[Ffmpeg-devel-irc] ffmpeg-devel.log.20150117

[burek]

[01:54] <heffer2k02> Does anyone understand AAC + MP4 internals a bit? FFmpeg is reported that my audio file has a "start offset" of 0.02333 (1024 samples / 44100), i.e. a single audio frame. Looking at ffprobe, it doesn't reveal and delay (first pts is 0). Digging further, the MP4 files first sample actually points to an AAC FIL frame with some string indentifying the encoder software, however it has a duration of 1024... Is this a bug in ffmpeg?
[01:54] <heffer2k02> Or is FIL frame interpreted as silence? Confused...
[02:02] <thardin> ask on the mailing list maybe?
[02:10] <aetasx> where does it report the start offset?
[02:11] <cone-576> ffmpeg.git 03Michael Niedermayer 07master:040e5703e6b8: avcodec/flac_parser: print error messages in case of errors
[02:11] <cone-576> ffmpeg.git 03Michael Niedermayer 07master:c4d85fc23c10: avcodec/flac_parser: fix handling EOF if no headers are found
[02:18] <heffer2k02> aetasx: ffmpeg -i gives a line like: Duration: 00:00:16.19, start: 0.023220, bitrate: 78 kb/s
[02:20] <heffer2k02> aetasx: I've just been doing some ffprobes and managed to get this line: "start_time": "-0.023220". Which is negative, so now I'm more confused. The negative may make sense though as it offset the apparent "delay" inserted by the MP4 giving the FIL frame a duration of 1024. Although I might be reaching.
[02:21] <heffer2k02> Not sure how ffmpeg is computing these values.
[02:27] <aetasx> you sure thats not just the information you see when pulling up an mp4 file in mediainfo?
[02:28] <heffer2k02> mediainfo doesn't actually say anything about start offsets...
[02:29] <aetasx> in ffprobe, whats the actual file start_time listed as?
[02:29] <aetasx> vs the stream start_time
[02:30] <heffer2k02> FORMAT section and STREAM section both report: start_time=-0.023220
[02:31] <aetasx> and any other streams in there?
[02:31] <heffer2k02> Nope
[02:32] <heffer2k02> What does a negative start time usually acheive? Is it just applied as an offset to all pts?
[02:32] <aetasx> usually its when having multiple streams, not sure what players do with it, probably drop it
[02:33] <aetasx> try doing "ffmpeg -ss 0 -i <file> -c:a copy <outfile>"
[02:43] <kierank> heffer2k02: aac start time is not a trivial problem
[02:43] <kierank> https://developer.apple.com/library/mac/documentation/QuickTime/QTFF/QTFFAppenG/QTFFAppenG.html
[02:43] <heffer2k02> aetasx: Ah, so that produced a file that now reports 0 as start time. The first frame pts now reports as 0.02322. I should point out that this was originally extracted from a video, and it will be remuxed with it at a later time. Has this effectively changed the presentation time of the audio?
[02:43] <kierank> and https://developer.apple.com/library/mac/technotes/tn2258/_index.html
[02:46] <heffer2k02> kierank: thanks. these look useful.
[02:46] <aetasx> thanks kierank
[03:21] <heffer2k02> kierank: Thanks again. Was able to track this offset down to the elst box.
[03:21] <kierank> the offset is correct iirc
[13:30] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:7532f0e14c5e: avcodec/libmp3lame: use av_malloc_array()
[13:30] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:1f4ef3ccaee6: avcodec/libopusenc: Use av_mallocz_array()
[13:30] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:f8ff76199c38: avcodec/mss12: Use av_malloc_array()
[14:38] <arwa> does eq/eq2 filter need to be ported in one single filter?
[14:40] <compn> i dont think so
[14:40] <compn> but they are very similar
[14:40] <compn> iirc
[14:50] <wm4> having two of them makes no sense at all
[14:51] <wm4> and it was probably only done because of shitty mplayer dev culture
[14:51] <arwa> okay thanks :)
[15:20] <ubitux> compn: yes they need to, obviously.
[15:20] <ubitux> arwa: they're probably going to be merged into hue filter
[15:20] <ubitux> one is a superset of the other
[15:21] <arwa> Okay, so should I directly go for the merged filter?
[15:21] <ubitux> i thought there was mail exchange about this
[15:23] <arwa> Yeah, there was. But there was nothing mentioned about the merging part.
[15:26] <ubitux> we just need the feature from eq & eq2 (which are the same basically) into hue
[15:26] <ubitux> i think that was the plan, and you can look at J. Darnley code
[15:26] <arwa> Yes, I was going through it when I realized that it is merged
[15:28] <arwa> Does the arm code need to be ported?
[15:29] <wm4> arm code? or x86 asm code?
[15:29] <arwa> sorry the asm code
[15:30] <wm4> I think porting the C first, and adding asm as yasm later would be a reasonable proposition
[15:30] <wm4> although it could well be that this old mmx code is slower than the C code
[15:30] <arwa> okay.
[15:33] <ubitux> hue works in a LUT approach iirc
[15:44] <tim> Hi!
[15:51] <tim> nobody response?
[15:52] <tim> *àJ†
[15:53] <compn> hey tim
[15:53] <compn> what country are you from ?
[15:53] <tim> china
[15:53] <tim> and you ?
[15:53] <compn> awesome. i'm from usa
[16:33] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:353cf95f948e: avfilter/vf_frei0r: also set AVFilterLink.frame_rate
[16:33] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:e0623953e1e6: avfilter/vf_frei0r: fix build with DEBUG enabled
[16:33] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:edf1cb7aeea0: avformat/nsvdec: Check av_get_packet() return value
[16:33] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:8de0990e9f22: avcodec/dvbsubdec: Check get_region() return value
[18:31] <arwa> I added the attachments to a wrong page. How can I delete those attachments?
[18:34] <cone-46> ffmpeg.git 03Timo Rothenpieler 07master:bc3f76779b6c: avcodec/nvenc: Add support for nvenc api version 5
[18:38] <michaelni> arwa, deleted them, only admins can delete stuff
[18:39] <arwa> Okay, thanks and sorry.
[20:28] <kierank> so what exactly is wrong with the nvidia SAR/DAR?
[20:29] <kierank> anyhow the important thing is that the aspect ratio applies to the 702 pixels
[20:30] <nevcairiel> but that stops being relevant for other resolutions
[20:30] <nevcairiel> and apparently its applied everywhere
[20:30] <kierank> ok
[20:53] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:4adf75cade69: avformat/rtpdec_hevc: Fix 2 "may be used uninitialized in this function" warnings
[20:53] <cone-46> ffmpeg.git 03Michael Niedermayer 07master:ab161bfa4b3e: avformat/asfdec: Check the av_get_packet() return value
[21:04] <kierank> sigh
[21:04] <kierank> just because you can make a 720x576 testsrc in ffmpeg doesn't change the fact that the rest of the world does it right
[21:04] <kierank> and has a 702 active picture
[21:05] <nevcairiel> arent dvds typically 720x576
[21:06] <JEEB> according to the specs they should be with the active picture based AR though (often simplified)
[21:06] <kierank> yes with 702 active pixels
[21:06] <kierank> look at the downscaled parkjoy samples
[21:06] <kierank> they do it correctly
[21:06] <kierank> 9 pixels of grey on either side
[21:06] <JEEB> that's also why the PSP crops the picture
[21:06] <JEEB> when you encode 480p or 576p
[22:41] <wm4> talking about ffmpeg advancements on #libav-devel is mean
[22:47] <ubitux> yeah, sorry
[22:47] <ubitux> i was highlighted there
[22:49] <ubitux> j45: https://github.com/ubitux/FFmpeg/commit/a73a19cfc5c5e27d06b6e0cdb4ec06b1c45c8240 just an experiment, but you get the idea
[22:49] <ubitux> typically all the timing go out of the decoders
[22:49] <ubitux> and you can see the readorder heuristic
[22:50] <ubitux> the dts being in a bunch of demuxers (the one for standalone subs) handled @ https://github.com/ubitux/FFmpeg/commit/a7d1c5917c36b66f3bcdcaf369f12edbc3450e8e (but i need to do that properly in mkv demuxer at least, and probably several others)
[22:50] <ubitux> i could probably extract this https://github.com/ubitux/FFmpeg/commit/a73a19cfc5c5e27d06b6e0cdb4ec06b1c45c8240#diff-d7031e4f689033d5546668b25eb9dd30R2708 in a commit
[22:50] <ubitux> to at least have the code properly refactored 
[22:51] <ubitux> but the AVTextSubtitles thing wasn't a good idea, i realized that too late
[22:51] <ubitux> anyway, experiments.
[22:51] <ubitux> will try to get done with this... but i'm on something more fun :)
[22:56] <cone-46> ffmpeg.git 03Martin Storsjö 07release/2.4:932caa50a5de: libavformat: Check for malloc failures in avformat_new_stream
[22:56] <cone-46> ffmpeg.git 03Vittorio Giovara 07release/2.4:61335ab33e59: opt: check memory allocation
[22:56] <cone-46> ffmpeg.git 03Vittorio Giovara 07release/2.4:3280b86c85f5: swscale: check memory allocations
[22:56] <cone-46> ffmpeg.git 03wm4 07release/2.4:b96163f055ee: avformat/utils: check for malloc failure
[22:56] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:cdf37f25489f: Merge commit '932caa50a5dec7f2a69da8a417ccae84d19ae5e5' into release/2.4
[22:56] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:ee910aef38cb: Merge commit '61335ab33e594d47b84fde7367ee10dc66fc3465' into release/2.4
[22:56] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:dd4adba8112b: Merge commit '3280b86c85f5a88397fce0230d1878fee79fbb93' into release/2.4
[23:07] <cone-46> ffmpeg.git 03Luca Barbato 07release/2.4:f6c82b34a320: segment: Fix the failure paths
[23:07] <cone-46> ffmpeg.git 03Anton Khirnov 07release/2.4:bfe18be88a66: mpeg4audio: check the init_get_bits() return value
[23:07] <cone-46> ffmpeg.git 03Alexandre Colucci 07release/2.4:3d0752d82f8e: xsub: Support DXSA subtitles
[23:07] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:cc598bf7dc8a: Merge commit 'f6c82b34a320f105af266997f5951cbe7dfc8a05' into release/2.4
[23:07] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:05b1ac24709b: Merge commit 'bfe18be88a66da25b60a091de6011197dcb231fd' into release/2.4
[23:07] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:3cc8822c93e9: Merge commit '3d0752d82f8eaa326cff306ae50b0186a5b4d304' into release/2.4
[23:20] <cone-46> ffmpeg.git 03Vittorio Giovara 07release/2.4:76e9a17f3392: lavfi: always check av_expr_parse_and_eval() return value
[23:20] <cone-46> ffmpeg.git 03Vittorio Giovara 07release/2.4:1411f073fdce: aacenc: correctly check returned value
[23:20] <cone-46> ffmpeg.git 03Vittorio Giovara 07release/2.4:cbfdbba58e14: cmdutils: check file access functions return values
[23:21] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:98d4a07ccf97: Merge commit '76e9a17f3392e752193015765e9216c2f0716b96' into release/2.4
[23:21] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:0afeb76e75f0: Merge commit '1411f073fdceeff1f39dbaa035c3c0275f69095f' into release/2.4
[23:21] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:e7b8fa2c005b: Merge commit 'cbfdbba58e1460bd0791911ad84a6c76b5500a0e' into release/2.4
[23:55] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:21683549edf4: h261dec: Fix context initialization sequence
[23:55] <cone-46> ffmpeg.git 03Martin Storsjö 07release/2.4:cbf31d5f1577: rtpdec_h263_rfc2190: Clear the stored bits if discarding buffered data
[23:55] <cone-46> ffmpeg.git 03Luca Barbato 07release/2.4:9bd4561d74a4: tiff: Check the check_size() return value and forward it
[23:55] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:3c33983b7435: Merge commit '21683549edf436024cc3efbc4f8d1d55221c5336' into release/2.4
[23:55] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:c69fff3790ae: Merge commit 'cbf31d5f15774b3ffd1e2009159dc7154a767b09' into release/2.4
[23:55] <cone-46> ffmpeg.git 03Michael Niedermayer 07release/2.4:0295615f0a01: Merge commit '9bd4561d74a46dd6557140e286228e09e380674c' into release/2.4
[00:00] --- Sun Jan 18 2015