[Ffmpeg-devel-irc] ffmpeg-devel.log.20150421

[burek]

[00:12:20 CEST] <cone-905> ffmpeg 03Michael Niedermayer 07master:e6e8cc8ce9c2: avformat/img2dec: do not rewind custom io buffers
[00:35:28 CEST] <kierank> wm4: so as far as I can tell it's not possible to link to avcodec in /tests
[00:35:32 CEST] <kierank> which is annoying
[00:41:58 CEST] <cone-905> ffmpeg 03Deb Mukherjee 07master:26a0bcb15b83: Support for VP9 high-color/high-bit-depth encoding
[00:49:56 CEST] <kierank> anyone knowledgable about the Makefile system?
[01:04:41 CEST] <jamrial> kierank: check doc/examples/Makefile, it links to libraries outside of their respective folders to compile example programs
[01:05:00 CEST] <jamrial> it could be adapted to compile these api test programs from inside tests/
[01:05:25 CEST] <kierank> thanks
[01:08:16 CEST] <cone-905> ffmpeg 03Michael Niedermayer 07master:ba631b791435: avformat/utils: Ensure that AVFMT_FLAG_CUSTOM_IO is set before use
[01:15:03 CEST] <wm4> kierank: huh, but aren't there other tests using lavf etc.?
[01:15:18 CEST] <kierank> they all use ffmpeg.c
[01:18:30 CEST] <wm4> I wasted so much time on ffmpeg.c's bullshit semantics for -ss
[01:29:41 CEST] <cone-905> ffmpeg 03Vignesh Venkatasubramanian 07master:30ba28fe8e57: webmdashenc: Fix potential memory leak
[02:03:01 CEST] <cone-905> ffmpeg 03Michael Niedermayer 07master:ecf34e8fc2f7: avformat/mp3dec: Adjust returned timestamp in the CBR seeking case
[02:25:46 CEST] <cone-905> ffmpeg 03Michael Niedermayer 07master:a924b83cccf2: avformat/nsvdec: remove trace log from probe(), this is better done generically not per format
[02:39:18 CEST] <BBB> is there a way to force ffmpeg to not resize?
[02:39:41 CEST] <BBB> like, if my input resizes from size axb to cxd and I just want these frames to be sent like that to the encoder (rawenc+framemd5 in this case)
[04:21:26 CEST] <ubitux> alright, i'm soon back in the game.
[04:23:20 CEST] <jamrial> \o/
[04:29:40 CEST] <ubitux> oh, colorkey filter, cool
[04:57:04 CEST] <ubitux> note: the style of the flac test is wrong; i think i saw someone raise the issue, it could be nice if it could be fixed
[05:24:04 CEST] <ubitux> rcombs: are you feeling like adding test(s) for the ASS thing?
[05:24:53 CEST] <ubitux> i think it would be nice to have a bunch of small mkv with all kind of crazy ass muxed in (with different pattern of extradata/codecprivate sickness)
[05:26:41 CEST] <rcombs> ubitux: that sounds like an excellent idea, if I can work up the motivation to learn to write fate tests
[05:26:54 CEST] <rcombs> or do you want unit tests
[05:26:57 CEST] <rcombs> (or, is fate unit tests?)
[05:27:02 CEST] <rcombs> (I don't know things)
[05:27:15 CEST] <ubitux> there are functionnal tests
[05:27:23 CEST] <ubitux> it's generally calls to ffmpeg
[05:27:31 CEST] <ubitux> to do very specific tasks
[05:27:38 CEST] <ubitux> (it has various other tools though)
[05:27:46 CEST] <ubitux> but in this case, a demux will be enough
[05:28:01 CEST] <ubitux> look at tests/fate/subtitles.mak
[05:28:45 CEST] <ubitux> and add maybe fate-mks-[1..N] tests, that will be triggered by make fate-subtitles (and thus make fate)
[05:29:02 CEST] <ubitux> see http://ffmpeg.org/fate.html
[05:29:10 CEST] <ubitux> if you need to upload one or more sample, ask michaelni 
[05:29:17 CEST] <ubitux> ideally, small ones
[05:30:26 CEST] <ubitux> typically, i think having extradata with various crap like no [Events] header, some with V4 style, some with V4+ styles, and that sort of madness might help
[05:30:33 CEST] <ubitux> but anyway... :)
[05:31:00 CEST] <ubitux> i'm not sure how to construct easily such stream, so i didn't bother doing it
[07:16:07 CEST] <Timothy_Gu> wait onemda is from Croatia right?
[09:47:44 CEST] <wm4> michaelni: what happens if non-mp3 data is fed to the mp3 parser or the dcoder? like a id3v1 tag?
[09:47:47 CEST] <wm4> *decoder
[09:50:18 CEST] <wm4> ubitux: didn't you want some subtitle tests from me? I feel like adding some now
[10:43:27 CEST] <durandal_1707> wm4: where you found 24bit Apng sample?
[11:05:53 CEST] <wm4> durandal_1707: not sure, and I don't have it anymore
[11:07:48 CEST] <michaelni> parsers should generally pass data through they dont recognice, the decoder should ignore it i think but i dont know if it does
[11:09:18 CEST] <wm4> I guess I'll check...
[14:27:09 CEST] <cone-874> ffmpeg 03Michael Niedermayer 07master:8f9a3817d502: avfilter/vf_qp: fix cleanup on error
[15:00:42 CEST] <wm4> michaelni: with our mp3 test sample, the CBR seek actually gets off by one frame towards the end of the file
[15:01:11 CEST] <wm4> michaelni: I found out that mpg123 actually indexes the file while seeking (instead of skipping it without reading the data)
[15:01:29 CEST] <wm4> michaelni: I want to implement the same thing for libavformat (we discussed it to some degree)
[15:06:38 CEST] <michaelni> libavformat should already index and use the index if seek otherwise fails (like no cbr no toc)
[15:07:57 CEST] <wm4> well, I'm not sure what exactly it does in this case
[15:08:06 CEST] <nevcairiel> unfortunately "should" is not a "actually does" =p
[15:08:57 CEST] <nevcairiel> also, it only indexes parts you already played, seeking into the future doesnt work that way ... plus on vbr samples, it doesnt even figure out how long an audio file actually is, which a full index would also cure
[15:09:24 CEST] <wm4> I suppose there's some chance what I want it to do is already buried there (like if the seek target is beyond the last index entry, it keeps reading packets until the target is reached)
[15:09:45 CEST] <wm4> i.e. lazy indexing
[15:09:52 CEST] <wm4> a full index would be nice too
[15:10:02 CEST] <wm4> it'd be possible to actually seek in files with timestamp resets
[15:10:26 CEST] <nevcairiel> for some reason people want to keep using my stuff for audio files, including vbr stuff, and they always whine when the duration is off :(
[15:10:42 CEST] <wm4> what file types does this affect?
[15:10:51 CEST] <wm4> mp3 with xing headers should be ok (just seeking is broken)
[15:10:54 CEST] <nevcairiel> usually thats raw audio files
[15:11:00 CEST] <wm4> ah that stuff
[15:11:01 CEST] <nevcairiel> ie. raw aac
[15:11:14 CEST] <wm4> apparently it's popular to use such files as external audio tracks
[15:11:27 CEST] <nevcairiel> i keep telling them to shove it into m4a, but they wont listen
[15:11:38 CEST] <wm4> so it's for use with video, and naturally this requires accurate seeking
[15:11:44 CEST] <wm4> stupid crap
[15:12:05 CEST] <nevcairiel> i'm kinda surprised i dont get complaints from the external audio track crowd
[15:21:35 CEST] <cone-874> ffmpeg 03Paul B Mahol 07master:77e794908b18: avcodec/hq_hqa: remove superfluous log message
[15:32:21 CEST] <wm4> ok it seems to work
[15:32:34 CEST] <wm4> I only had to remove the seek callback and not add the xing entries
[15:33:00 CEST] <wm4> michaelni: so what do you say, use the generic index creation & seek code, and remove all the mp3 seek code?
[15:33:23 CEST] <wm4> or shall there be a compatibility option to use the old mode (on or off by default?)
[15:36:39 CEST] <BBB> michaelni: whats the practical effect of your patch?
[15:36:45 CEST] <BBB> michaelni: the remove-memalign-warning one
[15:36:54 CEST] <BBB> michaelni: youre not stating what your actual goal is
[15:37:05 CEST] <BBB> michaelni: do you want to support mixing av_realloc and av_malloc?
[15:37:37 CEST] <wm4> BBB: I believe that's the intention
[15:38:40 CEST] <BBB> so is mac an affected platform?
[15:39:20 CEST] <wm4> realloc removes the alignment (I think), so this patch can unfortunately not work (I think)
[15:45:36 CEST] <kierank> interesting how nicholas suddenly cares about the POSIX spec
[15:46:56 CEST] <wm4> he didn't before?
[15:49:16 CEST] <kierank> no but he complains that I follow specs too closley
[15:49:19 CEST] <kierank> closely
[15:50:29 CEST] <wm4> oh I see
[16:04:36 CEST] <robUx4> hi
[16:06:41 CEST] <robUx4> do you have any sample where the decoder profile changes mid-stream ?
[16:08:03 CEST] <robUx4> H264, MPEG4.2 or any video codec where it's possible
[16:09:03 CEST] <nevcairiel> http://fate-suite.ffmpeg.org/h264/reinit-small_420_8-to-large_444_10.h264
[16:09:40 CEST] <robUx4> thx
[16:09:40 CEST] <nevcairiel> this changes practically everything there is to change, size, chroma, bitdepth
[16:10:56 CEST] <michaelni> BBB, i dont want to mix them, but rather they are likely mixed in various codepathes and finding and avoiding it all costs time that could be better spend elsewheer IMHO
[16:11:24 CEST] <BBB> michaelni: I dont see why it matters then
[16:11:31 CEST] <BBB> michaelni: as in, I dont think the patch makes sense
[16:11:33 CEST] <robUx4>     Stream #0:0: Video: h264 (High 4:4:4 Predictive), yuv420p, 240x196, 25 fps, 25 tbr, 1200k tbn, 50 tbc
[16:11:33 CEST] <robUx4> [h264 @ 04a97320] Reinit context to 352x288, pix_fmt: yuv444p10le
[16:11:43 CEST] <robUx4> that's what I needed, thx
[16:11:45 CEST] <BBB> michaelni: its not like youre actively spending time trying to find and identify such codepaths, right?
[16:11:55 CEST] <michaelni> BBB, yes i see what you mean
[16:11:59 CEST] <nevcairiel> I think the warning should stay in, if it happens to cause problems in practice we can always fix them once reported as such, i dont think anyone is actively chasing after such cases
[16:11:59 CEST] <BBB> michaelni: just in some cases youll find a crash, and itll turn out to be related to this
[16:12:03 CEST] <BBB> michaelni: so you fix the bug
[16:12:12 CEST] <BBB> michaelni: so I dont see how it costs time to keep the comments where they are
[16:12:42 CEST] <nevcairiel> ..not that I care personally, Windows has an aligned realloc :D
[16:12:47 CEST] <BBB> kierank: I think you sometimes just need a big teddybear hug
[16:12:58 CEST] <BBB> nevcairiel: but we dont use it, do we?
[16:12:59 CEST] <BBB> :-p
[16:13:03 CEST] <nevcairiel> i think we do
[16:13:06 CEST] <BBB> oh rly
[16:13:08 CEST] <BBB> thats cool
[16:13:30 CEST] <nevcairiel> yes we do
[16:13:33 CEST] Action: nevcairiel checked
[16:20:27 CEST] <michaelni> wm4, about mp3, users should have the option to use cbr/toc based seeking if they dont want to index the whole file for a seek to near the end, i have no oppinion on what should be default
[16:21:35 CEST] <wm4> then I'd say let's try with actually correct seeking on by default
[18:05:38 CEST] <nalply> Did I spam ffmpeg-devel? If so I am sorry for my mistake.
[18:06:56 CEST] <Compn> im not seeing any large spams...
[18:07:39 CEST] <wm4> git send-email misusage?
[18:07:50 CEST] <nalply> wm4: Yes. Some members got spam from my in form of CC emails.
[18:08:02 CEST] <wm4> yeah, I got some too
[18:08:08 CEST] <wm4> but nothing on the actual list
[18:08:08 CEST] <nalply> Terribly sorry!
[18:08:45 CEST] <nevcairiel> the auto-CC behavior of git send-email is kinda annoying
[18:09:11 CEST] <nalply> It's because I did a merge
[18:09:32 CEST] <Timothy_Gu> lol
[18:10:16 CEST] <Timothy_Gu> nalply: are you "Daniel Ly"?
[18:10:20 CEST] <nalply> Yes
[18:10:35 CEST] <Timothy_Gu> ah yep i got one too
[18:10:50 CEST] <Timothy_Gu> You should probably say sorry on #libav-devel too
[18:11:43 CEST] <nalply> Thanks for the tip.
[18:12:26 CEST] <nalply> I will be back Thursday and retry.
[18:49:32 CEST] <kierank> michaelni: so what do I need to do for outreachy
[18:49:35 CEST] <kierank> I don't see this button
[19:01:54 CEST] <michaelni> kierank, on https://outreachy.gnome.org/?q=view_projects&prg=4&p=717 theres a "Willing to mentor" button somewhere at the bottom
[19:02:04 CEST] <kierank> ah thanks
[19:02:38 CEST] <kierank> ah you click on the project name
[19:03:17 CEST] <michaelni> thanks, it worked, it lists you now
[19:04:26 CEST] <kierank> wtf R = floor (10 * min (1, 1 / average_reply_delay)) [use days for all units]
[19:09:00 CEST] <kierank> reynaldo: ping
[20:28:07 CEST] <cone-874> ffmpeg 03Luca Barbato 07master:9b2c57bef5e2: drawtext: Add an alpha option
[20:28:08 CEST] <cone-874> ffmpeg 03Michael Niedermayer 07master:122150168f7d: Merge commit '9b2c57bef5e2f3f61a5fd708ba7d5351f5b3f386'
[20:37:38 CEST] <reynaldo> kierank: pong
[20:37:50 CEST] <reynaldo> sry for the delay, was having lunch
[20:37:54 CEST] <reynaldo> kierank: still there?
[21:17:29 CEST] <kierank> reynaldo: YES
[21:17:32 CEST] <kierank> whoops
[21:17:34 CEST] <kierank> caps lock
[21:31:13 CEST] <reynaldo> kierank: sec please, will ping you in a bit
[21:45:46 CEST] <reynaldo> kierank: ok, here now. might take a minute or two to reply but go ahead
[21:45:58 CEST] <kierank> reynaldo: can you explain how the spreadsheet works
[21:46:10 CEST] <kierank> because there are candidates who haven't submitted anything who have a high score
[21:46:20 CEST] <reynaldo> no no
[21:46:50 CEST] <reynaldo> those "scores" you are seeing are just leftover data from the previous page of the spreadsheet where I copied this from
[21:47:06 CEST] <reynaldo> just fill the non calculated fields for your intern and please ignore the rest
[21:47:45 CEST] <reynaldo> remember you are supposed to be working on the second sheet, the first one is the previous opw/outreachy term
[21:48:43 CEST] <kierank> oic
[22:01:38 CEST] <reynaldo> kierank: coolio, let me know if you need help with anything else
[22:01:41 CEST] <reynaldo> and thanks for handling this
[22:01:54 CEST] <reynaldo> will you be around for the meeting latter ?
[22:02:19 CEST] <reynaldo> its in one hour from now
[22:02:36 CEST] <kierank> yes
[22:02:41 CEST] <kierank> a bit late but i should be there
[22:22:46 CEST] <Timothy_Gu> michaelni: can you backport 7ad27f1221562b037178d1fb605510ba349f04cc?
[22:23:51 CEST] <Timothy_Gu> it would be cool if you can do that now cuz I manually fixed everything in fate.ffmpeg.org to form a proper display: http://fate.ffmpeg.org:8080/
[22:24:04 CEST] <Timothy_Gu> but then some new results came in and messed it up again.
[22:38:37 CEST] <michaelni> Timothy_Gu, locally backported to 2.6 - 2.3, it doesnt apply cleanly to 2.2
[22:38:38 CEST] <Timothy_Gu> michaelni: also: diff -u <(curl -q 'http://fate.ffmpeg.org:8080/log/mips-ubuntu-qemu-gcc-4.4/20150421091735/comp
[22:38:41 CEST] <Timothy_Gu> ile' | grep warning | sort) <(curl -q 'http://fate.ffmpeg.org:8080/log/mips-ubuntu-qemu-gcc-4.4/20150421171722/compile' | grep warning | sort)
[22:38:45 CEST] <Timothy_Gu> michaelni: thx
[22:38:51 CEST] <Timothy_Gu> ooh
[22:39:31 CEST] <Timothy_Gu> diff -u <(curl -q 'http://fate.ffmpeg.org:8080/log/mips-ubuntu-qemu-gcc-4.4/20150421091735/compile' | grep warning | sort) <(curl -q 
[22:39:34 CEST] <Timothy_Gu> 'http://fate.ffmpeg.org:8080/log/mips-ubuntu-qemu-gcc-4.4/20150421171722/compile' | grep warning | sort)
[22:39:37 CEST] <Timothy_Gu> wat
[22:39:40 CEST] <Timothy_Gu> but you get the idea
[22:40:21 CEST] <Timothy_Gu> http://sprunge.us/FHdE?diff
[22:44:09 CEST] <michaelni> thx, will fix it in a moment
[22:47:34 CEST] <jamrial> Timothy_Gu: that's looking really nice. glad to finally see branches separate from master :p
[22:49:01 CEST] <Timothy_Gu> jamrial: yeah
[22:50:42 CEST] <jamrial> there's a 2.2 slot listed under master, though
[22:55:01 CEST] <cone-874> ffmpeg 03Michael Niedermayer 07master:a79cbc0b8e77: avfilter/vf_qp: Fix use of uinitialized variable
[22:55:56 CEST] <Timothy_Gu> jamrial: see the discussion above. Michael didn't backport the commit that fixes it, so I had to manually fix all the summary files. And that result just came in so I didn't fix that one yet.
[22:57:11 CEST] <Timothy_Gu> jamrial: that result fixed
[22:58:51 CEST] <jamrial> ah ok
[23:03:08 CEST] <GabrielYYZ> hello
[23:03:36 CEST] <cone-874> ffmpeg 03Michael Niedermayer 07release/2.5:3f629c83b462: avformat/img2dec: do not rewind custom io buffers
[23:03:37 CEST] <cone-874> ffmpeg 03Michael Niedermayer 07release/2.5:5fd5b8ef79c0: avformat/utils: Ensure that AVFMT_FLAG_CUSTOM_IO is set before use
[23:03:38 CEST] <cone-874> ffmpeg 03Timothy Gu 07release/2.5:a2ad7fdb3a26: fate: Include branch information in the payload header
[23:06:03 CEST] <GabrielYYZ> i don't know if you've noticed but building with hardcoded-tables enabled is broken, i did a git bisect and it seems commit 8f7b022 is the bad one
[23:07:06 CEST] <kierank> how do I connect to the channel
[23:08:06 CEST] <wm4> GabrielYYZ: nice, a merge commit that basically touches everything
[23:09:35 CEST] <GabrielYYZ> wm4, i think (i'm guessing here) including internal.h in bitstream.c is the actual problem
[23:10:08 CEST] <GabrielYYZ> but, again, that's just me guessing
[23:12:40 CEST] <GabrielYYZ> the other alternative is qdm2_tablegen.c #defining hardcoded_tables to 0 unconditionally
[23:16:08 CEST] <kierank> reynaldo: ping
[23:16:32 CEST] <reynaldo> kierank: pong, -> ffmpeg-opw
[23:16:35 CEST] <kierank> how do I join
[23:16:41 CEST] <reynaldo> you are there already
[23:16:53 CEST] <kierank> my client just says "connecting"
[23:17:04 CEST] <reynaldo> switch channel window in your IRC client I guess
[23:27:37 CEST] <kierank> lglinskih: do you understand the response I emailed
[23:31:26 CEST] <lglinskih> kierank: =)) yes, now it's absolutely clear
[23:31:33 CEST] <kierank> ok :)
[23:59:24 CEST] <cone-874> ffmpeg 03Chris Watkins 07master:0703c44692d1: avformat/mov.c: Use %d to print an enum, not PRIu8
[00:00:00 CEST] --- Wed Apr 22 2015