[Ffmpeg-devel-irc] ffmpeg-devel.log.20151110

[burek]

[00:42:34 CET] <cone-516> ffmpeg 03Ganesh Ajjanagadde 07master:b87ca4bf25a3: swresample/resample: speed up upsampling by precomputing sines
[00:42:35 CET] <cone-516> ffmpeg 03Ganesh Ajjanagadde 07master:cf491a925e22: swresample/resample: speed up Blackman Nuttall filter
[01:32:21 CET] <cone-516> ffmpeg 03Ganesh Ajjanagadde 07master:6c2dbff7f08c: ffserver: fix incorrect strlcpy usage
[02:10:03 CET] <atomnuker> michaelni: on it, some parts don't look like good hacks, looking into the specs
[02:15:18 CET] <atomnuker> also thanks for whoever put me on this channel's access list
[02:35:38 CET] <cone-516> ffmpeg 03Andreas Cadhalpun 07master:ef7fe9851e09: aacps: avoid division by zero in stereo_processing
[03:18:20 CET] <cone-516> ffmpeg 03Andreas Cadhalpun 07master:d7f29bfa6985: aacdec: don't return frames without data from aac_decode_er_frame
[04:24:04 CET] <atomnuker> "aacps_fixed_tablegen.c:(.text.startup+0x9e1): undefined reference to `av_log'"
[04:24:12 CET] <atomnuker> when compiling with hardcoded tables
[05:04:46 CET] <cone-516> ffmpeg 03Michael Niedermayer 07master:b83c849e8797: tests/fate/avformat: Fix fate-lavf
[05:45:33 CET] <cone-516> ffmpeg 03Timothy Gu 07master:480bad7c6d08: avformat: Add v210[x] demuxer
[11:21:39 CET] <nevcairiel> dont pay attention for a couple days, and half my fate boxes show random errors from code breakages <.<
[12:41:47 CET] <JoshX> what do those mean btw?
[12:42:08 CET] <JoshX> whoops :) sorry :)
[13:20:23 CET] <Compn> guess llogan is still away
[13:20:30 CET] Action: Compn reviews some mod queuews
[13:20:52 CET] <Compn> is there a list of lists i should be checking
[13:21:05 CET] <Compn> ffmpeg-user , libav-user , ffmpeg-devel ... what else
[13:21:44 CET] <Compn> ets review -cvslog for fun i gues
[13:26:26 CET] <Compn> theres one mail in russian with a link to an ogg of major lazer's 'lean on'. so confusing.
[13:28:56 CET] <Compn> nah just spam
[13:46:21 CET] <cone-726> ffmpeg 03Michael Niedermayer 07master:8a024f6a4344: avcodec/aacsbr_fixed: Try to initialize sum[0..1] differently to fix build with VS2012
[13:47:28 CET] <cone-726> ffmpeg 03wm4 07master:dafe4cd29cad: mmaldec: send only a single EOS packet on flushing
[14:00:58 CET] <wm4> so how does our current decode API work again
[14:01:15 CET] <wm4> as soon as got_picture returns 0, and the input was a flush packet, it means the decoder is fully flushed, right?
[14:01:31 CET] <nevcairiel> yes
[14:01:49 CET] <nevcairiel> or an error return
[14:02:03 CET] <wm4> and what is the behavior if you sent more flush packets to a flushed decoder, or send new input data to a flushed decoder?
[14:03:34 CET] <nevcairiel> more flush packets just yield nothing, more got_picture 0 returns, i suppose
[14:03:47 CET] <rcombs> probably not reliable
[14:03:47 CET] <nevcairiel> its technically an invalid call, but probably harmless
[14:04:22 CET] <wm4> is it ok to send a flush packet to an "empty" decoder?
[14:04:41 CET] <nevcairiel> i dont see why not, thats strictly speaking possible
[14:04:47 CET] <wm4> I'd also say yes, but it's quite a corner case actually
[15:35:27 CET] <durandal_1707> what you insensitive folk have against OSS?
[15:37:00 CET] <J_Darnley> What?
[15:37:27 CET] <wm4> wrong channel
[15:40:19 CET] Action: J_Darnley nods
[15:44:22 CET] <wm4> nevcairiel: what do we say about sending new packets to a partially flushed decoder?
[15:45:59 CET] <durandal_1707> wm4: you have nothing to say against OSS?
[15:46:07 CET] <nevcairiel> sounds like a crazy thing to do
[15:46:11 CET] <nevcairiel> why would you partially flush it
[15:46:49 CET] <wm4> nevcairiel: demuxer reports EOF, and then suddenly recovers
[15:46:57 CET] <nevcairiel> then de muxer is bugged
[15:47:04 CET] <wm4> not necessarily
[15:47:11 CET] <nevcairiel> either the file is over, or it aint
[15:47:13 CET] <wm4> packets could be sparse or stop for a while
[15:47:19 CET] <wm4> haha, good one
[15:47:48 CET] <nevcairiel> you wouldnt even keep calling the demuxer when it once reported EOF
[15:48:11 CET] <nevcairiel> also, why wouldnt you just flush the decoder entirely at least
[15:48:36 CET] <wm4> so you would say this API usage is invalid
[15:50:49 CET] <fritsch> from an audio pov it should not matter how often you call flush 
[15:50:56 CET] <fritsch> if it's already flushed it should be noop
[15:51:10 CET] <fritsch> if you see for example PA API which has a flush command
[15:51:23 CET] <fritsch> that you can call whenever and how often you want - after the stream is setup
[15:55:43 CET] <nevcairiel> the majority of audio decoders dont need a flush anyway since they dont have a concept of delay
[15:57:25 CET] <fritsch> PA is not a decoder, but a playback / audio engine api
[15:57:37 CET] <fritsch> so - they need it, if you want to play all samples
[15:57:51 CET] <nevcairiel> we're talking about ffmpeg, it does decoding =P
[15:58:01 CET] <fritsch> jep
[15:58:13 CET] <fritsch> though flushing an already flushed decoder
[15:58:19 CET] <fritsch> should just say: OKAY
[15:58:22 CET] <fritsch> or something :-)
[15:58:30 CET] <fritsch> and return
[15:58:58 CET] <nevcairiel> it probably will on most of them, but not sure its a guarantee
[15:59:12 CET] <fritsch> if the decoder can "flush" it self
[15:59:14 CET] <fritsch> in case of error
[15:59:18 CET] <fritsch> it even _must_ do that
[15:59:33 CET] <fritsch> cause if not - the API would not be well defined anymore
[16:01:37 CET] <durandal_1707> what demuxer are we talking about?
[16:02:07 CET] <fritsch> i think I skipped a bit of the talk and still answered to the question from this morning
[16:02:15 CET] <fritsch> where we talked about "the decoder"
[16:04:11 CET] <cone-726> ffmpeg 03Michael Niedermayer 07master:6e3244fb09f2: avcodec/g729postfilter: Avoid function calls in FFMAX() arguments
[17:38:19 CET] <cone-726> ffmpeg 03James Almer 07master:9f4a41bf9919: avutil/softfloat: use abort() instead of av_assert0(0)
[17:39:49 CET] <cone-726> ffmpeg 03Luca Barbato 07master:1e7ff5ac6923: nut: Use the correct codec_tag when multiple are available
[17:39:50 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:d786cdff1ca5: Merge commit '1e7ff5ac6923996f7292c82f102c68384fbc9d97'
[17:44:17 CET] <cone-726> ffmpeg 03Luca Barbato 07master:0b699920f3c0: lagarith: Correctly compute hash_shift
[17:44:18 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:2503aa6bd912: Merge commit '0b699920f3c0b2e30969b38ce59226653f9d80ec'
[17:54:38 CET] <cone-726> ffmpeg 03Luca Barbato 07master:60f50374f195: rpza: Check the blocks left before processing one
[17:54:39 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:8600fef123bf: Merge commit '60f50374f1955442dc987abc4a6c61c2109620c2'
[17:56:40 CET] <cone-726> ffmpeg 03Luca Barbato 07master:4dfbc7a7559c: msnwc_tcp: Correctly report failure
[17:56:41 CET] <cone-726> ffmpeg 03Luca Barbato 07master:f8d10511fef0: cosmetics: msnwc_tcp: Reformat
[17:56:42 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:2e82d51714c7: Merge commit '4dfbc7a7559ccab666a8fd39de4224eb4b02c768'
[17:56:43 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:618817b6a8f2: Merge commit 'f8d10511fef0bb3265f2d7bdfd38f90f60731d80'
[17:57:12 CET] <cone-726> ffmpeg 03Luca Barbato 07master:452d659aa4a7: msnwc_tcp: Use the correct return values
[17:57:13 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:b40e43c5c401: Merge commit '452d659aa4a742106c70ffe9ef3df47ef5b81ea6'
[17:58:12 CET] <cone-726> ffmpeg 03Luca Barbato 07master:4a326923381f: mimic: Always return on failure
[17:58:13 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:446ac80ea594: Merge commit '4a326923381f2bc0fb7d960c8f8366a5456bc4ab'
[17:59:31 CET] <cone-726> ffmpeg 03Luca Barbato 07master:de41b555cdea: truemotion2: Fix the buffer check
[17:59:32 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:0c1a6f577bbc: Merge commit 'de41b555cdea2dcacbe98ee9edc83a8c15c73c4c'
[18:02:18 CET] <cone-726> ffmpeg 03Luca Barbato 07master:50d2a3b5f34e: flashsv: Initialize the block array
[18:02:19 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:3a8b37c5904f: Merge commit '50d2a3b5f34e6f99e5ffe17f2be5eb1815555960'
[18:08:25 CET] <cone-726> ffmpeg 03wm4 07master:f128b8e19ac7: mov: detect cover art pictures by content
[18:08:26 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:df06cb561102: Merge commit 'f128b8e19ac7f702adae899ab91cc1e80f238761'
[18:08:37 CET] <cone-726> ffmpeg 03Luca Barbato 07master:e2854e731f84: avresample: Document avresample_open() a little better
[18:08:38 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:240ac53a5ccc: Merge commit 'e2854e731f843906d9a9a5b882bed872341999fd'
[18:14:48 CET] <cone-726> ffmpeg 03John Stebbins 07master:2ec112f71cd0: vf_pad: fix x, y option expression evaluation
[18:14:49 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:43266457b4f9: Merge commit '2ec112f71cd03ccab1b6f9a00d29199a57bcc7a5'
[18:17:27 CET] <cone-726> ffmpeg 03Paul B Mahol 07master:c012c6f1a8b3: segafilm: implement seeking
[18:17:28 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:b3453f44edae: Merge commit 'c012c6f1a8b34828a7870dc1854422934f14b79a'
[18:17:59 CET] <cone-726> ffmpeg 03Paul B Mahol 07master:1e791ee3aae6: segafilm: set video and audio stream duration
[18:18:00 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:0e29820cd164: Merge commit '1e791ee3aae63122afb10cbabe86a747aea58992'
[18:19:02 CET] <cone-726> ffmpeg 03Michael Niedermayer 07master:563e6d860391: segafilm: drop the "song and dance" for cinepak
[18:19:03 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:42a088784b94: Merge commit '563e6d860391bac0511984e5c0842320b5c94d2d'
[18:19:18 CET] <cone-726> ffmpeg 03Michael Niedermayer 07master:fcc49924db6c: vdpau: Remove a spurious CONFIG_H263_VDPAU_HWACCEL
[18:19:19 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:947447bfdc69: Merge commit 'fcc49924db6c5530a3346c45a6999244d92eaf48'
[18:20:14 CET] <cone-726> ffmpeg 03Michael Niedermayer 07master:cc96018c79cd: libopenh264enc: Set AVOption data type
[18:20:15 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:23277aa24c1d: Merge commit 'cc96018c79cd4dc823e0cef938da2131e4660fe5'
[18:24:30 CET] <cone-726> ffmpeg 03Anton Khirnov 07master:a0c71a575109: lavf: initialize cur_dts to AV_NOPTS_VALUE for muxing
[18:24:31 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:de0e219a8aba: Merge commit 'a0c71a575109f123978e345fa7eb4ac03cd4d3c3'
[18:38:29 CET] <cone-726> ffmpeg 03Anton Khirnov 07master:3efd71b4d0b4: avconv: set packet duration for CFR video streams
[18:38:30 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:d3d4bc4784d9: Merge commit '3efd71b4d0b4a73ccbbbdc092e6bbd54d92633f4'
[18:44:00 CET] <cone-726> ffmpeg 03Anton Khirnov 07master:aeda1121c913: lavf: factor out freeing an AVStream
[18:44:01 CET] <cone-726> ffmpeg 03Hendrik Leppkes 07master:3df9ec5ce7bc: Merge commit 'aeda1121c91395d1f0d434bc7f227423aa2ff97d'
[19:43:12 CET] <cone-726> ffmpeg 03Paul B Mahol 07master:a0f75c46d31c: avformat/wavdec: parse XMA2 tag
[19:43:13 CET] <cone-726> ffmpeg 03Paul B Mahol 07master:4e583eda5f5c: avformat/wavdec: allow data chunk to be before fmt/xma2 chunk
[21:02:07 CET] <cone-726> ffmpeg 03Paul B Mahol 07master:5b349c8d7cc5: aviflter/vf_waveform: speed up non-color filters
[21:06:13 CET] <jamrial> http://chrome.blogspot.com/2015/11/updates-to-chrome-platform-support.html maybe we could follow suit?
[21:10:14 CET] <wm4> michaelni_: is it normal that if I use swr_set_compensation(), the output is usually 1 sample different from what I requested, or am I making a stupid error?
[21:22:22 CET] <michaelni_> maybe theres a off by 1 error, how can this be reproduced?
[21:22:25 CET] <rcombs> anyone know if Tinglin Liu is on IRC?
[21:22:48 CET] <rcombs> looking at an oddity in 9ea812692c38c58a343cec5f66e8949655906edc
[21:23:41 CET] <J_Darnley> jamrial: what features do you want to drop or include that require newer than XP?
[21:24:41 CET] <jamrial> at the least it would simplify the threading code
[21:25:35 CET] <rcombs> https://gist.github.com/a1046300acc6e23d418a a couple of the checks seem to be swapped
[21:26:44 CET] <cone-726> ffmpeg 03Paul B Mahol 07master:9c66428016c3: avcodec/wmaprodec: XMA1 actually have packet sequence numbers
[21:27:16 CET] <wm4> michaelni_: I'm calling swr_set_compensation before every swr_resample call, with a delta of about 5 samples
[21:27:57 CET] <wm4> passing in an audio frame of 1024 (so delta is 5 pts units, distance is 1029 pts units)
[21:36:28 CET] <durandal_1707> pross: why 2018?
[21:37:43 CET] <rcombs> actually that patch is wrong
[21:38:49 CET] <rcombs> or, hmm
[21:39:21 CET] <rcombs> I dunno why "found_hdlr_mdta" exists at all
[21:41:46 CET] <nevcairiel> i think the current check is fine, if the function would continue with nb_streams == 0, it would just crash
[21:42:08 CET] <rcombs> nevcairiel: yeah, that's definitely wrong, my bad
[21:42:14 CET] <rcombs> mixed up 2 relevant functions
[21:42:39 CET] <rcombs> nevcairiel: but why set `found_hdlr_mdta` at all, and why check for it later?
[21:43:00 CET] <nevcairiel> i know shit about mov
[21:44:03 CET] <nevcairiel> are there actually any problems or just reviewing code
[21:44:54 CET] <rcombs> right now it doesn't parse quicktime meta if the first hdlr atom with type mdta occurs after the first trak
[21:45:50 CET] <rcombs> because found_hdlr_mdta only gets set if it's before the first trak, and the quicktime meta parsing code only runs if found_hdlr_mdta is set
[21:45:55 CET] <rcombs> and I have no idea why
[21:46:23 CET] <nevcairiel> i see
[21:51:47 CET] <rcombs> so either we move the `type == MKTAG('m','d','t','a')` check to before the `nb_streams < 1` check instead of inside it, or just remove `found_hdlr_mdta` and everything that checks it
[21:52:32 CET] <rcombs> the latter seems to make more sense to me, but I'm a little concerned about removing it without understanding what it's for
[21:54:10 CET] <nevcairiel> mail the guy and find out
[21:55:03 CET] <rcombs> guess that works
[22:07:35 CET] <cone-726> ffmpeg 03Andreas Cadhalpun 07master:7b67fe20f6c5: brstm: reject negative sample rate
[23:24:27 CET] <cone-726> ffmpeg 03Ni Hui 07master:c75640e9c9e1: avcodec/lzw: ff_lzw_decode_tail() now returns the consumed bytes in lzw decompress
[23:24:28 CET] <cone-726> ffmpeg 03Ni Hui 07master:3ea60c505f3f: avcodec/gifdec: skip the data lzw consumed
[23:45:59 CET] <cone-726> ffmpeg 03Michael Niedermayer 07master:d9726893f311: avformat/mxfenc: Only store user comment related tags when needed
[00:00:00 CET] --- Wed Nov 11 2015