[Ffmpeg-devel-irc] ffmpeg-devel.log.20160119

[burek]

[00:17:41 CET] <cone-294> ffmpeg 03James Almer 07master:c1a3683f26fb: configure: disable -Wunused-const-variable
[05:32:31 CET] <Lectem> Hi guys, I spotted a problem with Arm platforms that do not have VFP, and I'm not sure how to approach the issue. In https://www.ffmpeg.org/doxygen/2.8/fft__init__arm_8c_source.html#l00044 , ff_fft_calc_vfp is used without checking HAVE_VFP. The library builds fine, but when trying to link it in another program we get an undefined reference
[05:33:10 CET] <Lectem> would the fix only be to add #if HAVE_VFP ?
[05:33:59 CET] <Lectem> (it compiles fine with it, but I'm not sure if it breaks something or not)
[05:34:46 CET] <jamrial> that's weird, the have_vfp() check should be enough
[05:35:21 CET] <Lectem> at runtime yes, not when linking
[05:36:06 CET] <Lectem> ffmpeg/libavcodec/arm/fft_init_arm.c:46: undefined reference to `ff_fft_calc_vfp'
[05:37:43 CET] <jamrial> dead code elimination should take care of that, ah well
[05:38:12 CET] <jamrial> that fix is probably ok. send a patch to ffmpeg-devel so the arm maintainers can review it
[05:51:51 CET] <Lectem> ok, I wish ffmpeg would switch to a decent PR system, mailing lists are such a pain and flooding the inbox <.<
[06:05:38 CET] <jamrial> Lectem: well, for bug reports you have http://trac.ffmpeg.org/
[06:05:58 CET] <jamrial> but patches will still get the most attention if sent to the ml
[06:14:34 CET] <Lectem> well yeah I understand that, but its really not user friendly when people only want to make small patches, you receive mails for every single patch, and mailman still stores passwords in plaintext
[06:17:31 CET] <Lectem> (and since there's no real tracking, I have the feeling many patches are losts, such as https://ffmpeg.org/pipermail/ffmpeg-devel/2015-December/184675.html
[06:31:37 CET] <jamrial> you can send patches without subscribing to the ml. they just need to be approved before they make it through
[06:32:36 CET] <jamrial> you'll most likely not get replies, though, since they will be sent to the ml as well
[06:36:48 CET] <rcombs> I have a gmail rule that archives ML posts that don't mention me by name
[06:37:03 CET] <rcombs> which is _usually_ a pretty good heuristic for only getting replies to my own patches
[08:41:10 CET] <cone-987> ffmpeg 03Luca Barbato 07master:7ebf1b4adad4: avconv: Simplify poll_filters() return value check
[08:41:10 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:4440c8e9c98f: Merge commit '7ebf1b4adad40e2b0d88ebd5cfde821a7a0ceab2'
[08:42:45 CET] <cone-987> ffmpeg 03Luca Barbato 07master:63f7f8d1dbf6: avconv: Drop an impossible check
[08:42:46 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:7dae3461c0bb: Merge commit '63f7f8d1dbf6ce27440fdd56ef38c822fb11b9c2'
[08:43:50 CET] <cone-987> ffmpeg 03Maxym Dmytrychenko 07master:a6259a6ecbad: qsvenc: properly handle the warning from MFXVideoCORE_SyncOperation
[08:43:51 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:b5513441e477: Merge commit 'a6259a6ecbad18d58085e4ab23f56151e98d7b53'
[08:45:12 CET] <cone-987> ffmpeg 03Anton Khirnov 07master:d1cd20e4e33f: qsvenc: split encoding frames and reading from the async FIFO
[08:45:13 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:ed9cff9edf0f: Merge commit 'd1cd20e4e33f8139e150034b3c457302312d81bd'
[08:45:47 CET] <cone-987> ffmpeg 03Andreas Cadhalpun 07master:c69461d73797: asfdec: only set asf_pkt->data_size after sanity checks
[08:45:48 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:1ac8a1821abd: Merge commit 'c69461d73797e02e7a3ab4316050c241fa91f53f'
[08:45:58 CET] <cone-987> ffmpeg 03Andreas Cadhalpun 07master:2aec600ae7af: asfdec: reject size > INT64_MAX in asf_read_unknown
[08:45:59 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:202ed3e5e674: Merge commit '2aec600ae7af7d46a4877b5ccc263b39f05a91cb'
[08:48:02 CET] <cone-987> ffmpeg 03Alexandra Hájková 07master:066281372d90: asfdec: Remove unused function parameters
[08:48:03 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:be3963c8af6e: Merge commit '066281372d90d63ca021b659abcb8faefd6bc4a6'
[08:48:14 CET] <cone-987> ffmpeg 03Andreas Cadhalpun 07master:fa463aa83a49: avpacket: fix size check in packet_alloc
[08:48:15 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:774c5357dbe6: Merge commit 'fa463aa83a4920b0eed47ad1f79775dfc53d21ec'
[08:50:13 CET] <cone-987> ffmpeg 03Martin Storsjö 07master:26ec75aec357: checkasm: Check register clobbering on arm
[08:50:14 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:0b40e290e3cc: Merge commit '26ec75aec3576daea691dee53a78ec67c0dc4040'
[08:51:00 CET] <cone-987> ffmpeg 03Martin Storsjö 07master:fec76cd430f3: checkasm: Check register clobbering on aarch64
[08:51:00 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:7e2990352689: Merge commit 'fec76cd430f3c865183a6e5b4caec0743e055605'
[08:51:30 CET] <cone-987> ffmpeg 03Martin Storsjö 07master:73c8c0341cce: arm: Fix vfp dead code elimination with have_vfp_vm
[08:51:31 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:a04a9434b5ec: Merge commit '73c8c0341cce9e1a6c4169721f5123f97fc4be2f'
[08:53:57 CET] <cone-987> ffmpeg 03Luca Barbato 07master:40d43d25e76c: APIchanges: Add missing av_pix_fmt_get_chroma_sub_sample entry
[08:53:58 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:fea9243d53e0: Merge commit '40d43d25e76ca078e7665752e815fc8d96252f06'
[08:54:08 CET] <cone-987> ffmpeg 03Alexandra Hájková 07master:34d45b36164e: asfdec: Fix typo in comment
[08:54:09 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:25955c927fcf: Merge commit '34d45b36164eee044cfe55d22488b2b65e387872'
[08:59:29 CET] <cone-987> ffmpeg 03Clément BSsch 07master:7570c9e04f01: swfdec: support compressed swf
[08:59:29 CET] <cone-987> ffmpeg 03Hendrik Leppkes 07master:e816fe7401c9: Merge commit '7570c9e04f010c9b3bfdeb4338d330f2cdd25278'
[10:08:40 CET] <durandal_1707> what's point of codecpar?
[10:20:13 CET] <wm4> durandal_1707: untangling lavf and lavc
[11:03:51 CET] <cone-987> ffmpeg 03Carl Eugen Hoyos 07master:9cf8c3e62cf0: ape: Show more information for loglevel verbose.
[11:21:31 CET] <cone-987> ffmpeg 03Mats Peterson 07master:d6b3062b5a6a: lavf/qtpalette: Fix incorrect palettes
[13:20:43 CET] <wm4> for the mess that vaapi is, those h264 and h265 encoder patches look pretty nice
[13:31:48 CET] <jkqxz> Thanks for having a look.
[13:34:49 CET] <jkqxz> For H.264, baseline is not a strict subset of main.  It has a few extra features for real-time stuff which AFAIK noone actually uses.
[13:35:19 CET] <wm4> I know
[13:35:46 CET] <wm4> and AFAIK earlier versions of libva didn't report constrained baseline (which is a strict subset of main), so there's a slight mess
[13:36:01 CET] <wm4> (or was that vdpau only?)
[13:37:40 CET] <jkqxz> Things I've seen do the right thing there.  (Report constrained baseline but not baseline.)
[13:39:00 CET] <jkqxz> But I haven't looked back at old versions.  I'm going to make the configure test force you to be relatively up to date because old versions will just be horrible to support.
[13:43:00 CET] <durandal_1707> what happened to ubitux?
[13:47:14 CET] <Daemon404> srt killed him
[14:06:13 CET] <Compn> [06:20] <ubitux> "yes" is not a valid answer here :p
[14:06:16 CET] <Compn> was his last message
[14:06:28 CET] <Compn> rip
[14:07:22 CET] <J_Darnley> RIP IN PEACE UBITUX ??-06:20
[14:15:46 CET] <cone-987> ffmpeg 03Michael Niedermayer 07master:22f64c23244f: avcodec/ccaption_dec: Fix mixed declarations and code
[16:09:38 CET] <J_Darnley> Thanks Thundernird!  Did you send that email before you crashed?
[16:13:01 CET] <durandal_1707> nope
[16:16:16 CET] <J_Darnley> Yeah!  You did it!  Good job!
[16:26:15 CET] <kierank> who is going to fosdem?
[16:27:19 CET] <Rodeo> not me, apparently :(
[16:27:25 CET] <Rodeo> FWIW
[16:29:18 CET] <durandal_1707> who is going to me?
[16:31:48 CET] <Rodeo> ???
[16:33:01 CET] <durandal_1707> one ? is enough
[16:33:06 CET] <kierank> durandal_1707: y u no go to conf?
[16:33:12 CET] <kierank> you are one of the few devs I have not met
[16:34:05 CET] <durandal_1707> because I eat small people for breakfast
[16:34:39 CET] <kierank> atomnuker would like to meet you for sure
[16:34:57 CET] <atomnuker> all hail our libavfilter overlord
[16:38:02 CET] <durandal_1707> I'm busy writting 300th filter
[16:39:06 CET] <durandal_1707> there are 4 filters I havent committed
[16:39:18 CET] <J_Darnley> I'll probably turn up.
[16:39:31 CET] <J_Darnley> It's not like I do anything more interesting with my life.
[16:40:53 CET] <durandal_1707> when and where is fosdem?
[16:43:20 CET] <durandal_1707> ah, its timing bad and too far away
[16:44:17 CET] <durandal_1707> Just put it in Italy  for example and I will come
[16:44:38 CET] <cone-987> ffmpeg 03Neil Birkbeck 07master:d91718107c33: libavutil: add mastering display metadata sidedata
[16:58:02 CET] <jamrial> michaelni: where is the fixed_dsp checkasm patch failing to build for you? it works for me linux x86_64 and mingw
[17:03:26 CET] <michaelni> tests/checkasm/fixed_dsp.c:53 <--
[17:03:41 CET] <michaelni> jamrial, gcc (Ubuntu/Linaro 4.6.3-1ubuntu5) 4.6.3
[17:04:03 CET] <jamrial> so, old gcc then...
[17:04:25 CET] <jamrial> ok, thanks
[17:35:28 CET] <cone-987> ffmpeg 03Michael Niedermayer 07master:b0c57206d583: avformat/hls: Require the file extension to be m3u / m3u8 for probing to succeed
[19:59:11 CET] <durandal_1707> can I get comments for streamselect filter?
[20:19:54 CET] <Savago> Good morning. I got one question related to using libav* libraries in a C app. I want to cut some parts of a video (similar to what ffmpeg allows to do with -ss xx:xx -t duration). My understanding is that the steps would be roughly: a) init ffmpeg decoders b) open file/search streams c) open decoder d) read the frames c) encode an output video?
[20:50:13 CET] <an3k> nevcairiel ... I knew I know your nick from somewhere ... doom9? :)
[20:52:16 CET] <wm4> yes he's there
[20:52:43 CET] <wm4> is there a precedent of ffmpeg APIs which are only available on specific platforms?
[20:54:02 CET] <wm4> libavcodec/vdpau.h is probably not always available, but for some reason libavcodec/d3d11va.h always provides the function it declares
[21:01:05 CET] <nevcairiel> public api should always be available, even if it  only return NULL or ENOSYS
[21:02:19 CET] <wm4> I doubt that's currently the case with at least vdpau
[21:03:32 CET] <nevcairiel> the main reason for that is that we dont necessarily have "platform" features, but just features, so if i build ffmpeg without vdpau, would the API suddenly vanish? that sounds like ABI trouble
[21:08:59 CET] <cone-120> ffmpeg 03Paul B Mahol 07master:307b84821816: avfilter/af_dynaudnorm: call uninit() from config_input()
[21:42:32 CET] <wm4> J_Darnley: is there no good mail client? currently I'm waiting for claws-mail to get out of its fucking state and send my mail already
[21:42:46 CET] <wm4> I think it got into some network deadlock...
[21:43:11 CET] <wm4> its network code must be hilariously bad
[21:56:38 CET] <ubitux> hi, i'm back from the deads
[21:57:02 CET] <durandal_1707> :o
[21:57:09 CET] <ubitux> durandal_1707: nlmeans ETA unknown but i do not forget
[21:57:43 CET] <ubitux> wm4: the user is likely not to be supposed to free subtitle_header (i'm assuming this based on the use in vf_subtitles :p)
[21:57:58 CET] <ubitux> durandal_1707: nice for streamselect, i'll have a look
[21:58:30 CET] <wm4> ubitux: dunno, but avcodec_free_context frees it
[21:59:02 CET] <ubitux> is it a problem? (i don't remember it's actual use over extradata)
[21:59:35 CET] <ubitux> sorry for the abscence btw.
[21:59:45 CET] <wm4> huh, does vf_subtitles leak the codec context?
[22:00:01 CET] <wm4> ah no
[22:00:10 CET] <wm4> it reuses the AVStream context
[22:02:00 CET] <ubitux> durandal_1707: did you rewrite streamselect?
[22:02:12 CET] <J_Darnley> wm4: no there probably isn't a good one.
[22:03:07 CET] <wm4> sad
[22:06:27 CET] <durandal_1707> ubitux: no, added more code
[22:06:46 CET] <ubitux> i wouldn't mind being in signed off then ;)
[22:07:15 CET] <ubitux> durandal_1707: so, where is the trick?
[22:11:23 CET] <durandal_1707> ubitux: used request_frame from framesync
[22:11:59 CET] <durandal_1707> I don't get vava:va syntax
[22:12:01 CET] <ubitux> oh, ok
[22:12:37 CET] <durandal_1707> and sendcmd works fine with vv:v 
[22:13:14 CET] <ubitux> >  ffmpeg -i ... -i ... -lavfi "[0:0][0:1][1:0][1:1]streamselect=in=vava:out=va:map=0 1" 
[22:13:22 CET] <ubitux> > then send commands such as "map 0 1" or "map 2 3" to select one or another source.
[22:13:25 CET] <ubitux> you don't understand this?
[22:14:10 CET] <ubitux> let's say you have 2 live streams with an audio and video stream in each
[22:14:19 CET] <durandal_1707> no, need for one big multimedia filter instead of two smaller ones
[22:14:25 CET] <wm4> what happens to the selected sources?
[22:14:27 CET] <wm4> err
[22:14:29 CET] <wm4> what happens to the unselected sources?
[22:14:36 CET] <ubitux> wm4: it should be consumed
[22:14:36 CET] <wm4> do they just discard input?
[22:14:41 CET] <ubitux> it should
[22:14:55 CET] <ubitux> durandal_1707: yeah i guess it could be split
[22:15:01 CET] <wm4> doesn't that mean if the input is sourced from an independent file, it will cause that file to be read completely until EOF is reached?
[22:15:32 CET] <ubitux> it's timed based, it's supposed to drop in //
[22:15:58 CET] <ubitux> like, if you pick the stream 0 from t=[0..5] it will drop 5sec of the 2nd stream
[22:17:02 CET] <durandal_1707> so just many to 1 V filter, instead of many to less many
[22:17:58 CET] <ubitux> yeah that's fine with me i guess
[22:18:39 CET] <durandal_1707> imagine 2x2 videos stacked together and you switch streams with streamselect for each input
[22:20:09 CET] <durandal_1707> actually M to N could also work, need single command instead of 4
[22:20:47 CET] <durandal_1707> but sendcmd doesn't like space in command args 
[22:21:14 CET] <durandal_1707> so map 0 1 command doesn't work 
[22:28:55 CET] <wm4> hm what do I do in cases where avcodec_decode_audio4() returns 0 forever?
[22:29:29 CET] <wm4> note that some decoders actually return 0 a few times for the first packet, and then start working
[22:29:58 CET] <wm4> but I also have a sample that returns 0 forever for the first packet
[22:32:23 CET] <Compn> upload sample, report bug and wait for carl to close it of course :P
[22:32:28 CET] Action: Compn jokes and runs
[22:32:35 CET] <wm4> the file is from the sample server
[22:35:35 CET] <Lectem> fincs: BPB uses the following nngxCmdlistStorage(0x80000u,0x200u)
[22:35:59 CET] <Lectem> woops, wrong channel sorry
[22:36:39 CET] <Compn> wm4 : forever as in loops or just for the entirety of the file ?
[22:36:54 CET] <durandal_1707> wm4: take another packet
[22:36:57 CET] <wm4> Compn: as I said
[22:37:00 CET] <wm4> durandal_1707: when?
[22:37:37 CET] <durandal_1707> after 17
[22:37:43 CET] <wm4> why 17?
[22:38:12 CET] <durandal_1707> reasonable small
[22:38:46 CET] <durandal_1707> you can pick smaller
[22:39:09 CET] <ubitux> durandal_1707: designing a filter because of a but in another is not really good design :p
[22:39:25 CET] <ubitux> (not saying there aren't good reasons to design as such)
[22:40:42 CET] <durandal_1707> ubitux: it can do many - many just sendcmd need fix
[22:42:19 CET] <durandal_1707> just doing A and V at same time seems impossible
[22:43:10 CET] <durandal_1707> unless I'm missing something obvious
[22:52:54 CET] <durandal_1707> There is bsd licensed opencl nlmeans filter
[23:11:29 CET] <kierank> lotharkript: hello
[23:19:18 CET] <lotharkript_> kierank: hello
[23:21:57 CET] <kierank> I should probably send a new cineform patch
[23:29:54 CET] <cone-120> ffmpeg 03Vittorio Gambaletta (VittGam) 07master:cde4b99f4768: ffplay: seek only when pressing the right mouse button on the video window
[23:29:55 CET] <cone-120> ffmpeg 03Vittorio Gambaletta (VittGam) 07master:acd37069aec2: ffplay: toggle full screen when double-clicking the video window with the left mouse button
[23:32:27 CET] <cone-120> ffmpeg 03Vittorio Giovara 07master:ff8c2c410302: yuv2rgb: Document the color space coefficients
[23:32:28 CET] <cone-120> ffmpeg 03Bela Bodecs 07master:cc83177db4e0: vf_overlay: handles expression evaluation of frame size change in frame-by-frame evalutaion mode
[00:00:00 CET] --- Wed Jan 20 2016