpeg?
bsFFFFFF has quit [Ping timeout: 276 seconds]
<JEEB> pushing a stream onto a streaming server
bsFFFFFF has joined #ffmpeg-devel
<kasper93> rtmp goes brr
<JEEB> yea there are multiple alternatives, in non-OSS things fragmented mp4 over http is also a common thing
<JEEB> then there's the broadcast stuff like MPEG-TS over UDP
<JEEB> for newer things there's SRT
___nick___ has joined #ffmpeg-devel
Chagall has joined #ffmpeg-devel
<BtbN> Can't you also use WebRTC stuff to directly stream to a browser?
<JEEB> I've not seen that really?
<JEEB> as in not directly from ffmpeg
<BtbN> ah, shame
<BtbN> I tried streaming to twitch with the WHIP thing
<BtbN> but couldn't get it to work
<JEEB> mediamtx can do webrtc tho, so you can feed WHIP and output webrtc to browser I think
<BtbN> problem is, Twitch reports what went wrong in the response body of a failing HTTP request
<BtbN> and our http.c has no way to get that
<JEEB> aye
<BtbN> if there's a failing error code, it gives up on parsing any more
<kasper93> sounds very not user friendly
<BtbN> For users who manually send http requests? :D
cone-441 has joined #ffmpeg-devel
<cone-441> ffmpeg James Almer master:80a05bea4f09: avutil: add an API to handle 3D Reference Displays Information
<cone-441> ffmpeg James Almer master:b2e4b0e2826f: avutil/frame: add a 3D Reference Displays Information side data type
<cone-441> ffmpeg James Almer master:2122f04496e5: avformat/dump: add support for 3D Reference Displays Information side data
<cone-441> ffmpeg James Almer master:39d5a998bd2c: avcodec/packet: add a 3D Reference Displays Information side data type
<cone-441> ffmpeg James Almer master:dbe347f07460: avfilter/vf_showinfo: add support for 3D Reference Displays Information side data
<cone-441> ffmpeg James Almer master:9c2028b80677: avcodec/hevc/hevcdec: export 3D Reference Displays side data
<cone-441> ffmpeg Diego de Souza master:5083f4ad8e26: avcodec/nvenc: add MV-HEVC encoding support
Everything has quit [Quit: Lost terminal]
mark4o has joined #ffmpeg-devel
___nick___ has quit [Ping timeout: 252 seconds]
markh has quit [Ping timeout: 260 seconds]
mark4o is now known as markh
lemourin has quit [Quit: The Lounge - https://thelounge.chat]
lemourin has joined #ffmpeg-devel
mkver has quit [Ping timeout: 248 seconds]
lemourin has quit [Quit: The Lounge - https://thelounge.chat]
mkver has joined #ffmpeg-devel
bsFFFFFF has quit [Quit: bsFFFFFF]
lemourin has joined #ffmpeg-devel
lemourin has quit [Quit: The Lounge - https://thelounge.chat]
<BtbN> Something that baffles me about schannel is how there's seemingly no way to run a TLS server without persisting the unencrypted private key to a well known location on disc that any process can freely access oO
<BtbN> any process of that user or higher permissions
<kasper93> isn't it using some credential store in windows?
<BtbN> It is
<BtbN> But that "credentials store" is just plain text files sitting in %APPDATA%\Microsoft\Crypto\Keys\
<BtbN> There's a bunch of files there already, who knows what application put them there
<BtbN> I can open then all freely
lemourin has joined #ffmpeg-devel
Marth64[m] has joined #ffmpeg-devel
Marth64 has quit [Ping timeout: 276 seconds]
michaelni has quit [Ping timeout: 252 seconds]
<ePirat> BtbN, that sounds very secure
<BtbN> I've just made sure now in the code to delete the private key again asap
<BtbN> once SChannel has loaded it from there, it can go
<BtbN> so it'll sit there for a split second only
<BtbN> Still ridiculous
<BtbN> There is zero technical reason why it couldn't just load it from memory
michaelni has joined #ffmpeg-devel
<BtbN> Hm, do I really just dump 18 patches to the ML now?
<BtbN> https://github.com/FFmpeg/FFmpeg/compare/master...BtbN:FFmpeg:schannel_dtls accumulated quite a bunch of other fixes while at it
<kasper93> btw. if ACL is set correctly for those files, I'm not sure if that's an issue
<kasper93> your private ssh key is also in your home directory accessible by your user
<BtbN> but it's encrypted with a password I have to enter ever machine unlock
<BtbN> Those private keys sitting there are unencrypted and freely readable for my user
<jamrial> BtbN: yes
<BtbN> jamrial: I'm more thinking of sending them individually, the generic obvious fixes first, to get them LGTM'd and applies asap
<BtbN> like, fixing the log messages to not be silly, fixing a bunch of warnings due to wrong format strings, ...
<kasper93> BtbN: it is encrypted if you encrypt it. Anyway, have you tried creating in-memory CERT_CONTEXT, I guess it wasn't working?
<BtbN> The certificate is in-memory just fine
<BtbN> but the certificate refers to its private key only by its name in the store
<BtbN> despite the API having a way to link it directly, SChannel does not look at that
cone-441 has quit [Quit: transmission timeout]
<BtbN> why was it there?
<BtbN> Another weird property of SChannel is that it has an option to set the MTU
<BtbN> But can only set that after the handshake completed
<BtbN> So if the handshake packets are too big, you're screwed
<BtbN> and with big RSA keys that can easily happen