[Ffmpeg-devel-irc] ffmpeg-devel.log.20170216

[burek]

[05:33:09 CET] <FishPencil> I asked a benchmark question in #ffmpeg that might be more suited for developers to answer...
[05:38:35 CET] <cone-491> ffmpeg 03James Almer 07master:a50b3308591b: configure: remove unnecessary opus encoder dependency
[05:38:35 CET] <cone-491> ffmpeg 03James Almer 07master:ad1d0a1a06c8: configure: add missing mdct15 dependency
[10:53:08 CET] <cone-534> ffmpeg 03Matthieu Bouron 07master:5ef2f4f39480: swr/aarch64: add missing ret to ff_resample_common_apply_filter_x8_float_neon
[13:23:22 CET] <cone-534> ffmpeg 03Pavel Koshevoy 07master:e3f1b993341e: lavc/vda_h264_dec.c Fix NULL pointer dereference
[13:23:23 CET] <cone-534> ffmpeg 03Michael Niedermayer 07master:8365c1a15c94: avcodec/error_resilience: Fix "assignment from incompatible pointer type" warning
[13:48:22 CET] <cone-534> ffmpeg 03Michael Niedermayer 07master:1fa8884aeb94: avcodec/opusenc: Add () protecting macro arguments
[14:15:08 CET] <J_Darnley> Oh, I forgot to add a "version 2" to those emails.
[17:28:49 CET] <cone-534> ffmpeg 03Rostislav Pehlivanov 07master:ada281dbe2e3: Changelog: add entry for the native Opus encoder
[18:08:56 CET] <michaelni> atana, do you have time to continue working on af_peakpoints2 ?
[18:11:11 CET] <atana> yes, I am working on it. trying to fix the issue 'when file is not present it doesn't enter the fp==-1 criteria
[18:24:45 CET] <michaelni> atana, seems working fine here, avpriv_open(filename, O_RDONLY, 0444); returns -1 if the file isnt there
[18:24:59 CET] <michaelni> also you dont need the 0444
[18:30:03 CET] <atana> some it does enter retval<0 condition and keep on printing "No content read"
[18:30:10 CET] <atana> when file is not present
[18:30:23 CET] <atana> I will remove permission
[18:32:20 CET] <michaelni> atana, not sure i have the latest code but the check for fp==-1 just prints a message it doesnt free the filename and continue with the next file
[18:34:33 CET] <atana> when file is not present then I should free the filename too right?
[18:37:47 CET] <michaelni> atana, you allocate a new one for each filename, the old should be freed before or you could always reuse the memory
[19:08:07 CET] <atana> michaelni, repo is updated. Also the logic to skip entries with less than 3 peaks added
[19:16:39 CET] <michaelni> atana, if you "break;" on missing files then you would skip testing the next hash values. a missing file is completely normal, the looked up audio will not be bit by bit identical to what was stored in the database
[19:17:18 CET] <michaelni> atana, the code should continue with the next point / hash
[19:17:24 CET] <atana> oh. will fix it
[19:58:17 CET] <atana> updated repo
[20:04:42 CET] <Chloe> What would be a good way to export a language -> page id mapping from mpegts, so it is accessible from an avstream?
[20:04:59 CET] <michaelni> atana, on fp == -1 do av_freep(&filename); continue;
[20:06:22 CET] <atana> done
[20:11:05 CET] <michaelni> atana, ok, next we should force a specific sample rate otherwise the samplerate when building the database can mismatch when looking a song up which would not work
[20:11:37 CET] <atana> ok this should be done in query_format?
[20:11:47 CET] <michaelni> Yes, exactly
[20:11:57 CET] <atana> what should be sampling rate?
[20:12:53 CET] <michaelni> it doesnt matter much, 11025 or 8000 are probably good choices, more than that is probably a waste of coputation time
[20:13:10 CET] <michaelni> coMputation time
[20:14:39 CET] <michaelni> atana, you can look at query_formats() of libavfilter/af_sofalizer.c as example
[20:15:16 CET] <michaelni> af_sofalizer.c is more complex than what we need though
[20:15:21 CET] <pfanous> JEEB: do you know if there's been any bugs in H.264 decoding of NAL units in v3.2?
[20:15:47 CET] <JEEB> does FATE successfully finish?
[20:16:11 CET] <JEEB> http://fatebeta.ffmpeg.org/v3.2
[20:16:14 CET] <JEEB> seems like it does :P
[20:16:45 CET] <JEEB> but no, always always if there are bugs they are not known about. if you think you have found an issue, test with ffmpeg.c and if that breaks as well, post a bug report on trac
[20:17:14 CET] <JEEB> pfanous: also most developer-like people are useless to ask about releases since no developer uses releases, everyone uses the latest HEAD of master they've tested
[20:17:42 CET] <JEEB> releases are made purely for distros etc and can be considered "barely" maintained
[20:17:48 CET] <pfanous> ok. i see your point.
[20:18:33 CET] <pfanous> it might be a camera issue and not an ffmpeg issue, since it occurs on an infrequent basis and usually after several hours of running
[20:20:01 CET] <pfanous> the problem I'm refering to looks like an array overrun in h2645_pars.c
[20:20:37 CET] <pfanous> might be that the camera is encoding the NAL data improperly
[20:22:10 CET] <pfanous> in ff_h2645_extract_rbsp()
[20:26:48 CET] <JEEB> Chloe: check how the MPEG-TS PID side data was pushed? I think that was on the AVStream?
[20:27:43 CET] <nevcairiel> its on the packet as well
[20:28:04 CET] <nevcairiel> but yes we do have stream-global sidedata
[20:28:43 CET] <JEEB> because if new stuff has to be exported from AVStream it could be either a new field, or new type of side data
[20:29:07 CET] <nevcairiel> fields for singular use-cases are frowned upon
[20:29:28 CET] <JEEB> yeah
[20:31:19 CET] <Chloe> I thought as much. I guess I can just add another AVPacketSideDataType, and export the mapping through there AVStream side_data
[20:38:00 CET] <JEEB> sounds good
[21:06:34 CET] <atana> michaelni, repo updated 
[21:17:56 CET] <michaelni> atana, ok, you should test the return code of ff_set_common_samplerates() for failure
[21:18:36 CET] <atana> failure is ret==-1?
[21:19:12 CET] <michaelni> failure is ret <0, also you could similar to the sample rate also force a channel layout of mono, none of the fingerprint algos discussed does anything with more than 1 channel
[21:20:29 CET] <michaelni> atana, that is by using ff_set_common_channel_layouts()
[21:20:51 CET] <atana> ok
[21:32:55 CET] <llogan> we should make our own version of KernelPRBot: https://github.com/torvalds/linux/pull/389#issuecomment-277539188
[21:35:03 CET] <wm4> we can't steal theirs?
[21:36:09 CET] <llogan> only the message itself seems to be shared and not the mechanism behind the account: https://github.com/ajdlinux/KernelPRBot
[21:37:06 CET] <llogan> i can ask the author if he will provide details
[21:40:50 CET] <llogan> so many people think github *is* git...
[21:49:11 CET] <atana> michaelni, updated repo. I am not quite sure how to add mono layout
[21:50:06 CET] <wm4> llogan: terrifying
[21:58:25 CET] <llogan> wm4: email sent to bot author. i'll let you know what he says.
[22:02:30 CET] <wm4> nice
[22:51:35 CET] <Chloe> Looks like the extra data was already being copied. Does this look correct for documenting it? https://0x0.st/4Wp.patch The reason why I am doubtful is that the actual teletext pages have been >255, thus not just one bit, so I'm a bit confused on how the pages are being worked out.
[23:45:18 CET] <cone-179> ffmpeg 03Mulvya 07master:64b79535d7c4: doc: correct order of options for channelmap filter
[00:00:00 CET] --- Fri Feb 17 2017