[Ffmpeg-devel-irc] ffmpeg-devel.log.20150523

burek burek021 at gmail.com
Sun May 24 02:05:02 CEST 2015


[00:40:08 CEST] <cone-682> ffmpeg 03Michael Niedermayer 07master:ebb0ca3d7046: avformat/nutdec: Check X in 2nd branch of index reading
[03:44:29 CEST] <cone-195> ffmpeg 03Urvang Joshi 07master:02cf59f3a619: WebP encoder: use WebPAnimEncoder API when available.
[04:21:48 CEST] <cone-195> ffmpeg 03Michael Niedermayer 07release/2.5:3ea101b23448: avformat/nutdec: Check X in 2nd branch of index reading
[04:21:49 CEST] <cone-195> ffmpeg 03Andreas Cadhalpun 07release/2.5:ffb4362bf0b6: nutdec: fix infinite resync loops
[04:21:50 CEST] <cone-195> ffmpeg 03Andreas Cadhalpun 07release/2.5:af27254eb30c: nutdec: stop skipping bytes at EOF
[04:21:51 CEST] <cone-195> ffmpeg 03Andreas Cadhalpun 07release/2.5:21d0ae829f72: nutdec: abort if EOF is reached in decode_info_header/read_sm_data
[04:59:03 CEST] <cone-195> ffmpeg 03Urvang Joshi 07n2.5.7:HEAD: WebP encoder: use WebPAnimEncoder API when available.
[05:17:25 CEST] <rcombs> http://puu.sh/hX94U/b2429e5ba9.dts <-- lavf misidentifies this wav file's stream as PCM instead of DTS, and I can't tell how it's supposed to know
[06:09:04 CEST] <rcombs> wm4: around?
[06:13:33 CEST] <Compn> rcombs : that sounds like an old bug
[06:13:37 CEST] <Compn> latest ffmpeg ?
[06:13:47 CEST] <Compn> filename extension detection ?
[06:13:53 CEST] <Compn> if .dts ==== dts
[06:13:54 CEST] <Compn> :P
[06:14:00 CEST] <rcombs> Compn: as of a few days ago, anyway
[06:14:03 CEST] Action: Compn wonders why so much raw .dts files
[06:14:25 CEST] <rcombs> well it appears to actually be a wave file containing DTS audio
[06:14:31 CEST] <rcombs> because apparently that's a thing
[06:14:32 CEST] <Compn> yeah
[06:15:02 CEST] <rcombs> VLC handles it, but I don't see where it does anything different
[06:15:23 CEST] <rcombs> and I'd _really_ rather not build VLC locally so I can debug and figure out what it does right
[06:15:49 CEST] <rcombs> (also, VLC has its own wav demuxer because why not?)
[06:16:38 CEST] <rcombs> (I mean, I guess for files like this, but you'd think fixing whatever the bug in lavf is would be easier)
[06:16:56 CEST] <Compn> there was a patch in 2010 to fix this
[06:17:00 CEST] <Compn> i see in the googles
[06:17:12 CEST] <Compn> https://trac.ffmpeg.org/ticket/1747
[06:17:14 CEST] <Compn> theres the ticket
[06:18:05 CEST] <Compn> another similar ticket
[06:18:05 CEST] <Compn> https://trac.ffmpeg.org/ticket/4492
[06:18:26 CEST] <Compn> probably some frame size probe change 
[06:18:46 CEST] <Compn> btw to find that i did this i ngoogle : intitle:dts site:trac.ffmpeg.org
[06:18:51 CEST] <Compn> but probably intitle:dca is better
[06:18:52 CEST] <Compn> derp
[06:18:55 CEST] Action: Compn sleps
[06:19:09 CEST] <rcombs> that second one looks unrelated
[06:19:29 CEST] <jamrial> michaelni: i'm not sure if adding a new libwebp encoder for the WebPAnimEncoder api was the best idea
[06:21:20 CEST] <Compn> rcombs : also is your vlc compiled against libav ? 
[06:21:26 CEST] <Compn> wonder if its regression...
[06:22:10 CEST] <jamrial> michaelni: the changes to the webp muxer to automatically detect animated webp streams are great, but I'd have preferred a single libwebp encoder using either the WebPEncode or WebPAnimEncoder APIs depending on availability
[06:22:36 CEST] <jamrial> for cleanliness sake, if anything
[06:22:54 CEST] <rcombs> Compn: I have whatever the official builds use
[06:24:10 CEST] <jamrial> michaelni: i assume the decision to have two was to let users choose between one or the other? is there any benefit of using the native muxer if WebPAnimEncoder is available?
[06:25:47 CEST] <rcombs> wow, silly file
[06:26:00 CEST] <rcombs> the patch from 2006 appears to be lost to the ages
[06:26:18 CEST] <rcombs> https://ffmpeg.org/pipermail/ffmpeg-devel/2006-February/010311.html
[06:33:09 CEST] <cone-195> ffmpeg 03James Almer 07master:98be2d94eaa1: libwebp: remove unneeded defines
[06:46:08 CEST] <cone-195> ffmpeg 03James Almer 07master:44ca8a6adbc3: libwebpenc_common: add header guards
[06:53:34 CEST] <cone-195> ffmpeg 03James Almer 07master:1096c46c55d6: libwebp: use a separate AVClass for each encoder
[10:23:56 CEST] <nevcairiel> rcombs: dts-in-wav is a horrible format, and auto-detection relies on a few things going right .. sometimes wav files have silence before the actual dts data starts, and you are screwed
[10:27:27 CEST] <nevcairiel> rcombs: in wavdec there is a function called handle_stream_probing, which sets a number of packets which should run through avformats probing logic, which would then use the dts/dca demuxer to probe the packets if they look like dca data
[10:28:23 CEST] <nevcairiel> maybe something in that logic broke? 
[10:28:28 CEST] <nevcairiel> i have my own dts-in-wav detection code =p
[10:33:35 CEST] <nevcairiel> debug logging suggests its only probing 2 formats o.o
[10:35:25 CEST] <nevcairiel> oh, i read it wrong
[10:39:07 CEST] <cone-459> ffmpeg 03Andreas Cadhalpun 07master:0b79a389ba48: nutdec: always check the get_str return value
[11:03:21 CEST] <nevcairiel> rcombs: this change breaks detection of that file: http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=6663205338f57eedb2392263dde48e2717c6e980
[11:03:51 CEST] <nevcairiel> i have no idea wtf its supposed to do
[11:03:55 CEST] <nevcairiel> ask michaelni perhabs : )
[11:24:43 CEST] <wm4> don't we love stuff like this
[11:35:27 CEST] <JEEBsv> wow
[12:23:18 CEST] <cone-459> ffmpeg 03Michael Niedermayer 07master:9f5769437aaa: avformat/wavdec: Increase probe_packets limit
[12:23:53 CEST] <michaelni> rcombs, fixed the dts detection
[13:49:47 CEST] <rcombs> michaelni: nevcairiel: thanks :)
[14:05:57 CEST] <rcombs> michaelni: https://gist.github.com/6e85572b6b35841ed3f4 <-- tested with the rest of my samples; this is the minimum value required to work for all of them
[14:07:41 CEST] <rcombs> also, I get these errors (2 at start and 2 at end) on all of them: https://gist.github.com/59e12a58f486a621fd5c
[14:19:07 CEST] <nevcairiel> i still find this patch i linked rather weird
[14:19:12 CEST] <nevcairiel> it detects the dca frames just perfectly fine
[14:19:17 CEST] <nevcairiel> but the odd check there breaks it
[14:28:52 CEST] Action: rcombs stares into the patch
[14:28:55 CEST] <rcombs> (the patch stares back)
[14:36:44 CEST] <sh4rm4^bnc> ubitux, which function of avformat would i pass the format string "v4l2" ?
[14:37:23 CEST] <sh4rm4^bnc> av_find_input_format("video4linux2") returns NULL
[14:37:33 CEST] <sh4rm4^bnc> (same with "v4l2")
[14:53:41 CEST] <wm4> ffmpeg -devices
[14:54:18 CEST] <sh4rm4^bnc>   E v4l2            Video4Linux2 output device
[14:54:18 CEST] <sh4rm4^bnc>  D  video4linux2,v4l2 Video4Linux2 device grab
[14:57:29 CEST] <sh4rm4^bnc> ah, got to call avdevice_register_all() first
[15:00:29 CEST] <wm4> right
[15:00:37 CEST] <wm4> this function has no reason to exist, but you need to call it anyway
[15:00:50 CEST] <wm4> it's no secret that libav*'s API is crap
[16:04:48 CEST] <cone-571> ffmpeg 03Rodger Combs 07master:40a3e1e9c549: avformat/wavdec: Increase dts packet threshold to fix more misdetections
[16:14:40 CEST] <rcombs> wm4: you want to split Secure Transport as well, or should I wait for the existing split patches to land and then do that?
[16:17:12 CEST] <wm4> I'd prefer the latter
[16:20:41 CEST] <rcombs> sure
[16:38:43 CEST] <cone-571> ffmpeg 03Michael Niedermayer 07master:4ae15605f629: avcodec/dcadec: Search and decode frame in case it starts later in a packet
[16:50:07 CEST] <ubitux> it seems ubsan detected slightly more stuff
[16:51:26 CEST] <ubitux> maybe just one
[17:44:19 CEST] <cone-571> ffmpeg 03Andreas Cadhalpun 07master:ed4932a6941f: nutdec: fix various memleaks on failure
[19:19:32 CEST] <cone-571> ffmpeg 03James Almer 07master:57d9788d1b06: configure: simplify libwebp checks
[19:47:09 CEST] <cone-571> ffmpeg 03James Almer 07master:8952254ffe7f: libwebp: simplify AVCodec.close functions
[19:58:44 CEST] <pkhgj> hi all
[19:58:54 CEST] <pkhgj> maybe someone can help me
[19:59:10 CEST] <pkhgj> ffmpeg can receive a webrtc live stream?
[19:59:19 CEST] <pkhgj> and transcode it to webm live?
[19:59:33 CEST] <pkhgj> I saw: https://groups.google.com/a/webmproject.org/forum/#!topic/webm-discuss/Y2At8cjIfW0 
[19:59:47 CEST] <pkhgj> the last post says it will be available soon
[19:59:51 CEST] <pkhgj> is this info correct?
[20:01:39 CEST] <JEEBsv> that's for muxing
[20:01:42 CEST] <JEEBsv> not demuxing
[20:02:00 CEST] <JEEBsv> and that's just muxing of "webm dash"
[20:02:04 CEST] <JEEBsv> nothing to do with webrtc
[20:03:11 CEST] <JEEBsv> http://git.videolan.org/?p=ffmpeg.git;a=commit;h=6fd300ac6c2c3871736ce0e6df95603255004dc6
[20:08:16 CEST] <pkhgj> so basically it can only mux wenm files?
[20:08:27 CEST] <pkhgj> *webm 
[20:10:54 CEST] <pkhgj> so I cannot send a live webrtc stream to ffmpeg / ffserver to transcode it to webm ?
[20:11:29 CEST] <wm4> this patch was obviously written by a google employee
[20:11:44 CEST] <wm4> they probably have no interest in implementing a demuxer in ffmpeg
[20:17:09 CEST] <jamrial> same with animated webp. they bothered adding the webpanim api to encode them (even though there already was code to do the same, albeit supposedly being less efficient at it), but nothing to decode them
[20:17:29 CEST] <jamrial> so ffmpeg can create animated webp but not read them
[20:18:01 CEST] <thardin> abimated.. webp?
[20:18:11 CEST] <thardin> why not webm?
[20:18:19 CEST] <wm4> that's why I'm saying ffmpeg should have a policy of not adding encoders for things it can't decode
[20:18:49 CEST] <jamrial> thardin: same as animated png. gif replacement without actually being video
[20:19:10 CEST] <thardin> sounds stupid
[20:19:14 CEST] <wm4> I really wonder what "without actually being video" means
[20:19:35 CEST] <thardin> cargo culting idiocy
[20:48:48 CEST] <pkhgj> i cannot understand how could I exchange the sdp object between a client and a server runing ffmpeg to start streaming live to the ffmpeg server ? is it possible? any lead?
[20:49:38 CEST] <pkhgj> and produce DASH to play in a DASH player
[21:55:08 CEST] <cone-571> ffmpeg 03Michael Niedermayer 07master:de0d3fe562e2: avcodec/y41pdec: Avoid using float for size test
[22:36:36 CEST] <cone-571> ffmpeg 03Michael Niedermayer 07master:3bf57acb9cb2: avformat/4xm: Change fps to AVRational
[23:54:37 CEST] <cone-571> ffmpeg 03Michael Niedermayer 07master:17264cd281c4: avformat/avienc: Avoid float usage
[00:00:00 CEST] --- Sun May 24 2015


More information about the Ffmpeg-devel-irc mailing list