[Ffmpeg-devel-irc] ffmpeg-devel.log.20161208

burek burek021 at gmail.com
Fri Dec 9 03:05:03 EET 2016


[00:21:02 CET] <Zeranoe> These IPs are all over the place....
[00:27:46 CET] <Shiz> if it's all the same UA it might be a targeted attack
[00:27:53 CET] <Shiz> good people don't spoof their UA...
[00:28:02 CET] <cone-479> ffmpeg 03Andreas Cadhalpun 07master:e3694478a98b: yuv4mpegdec: fix leaking pkt in yuv4_read_packet
[00:33:57 CET] <Zeranoe> Shiz: They're all pretty similar
[00:38:33 CET] <Zeranoe> Why would anyone target open source software....
[00:41:39 CET] <JEEB> people are bored, have some random idea of things...
[00:48:45 CET] Action: Compn forgot to spoof his ua when downloading log
[00:48:46 CET] <Compn> hehe
[00:48:57 CET] <Zeranoe> https://twitter.com/O_Shearer/status/806644040351576069
[00:49:20 CET] <Compn> LOL
[00:49:33 CET] <Compn> the official facebook 360 ?
[00:49:37 CET] <Compn> or just another app ?
[00:49:46 CET] <J_Darnley> Why isn't it shipping with ffmpeg?  Oh, is the GPL a scary nasty thing?
[00:50:02 CET] <Zeranoe> I can't confirm that yet
[00:50:03 CET] <Compn> https://facebook360.fb.com/
[00:50:30 CET] <j-b> http://people.videolan.org/~jb/Builds/360/ :)
[00:52:58 CET] <jamrial> j-b: i assume this is using koda's mov demuxer implementation? or are you using a custom demuxer?
[00:53:40 CET] <j-b> well, koda implementation is a bit too new to be in that build... So, custom one
[00:54:19 CET] <jamrial> just asking. he sent the first version a couple weeks ago after all
[00:54:38 CET] <j-b> yep
[00:54:40 CET] <jamrial> and for all i know, these builds could have gone live today :p
[00:55:13 CET] <j-b> yeah no. we did a bit of PR 2 weeks ago with those.
[00:55:44 CET] <Zeranoe> Yup, that's it. I see FBs 360 video trying to download it
[00:55:52 CET] <kierank> Zeranoe: do you want me to pass that on to fb
[00:55:54 CET] <jamrial> j-b: ah, missed that, sorry
[00:56:27 CET] <Shiz> .. with Fx UAs?
[00:57:57 CET] <Zeranoe> kierank: Do you know them or something? Either way, yes, please ask them to include their own copy
[00:58:17 CET] <kierank> yeah I will
[00:58:53 CET] <kierank> sent
[00:59:11 CET] <drv> it looks like something is trying to download /builds/[various different repetitions of builds]/win32/static/ffmpeg-20160712-df1dc52-win32-static.zip
[00:59:15 CET] <drv> somebody debugging their crappy downloader? :)
[00:59:21 CET] <Compn> Shiz : those UA's are usually just "compatability" user agents...
[00:59:30 CET] <RiCON> or some redirect loop
[01:02:58 CET] <Zeranoe> kierank: You could also ask them to give me a job, or set me up with a dedicated server :)
[01:12:01 CET] <jamrial> j-b: well, got a spherical sample that crashed it
[01:12:02 CET] <jamrial> i allowed it to send a report
[01:12:14 CET] <j-b> jamrial: please share the sample :)
[01:12:23 CET] <j-b> jamrial: we've fixed a few crashes already.
[01:13:28 CET] <jamrial> j-b: give me a few, the link i used to download it is no longer valid
[01:15:46 CET] <kierank> Zeranoe: do you have more detailed information
[01:16:10 CET] <Zeranoe> kierank: Regarding...
[01:16:20 CET] <kierank> what exactly is downloading ffmpeg
[01:16:31 CET] <kierank> like a url
[01:16:44 CET] <Zeranoe> It looks like "C:\Program Files\FB360 Spatial Workstation\Video Player\Video Player.exe"
[01:22:25 CET] <jamrial> j-b: https://transfer.sh/KaZ95/motherboard-cube-v2-metadata.mp4
[01:22:41 CET] <j-b> cube?
[01:22:56 CET] <jamrial> yeah, the second of three projections mentioned in the spec
[01:23:11 CET] <j-b> 01:12 < jamrial> j-b: well, got a spherical sample that crashed it
[01:23:16 CET] <j-b> so, not spherical :)
[01:23:35 CET] <jamrial> ah, that. my bad :p
[01:24:27 CET] <jamrial> called it that since the spec is called "Spherical video"
[01:26:36 CET] <j-b> yeah, I have the fix already :)
[01:26:37 CET] <j-b> thx
[01:26:42 CET] <j-b> will push soon
[01:26:45 CET] <jamrial> no prob
[01:27:05 CET] <wm4> so how does vlc render spherical video
[01:27:09 CET] <wm4> shader?
[01:30:01 CET] <kierank> Zeranoe: they are looking into it
[01:31:38 CET] <kierank> Zeranoe: see pm
[01:56:44 CET] <jamrial> j-b: where's the repository for these builds?
[02:43:35 CET] <cone-479> ffmpeg 03James Almer 07master:445204cd5777: avformat/matroskadec: add support for Spherical Video elements
[02:50:29 CET] <cone-479> ffmpeg 03Michael Niedermayer 07master:3ab1311aba74: avcodec/golomb: Consume invalid data in get_ur_golomb_jpegls()
[03:17:51 CET] <Zeranoe> So Facebook 360 will now include a copy of FFmpeg, but I'm not convinced they were the real cultrate. I'm considering requiring a referral to download a build
[03:17:51 CET] <Zeranoe> Unless I can figure out where these are coming from...
[05:26:29 CET] <cone-479> ffmpeg 03James Almer 07master:8396d54d3c62: avformat/matroskadec: fix memleak on stream side data failure
[05:26:30 CET] <cone-479> ffmpeg 03James Almer 07master:120471da2d39: avformat/matroskadec: remove unused variable
[05:29:18 CET] <Compn> Zeranoe : cat traffic.log | grep -v "Gecko/20100101" > traffic.logg
[05:29:20 CET] <Compn> 14 mb
[05:29:20 CET] <Compn> lol
[05:30:12 CET] <Zeranoe> Out of 300GB though that isn't a ton
[05:30:22 CET] <Compn> no i mean
[05:30:27 CET] <Compn> 300mb of that log is that user agent
[05:30:33 CET] <Compn> 14 mb of that 300mb log is other people :D
[05:30:38 CET] <Zeranoe> Oh yeah, -v
[05:30:42 CET] <Compn> -v ;)
[05:30:50 CET] <Zeranoe> Hmmm
[05:31:16 CET] <Zeranoe> That could be a valid UA though
[05:32:05 CET] <Zeranoe> I'll just block *Firefox*, that'll do it
[05:32:27 CET] <Compn> thats not a valid UA
[05:32:29 CET] <Compn> block it
[05:32:53 CET] <Zeranoe> https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/User-Agent/Firefox
[05:33:22 CET] <Compn> i mean its valid but ... not person web browser
[05:34:51 CET] <Zeranoe> I'd have to wildcard the front though, and it could be valid if it's tacked on to the beginning
[05:35:45 CET] <Zeranoe> OVH has a VPS with unlimited bandwidth https://www.ovh.com/us/vps/vps-ssd.xml but only a 100Mbps speed
[05:37:11 CET] <Zeranoe> which I guess is what I'm working with now... hmmm
[05:50:23 CET] <jamrial> Zeranoe: can firefox extensions spawn processes?
[05:50:32 CET] <jamrial> maybe it's some really shitty extension downloading a binary to do conversion out of youtube extracted videos or such
[05:50:44 CET] <Zeranoe> jamrial: Wowwww
[05:50:50 CET] <Zeranoe> jamrial: That's a hell of a good idea
[05:51:14 CET] <jamrial> assuming it's a popular one like Video Download Helper, it would explain this situation
[05:56:04 CET] <Zeranoe> It's not them http://www.downloadhelper.net/install-converter3.php
[05:56:21 CET] <Zeranoe> Though now I have to uninstall all this crap that it installs
[06:05:02 CET] <Zeranoe> You'd be amazed how many youtube download plugins there are...
[06:19:19 CET] <Compn> its interesting
[06:19:24 CET] <Compn> indonesia has 3000 hits 
[06:19:31 CET] <Compn> iraq! 
[06:20:03 CET] <Compn> nigeria
[06:20:23 CET] <Compn> all over the world, thats for sure
[06:21:29 CET] <Zeranoe> Right? It's weird, and the user agents are not the same
[06:21:54 CET] <Compn> i havent checked useragents yet. just checking the ip addy right now
[07:02:02 CET] <Zeranoe> We're back up. I put a referer check in place for the time being.
[11:24:12 CET] <cone-178> ffmpeg 03Srinath K R 07master:4df5a6a26a8a: avfilter/vf_hwupload_cuda: Add min/max limits for the 'device' option
[11:27:17 CET] <cone-178> ffmpeg 03Srinath K R 07release/3.2:314c425b16ae: avfilter/vf_hwupload_cuda: Add min/max limits for the 'device' option
[11:28:13 CET] <cone-178> ffmpeg 03Srinath K R 07release/3.1:8a4b18c639b5: avfilter/vf_hwupload_cuda: Add min/max limits for the 'device' option
[12:33:53 CET] <superware> can someone please (please, please) take a look at https://trac.ffmpeg.org/ticket/5615 ? it's easy to reproduce so maybe it's easy to fix.
[12:39:54 CET] <J_Darnley> Do you want my suggestion if you think it is easy?
[12:41:32 CET] <superware> I don't think it's easy, I just said it might be easy.
[12:41:40 CET] <superware> sure
[12:56:28 CET] <superware> J_Darnley?
[13:21:39 CET] <J_Darnley> My suggestion would have been "fix it yourself" if you thought it was easy but you don't think it would be easy so I won't make that suggestion.
[13:22:16 CET] <wm4> it probably is easy
[13:22:26 CET] <wm4> on the other hand, libavformat network code...
[13:56:46 CET] <Gramner> J_Darnley/BBB: yasm doesn't think that negative numbers are in fact numbers. that annoyed me so i fixed it, but then the main yasm dev disappeared so the fix isn't in any released version https://github.com/yasm/yasm/commit/32ad9c46cebd5bf07af9c05383d3c16e6cb1f17d
[13:56:54 CET] <Gramner> nasm doesn't have that bug afaik
[13:57:22 CET] <BBB> the workaround is fairly simple
[13:57:26 CET] <BBB> 0-number instead of -number
[13:57:30 CET] <BBB> its kind of silly but whatever
[13:57:42 CET] <BBB> we have worse hacks than that in our codebase...
[13:57:50 CET] <J_Darnley> We would need to use that as we allow people to use yasm 0.6
[13:58:00 CET] <J_Darnley> (the work-around I mean)
[14:38:59 CET] <Compn> yeah.
[14:46:03 CET] <superware> wm4: you said "it probably is easy", what did you mean?
[14:46:30 CET] <wm4> just what I wrote
[14:52:06 CET] <superware> it seems that when one of the UDP related functions is blocking, the abort mechanism doesn't work (when callback returns 1).
[14:55:09 CET] <wm4> ok then it's probably not easy
[16:10:49 CET] <nevcairiel> jkqxz: https://trac.ffmpeg.org/ticket/6003 .. needs a newer vaapi version check?
[16:11:15 CET] <wm4> I'm compiling ffmpeg on a linux machine, and it's built for the
[16:11:16 CET] <wm4> past several years. I just updated git-master after being gone
[16:11:16 CET] <wm4> for several months, and I get this.
[16:11:19 CET] <wm4> jkqxz ruined it all!
[16:15:45 CET] <BBB> wow nicolas is really raging at me, what did I do
[16:15:59 CET] <jkqxz> Looks like that is an ABI break in libva.
[16:16:48 CET] <jkqxz> I guess we have fix that with a conditional on the newer ABI, rather than just testing for the structure?
[16:18:59 CET] <jkqxz> Hmm.  I think it needs a separate configure check for VP9 + newer structure.
[16:19:15 CET] <wm4> BBB: disagree with him
[16:19:46 CET] <nevcairiel> dont worry BBB, he isnt the most logical person
[16:19:54 CET] <wm4> jkqxz: would an ifdef in the code not be sufficient? a bit inelegant, but more appealing than dealing with configure
[16:19:57 CET] <jamrial> what wm4 said. you disagreed with him. you're therefore wrong
[16:21:13 CET] <BBB> but Im trying to be nice
[16:21:27 CET] <BBB> Im genuinely trying to be nice and move on and stuff, I dont get the rage, what did I do
[16:22:32 CET] <ubitux> you lost the vote so now you have to kneel and admit defeat in a silent stance
[16:24:05 CET] <BBB> ...
[16:24:22 CET] <nevcairiel> the more i read from nicholas mails, it seems like he spent more time in rhetoric class then "decent person" class
[16:24:30 CET] <jamrial> BBB: literally, just ignore him
[16:24:41 CET] <ubitux> btw i had no answers to my questions so i couldn't vote (and the vote looks meaningless as ffserver technical limitations will likely not be met til the drop date)
[16:25:06 CET] <j-b> BBB: you exist.
[16:25:56 CET] <BBB> if michael can remove the ffm dependency by moving it to the tools (which isnt hard, really) and then remove ffserver-related network dealings or make them an acceptable public api and remove some hacks in ffmpeg.c dealing with the server, or make them maintainable...
[16:26:00 CET] <BBB> then its done, right?
[16:26:13 CET] <BBB> thats not hard, and michael has this nagging ability to do stuff that nobody likes doing
[16:26:20 CET] <BBB> so I dont see it as impossible
[16:26:33 CET] <BBB> and if he does it, I dont see the problem
[16:27:55 CET] <ubitux> me neither
[16:28:03 CET] <ubitux> did we identify all the issues?
[16:28:13 CET] <jkqxz> wm4:  The ifdef would do something funny if you build with older headers and run with a newer driver (since the profile would be there and appear to be working, but all the decoding would fail).
[16:28:28 CET] <jamrial> if someone's able to fix ffserver then horay, congrats
[16:28:31 CET] <jamrial> would have been nicer if it happened before we embarrased ourselves on the website and twitter, though
[16:28:58 CET] <jkqxz> Reading the i965 driver (the only implementation of VP9) it only reads that field if the profile is set, so it's probably ok.
[16:29:03 CET] <BBB> ubitux: I think that list I just gave is pretty complete, wbs had a more detailed list on IRC a few days ago, which was copied to the ML thread
[16:29:19 CET] <wm4> jkqxz: couldn't it just refuse init? or maybe exclude the hwaccel from some list (I don't know which)
[16:29:25 CET] <BBB> but anyway, I should go kneel in silence & :X
[16:29:25 CET] <wm4> but yeah maybe a configure check is cleaner
[16:29:33 CET] <ubitux> BBB: was the rtsp (and the like) api abuse mentioned? isn't this considered a technical issue to be solved as well? 
[16:29:46 CET] <wm4> BBB: actually your talk about Trump encouraged me trying to fight it out with Nicolas
[16:29:47 CET] <jkqxz> The ifdef could have an else case with an error return, but we would prefer to fail earlier than that.
[16:30:04 CET] <BBB> ubitux: thats network dealings, yes, its how it deals with rtsp/rtp/udp/tcp etc. by using internal API
[16:30:16 CET] <ubitux> ok
[16:30:23 CET] <ubitux> when is the due date already? next release?
[16:30:27 CET] <BBB> wm4: Im not sure that was my intention, the idea was just to go and vote
[16:36:59 CET] <BBB> ubitux: I believe it was next release, yes
[16:37:41 CET] <BBB> ubitux: but since it was made conditional on technical issues being fixed but the list of technical issues wasnt exactly defined, Im fearful well go through another madhouse of crazy by the time the release comes around
[16:37:44 CET] <BBB> :(
[16:38:41 CET] <ubitux> yes, as i said, the technical issues weren't clarified even though i asked for them
[16:38:59 CET] <ubitux> the vote looks invalid to me, so i ignored it
[16:39:14 CET] <jamrial> nicolas didn't answer when that was asked, so i think it's almost a given people like andreas will argue it should be at a major bump once the time comes, because exporting internal API on shared libraries and ffm* will remain as such until then
[17:57:27 CET] <BBB> wbs: http://ffmpeg.org/pipermail/ffmpeg-devel/2016-December/204049.html in case youre not reading the ML
[17:57:45 CET] <wbs> BBB: I noticed; it should be fixed by cherrypicking 6a62795d
[17:57:59 CET] <wbs> sorry for not noticing to include that in the cherrypick patchset
[17:58:07 CET] <BBB> no problem
[17:59:06 CET] <BBB> hm, your webgit is strange
[17:59:20 CET] <BBB> like, https://git.libav.org/?p=libav.git&a=search&h=HEAD&st=commit&s=6a62795d&sr=1 returns nothing
[18:00:17 CET] <wbs> https://git.libav.org/?p=libav.git;a=commit;h=6a62795d
[18:00:35 CET] <wbs> I don't think the search field actually looks at commit hashes, only commit content
[18:02:20 CET] <BBB> ok
[18:02:25 CET] <BBB> hows 10bit asm going?
[18:02:55 CET] <wbs> not started yet, but will hopefully start soon
[18:06:36 CET] <BBB> cool, Im looking forward to it
[18:06:47 CET] <BBB> I hear some people slowly getting excited about 10bit VP9
[18:06:53 CET] <BBB> so it may start getting some use next year
[18:07:15 CET] <wbs> nice. libvpx has been getting some opts to 10 bit for arm lately as well I think
[18:09:33 CET] <nevcairiel> youtube uses it for hdr streams now
[18:11:26 CET] <BBB> did they improve their 10bit idct performance?
[18:11:41 CET] <cone-633> ffmpeg 03Martin Storsjö 07master:da5c8284c02c: aarch64: h264idct: Use the offset parameter to movrel
[18:18:29 CET] <BBB> I think it still does this strange thing where it runs the C IDCT if the coefficients are larger than some kind of threshold value, at least in the 1.6.0 release I have installed here
[18:18:38 CET] <jamrial> BBB: HDR youtube content is vp9 10bit, so i can see where the interest comes from
[18:18:44 CET] <BBB> and the threshold value is quite low for larger transforms
[18:19:05 CET] <jamrial> oh lol, missed nevcairiel saying the same thing
[18:20:06 CET] <BBB> oh I thought that code looked familiar
[18:20:10 CET] <BBB> ; Part of the functions are originally derived from the ffmpeg project.
[18:20:15 CET] <BBB> isnt there a copyright concern?
[18:20:34 CET] <BBB> (like, our license is different from theirs, BSD vs. LGPL)
[18:20:50 CET] <durandal_1707> sue sue sue
[18:21:07 CET] <BBB> so american :-p
[18:25:06 CET] <jamrial> BBB: think you could add smpte2084 trc support to vf_colorspace?
[18:25:29 CET] <BBB> wasnt koda doing that?
[18:25:55 CET] <jamrial> no idea
[18:26:18 CET] <jamrial> if he is then cool
[18:40:04 CET] <cone-633> ffmpeg 03Michael Niedermayer 07master:1296f844955e: avformat: Add max_streams option
[18:42:41 CET] <BBB> jamrial: sounds like he has soem patches in that direction, so Ill defer that to him if thats OK
[18:43:04 CET] <jamrial> BBB: cool
[19:10:25 CET] <atomnuker> kierank: so you need someone to talk about av1 at fosdem?
[19:14:46 CET] <wm4> we should have a talk about ffmpeg drama at fosdem
[19:16:19 CET] <durandal_1707> that would be more like romantic comedy
[19:39:32 CET] <kierank> atomnuker: yeah would be interesting I think
[19:39:43 CET] <kierank> Please can you submit one via pentabarf
[19:46:03 CET] <atomnuker> kierank: done, will write a more detailed description in a couple of hours, gotta go
[19:46:32 CET] <kierank> Thanks!
[20:15:59 CET] <llogan> why does ffprobe seem to only show refs=1? ffprobe -v error -show_entries stream=refs -select_streams v input.mp4
[20:16:27 CET] <llogan> although input.mp4 may have ref=3
[20:18:29 CET] <BBB> kierank: youre doing av1?
[20:18:37 CET] <BBB> <- confused
[21:01:43 CET] <BBB> michaelni: can you provide a graph with memory usage depending on the number of streams?
[21:01:52 CET] <BBB> michaelni: I think based on that we can define a sensible default
[21:02:25 CET] <BBB> if we use 100GB with 10 streams, we should probably limit it to <10, but if memory usage doesnt change much between 10k and 100k streams, maybe a limit of 1M streams is acceptable
[21:02:35 CET] <BBB> michaelni: assuming the goal is to prevent OOM scenarios
[21:11:41 CET] <michaelni> i think one actual OOM case was with around 30k-60k  streams, the exact usage likely depends on the demuxer
[21:13:51 CET] <wm4> am I the only one who thinks it's stupid to try to prevent OOM for such files
[21:17:37 CET] <BBB> wm4: I personally dont care about it, but if others do care, I dont mind as long as it doesnt interfere with regular usage
[21:18:01 CET] <BBB> michaelni: so how about we set a limit of 10k or so? Maybe even 5k is ok, but 100 seems kinda low
[21:21:03 CET] <wm4> seems like a rather useful feature
[21:21:18 CET] <wm4> is it true that the files were generated by fuzzing? i.e. not even anything in the wild
[21:22:23 CET] <BtbN> https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/utils.c#L1729 does this check look wrong to anyone else? At least the error message makes no sense, and some people are having issues with packet sizes exceeding INT_MAX, which should be totally fine.
[21:24:32 CET] <nevcairiel> 2gb of packet seems  not that fine
[21:24:50 CET] <BtbN> can easily happen with lossless or uncompressed 4K or something
[21:25:01 CET] <nevcairiel> for one packet? not really
[21:25:22 CET] <BBB> wm4: of course :-p
[21:25:28 CET] <Fog^> I'm dealing with this currently
[21:25:49 CET] <philipl> BtbN: https://devtalk.nvidia.com/default/topic/980859/linux/linux-375-20-cuvid-hevc-10-bits-failed-to-decode-first-gop/
[21:25:56 CET] <philipl> I'm pretty sure it's nothing to do with us, but FYI
[21:26:15 CET] <Fog^> so the situation is we recently added a feature in Dolphin Emulator to dump at the internal resolution
[21:26:23 CET] <Fog^> without caring about window size
[21:26:53 CET] <Fog^> so we got a report on the issue tracker which points out that anything higher than 5X native fails with FFV1
[21:27:06 CET] <Fog^> (because people do the weirdest things)
[21:27:10 CET] <nevcairiel> what resolutiuon would that then be?
[21:27:26 CET] <nevcairiel> even 4k uncompressed at 16-bit per component is only 66mb or s o
[21:27:29 CET] <nevcairiel> per image
[21:27:54 CET] <BtbN> philipl, yeah, if it happens with 10 bit with 8 bit output as well, I'd be worried. But for now, I won't bother fixing bugs in unrelased features
[21:28:02 CET] <nevcairiel> 2gb for one image is unimagineably large
[21:28:05 CET] <Fog^> 3200x2640
[21:28:08 CET] <Fog^> is 5x
[21:28:09 CET] <Fog^> which works fine
[21:28:19 CET] <Fog^> 6x is 3840x3168
[21:28:40 CET] <Fog^> but I think that number is incorrect in our UI
[21:29:08 CET] <BtbN> Also what I mean completely unrelated to that, something seems off with that check in general. It talks about "Invalid minimum required packet size"
[21:29:12 CET] <BtbN> But it never checks that.
[21:29:14 CET] <wm4> the limit that hits sooner is about 16000x16000
[21:29:19 CET] <Fog^> because the output of the fifo file that I'm using is producing a 5632x3168 file
[21:29:20 CET] <wm4> I know this caused problems to some people
[21:29:34 CET] <wm4> and the limit is enforced because it "could" overflow some int calculations
[21:30:02 CET] <Fog^> which produces a value of 2497921024
[21:30:04 CET] <BtbN> So the error message is just wrong?
[21:30:52 CET] <cone-633> ffmpeg 03Paul B Mahol 07master:a25114983350: avfilter/vf_deband: fix silly mistakes
[21:31:20 CET] <Fog^> then again, the listed size in the UI is the framebuffer, not the actual output
[21:32:46 CET] <BtbN> nevcairiel, what I mean with "looks wrong" is that it says the min size is invalid. But it does not check the min size
[21:34:45 CET] <llogan> oh...there was a bug report about it already.
[21:35:17 CET] <llogan> (ffprobe refs...)
[21:48:55 CET] <BtbN> it is closed :o
[21:48:58 CET] <atomnuker> 3.5 years...
[21:49:27 CET] <llogan> /me jumps out window
[21:49:48 CET] <llogan> obviously i haven't been on irc much lately...
[22:02:09 CET] <BBB> atomnuker: was the issue actually fixed? :-p
[22:02:27 CET] <BBB> (kamedo2 never reported the post-i/s fix numbers)
[22:04:23 CET] <atomnuker> yeah, long ago, that thread was just used as a place to get feedback and work on the encoder
[22:09:57 CET] <cone-633> ffmpeg 03Lou Logan 07master:ee795d9d2674: doc/muxers: remove "-strict experimental" from tee muxer examples
[22:14:01 CET] <BBB> sweet
[22:50:00 CET] <nevcairiel> what kind of weird compuation does ffv1 use there
[22:50:08 CET] <nevcairiel> the raw uncompressed image is smaller then that
[22:51:16 CET] <Fog^> nevcairiel: it's passing maxsize directly to ff_alloc_packet2
[22:51:21 CET] <nevcairiel> why multiplied by 37
[22:51:32 CET] <nevcairiel> why would one frame get 37 times bigger then its raw input
[22:51:32 CET] <BtbN> probably some weird worst-case of the compression?
[22:51:38 CET] <Fog^> damned if I know
[22:54:15 CET] <Fog^> it actually used to be 35
[22:54:36 CET] <Fog^> according to the version of ffmpeg I used to debug this
[22:58:13 CET] <nevcairiel> guessing the final output size of a codec is not an easy problem, but assuming a worst case of a 100-something raw image turning into 2gb of encoded image seems rather out there =p
[22:58:24 CET] <nevcairiel> 100-something mb that is
[23:12:38 CET] <durandal_1707> option to not compress was added later just for this case, right michaelni ?
[23:13:23 CET] <durandal_1707> version 4
[23:14:41 CET] <michaelni> yes
[00:00:00 CET] --- Fri Dec  9 2016


More information about the Ffmpeg-devel-irc mailing list