[Ffmpeg-devel-irc] ffmpeg-devel.log.20150115

[burek]

[00:20] <iive> won't runtime lib loading be used as a workaround for non-free ffmpeg distributable.
[00:21] <iive> when external library have conflicting license.
[00:22] <iive> this way, ffmpeg and the library are distributed separately and same ffmpeg binary can work with and without the troubled library.
[00:42] <Daemon404> j-b, why did someone commit opus in mp4 to vlc
[00:42] <Daemon404> it is not a final spec.
[02:03] <compn> Daemon404 : because fuck specs!
[02:05] <kierank> well yeah look at libavformat and the mpegts
[02:05] <kierank> who cares if there isn't pmt
[02:06] <kierank> let's just guess the streams
[02:06] <compn> lets instead just reject 50% of mpegts because the spec says so ?
[02:12] <kierank> lol 50%
[02:12] <kierank> more like 0.0000000001%
[02:12] <kierank> and break valid streams for no reason by guessing ghost pids
[02:31] <compn> kierank : why not submit patch to make -ghostpid enable that feature and disable it the rest of the time ?
[02:36] <cone-975> ffmpeg.git 03Vittorio Giovara 07master:456ffec35d55: img2dec: correctly use the parsed value from -start_number
[02:36] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:140dfd14fdab: Merge commit '456ffec35d55288d589aa1901919e038c3427f30'
[02:39] <michaelni> kierank, which ticket is that where a valid stream (with pmt id assume from "valid") is broken ?
[02:40] <kierank> will have to look
[02:40] <kierank> but iirc there is a ticket with a ghost pid that is probed and ends up as mp3 as usual
[02:41] <michaelni> if you point me to it i can take a look, btw whats a ghost pid ?
[02:42] <kierank> a pid that is in the stream but not referenced in the pmt
[02:42] <kierank> i.e one the decoder is meant to ignore
[02:44] <michaelni> where does the spec say the decoder MUST ignore such streams ?
[02:44] <kierank> well they are not part of the program
[02:45] <michaelni> yes, they should be ignored if the user wants just that program
[02:45] <kierank> but you don't know what the data is
[02:45] <kierank> it could be anything 
[02:45] <kierank> but if it's not in the pmt you don't know
[02:46] <michaelni> i also know this IRC log is english even though it doesnt say it is english
[02:46] <kierank> so you should probe random data and guess incorrectly as mp3
[02:46] <kierank> because you don't know what it is
[02:46] <michaelni> no
[02:49] <michaelni> ffmpeg should ideally detect it as unknown data
[03:01] <cone-975> ffmpeg.git 03Vittorio Giovara 07master:1e763454322f: png: improve signature check
[03:02] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:cd960c8a4f7c: Merge commit '1e763454322f7fbc7799f6009bf2e11d7a3b9821'
[04:04] <j-b> Daemon404: why not?
[04:34] <cone-975> ffmpeg.git 03Vittorio Giovara 07master:cf70ba37ba74: mov: Check angle rather than full matrix when updating SAR
[04:34] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:89b6382bec08: Merge commit 'cf70ba37ba74089a18295b29e77dead0a3222c9e'
[04:34] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:47795505acea: avformat/mov: rewrite aspect from  matrix code
[04:34] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:cc18ea8de058: avformat/mov: Do not set aspect ratios that would contain zeros
[05:14] <cone-975> ffmpeg.git 03Stefano Sabatini 07master:e2ad0b66fa27: imgutils: create misc functions for dealing with buffers
[05:14] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:55dc253c24f2: Merge commit 'e2ad0b66fa273c5c823978e8f601f2c0d9ee42f8'
[05:34] <cone-975> ffmpeg.git 03Paul B Mahol 07master:0e6c7dfa650e: lavu: add planar 16 bit and 8 bit rgb with alpha
[05:34] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:ada0fc4efa25: Merge commit '0e6c7dfa650e8b0497bfa7a06394b7a462ddc33a'
[05:53] <cone-975> ffmpeg.git 03Paul B Mahol 07master:9a4aaae3b293: lavc: add GBRAP to avcodec_align_dimensions2
[05:53] <cone-975> ffmpeg.git 03Paul B Mahol 07master:928061670e87: libswscale: GBRAP input & output and GBRAP16 input support
[05:53] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:5edbd4f4106f: Merge commit '9a4aaae3b29392b6d74997e06be249d7d54a7394'
[05:53] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:27513846f66b: Merge commit '928061670e873e816daa14827853b7e11221ff5f'
[06:29] <cone-975> ffmpeg.git 03Paul B Mahol 07master:f86f39cb9b1f: tiff: support decoding GBRP and GBRAP formats
[06:29] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:3779973a9b8e: Merge commit 'f86f39cb9b1fcd063d5e4812132a75c06cc7acd2'
[10:17] <__gb__> hi compn, I am not going to comment on that right now
[10:20] <__gb__> compn, for testing hwaccel, I did something here: https://github.com/gbeauchesne/mvt_tools (granted, vaapi centric atm)
[10:21] <__gb__> FATE could work as well, there is only one point of possible failure I am aware of for h264 conformance tests
[10:25] <ubitux> michaelni: i'd like to improve pal8 support; i don't understand why sws says it doesn't support pal8 as output, it seems to though (what happens when you request pal8 as output?)
[10:28] <nevcairiel> using fate for hwaccel testing works quite well, just have to skip the cropping test. the only problem is that most/many fate boxes are headless systems or VMs, and hwaccel doesnt necessarily work on those. I know my fate Windows  VM doesn't do hwaccel for testing.
[10:28] <nevcairiel> .. so i only test it manually when i do changes
[10:28] <ubitux> mmh we don't have a apng muxer, too bad
[10:31] <wm4> ubitux: to handle pal8, you need to decide on a palette
[10:31] <ubitux> i'd like to start with a palette per frame
[10:31] <wm4> ubitux: so you'd have to provide sws one or so
[10:32] <ubitux> mmh?
[10:32] <ubitux> i mean defining a palette within swscale for a given frame
[10:33] <ubitux> (i can use elbg for that, no?)
[10:33] <wm4> I suppose palette per frame is much simpler
[10:34] <ubitux> maybe a video filter would be more appropriate
[10:36] <__gb__> nevcairiel, FATE/hwaccel over Xen, KVM or VMware could indeed be good tests for vGPU engines in there
[10:37] <__gb__> as to headless systems, this could work just fine. MVT runs in headless mode actually (VA/DRM)
[10:38] <__gb__> nevcairiel, including, windows over XenGT could be worth testing if you have some time to spare
[10:38] <wm4> __gb__: so what's your opinion about libyami? should we be using it instead?
[10:45] <nevcairiel> __gb__: the only problem with that would be that i cannot be sure if i'm testing our code or the VMs emulation capabilities.... :)
[10:49] <__gb__> from older discussions with original xen-gt architect a couple of years ago, there were indeed issues, but I believe they have been addressed since the public release :)
[10:50] <__gb__> hence, "worth testing if you have time" but don't get too upset if this fails right now
[11:24] <wm4> no opinion apparently
[11:40] <cone-975> ffmpeg.git 03Clément BSsch 07master:3b3229aef322: avfilter/edgedetect: fix indent
[12:04] <ac_slater> hey all. I can't really seem to figure out how to cross compile ffmpeg for 32bit from 64bit linux
[12:04] <ac_slater> the --host{cflags,ldflags}=-m32 aren't doin it
[12:05] <nevcairiel> you dont want --host, you want --extra-cflags
[12:06] <ac_slater> nevcairiel: damn! Thanks!
[12:06] <ac_slater> nevcairiel: I meant to ask this in the users channel, my mistake
[12:43] <michaelni> ubitux, the "non support" was probably meant as not supporting finding and using arbitrary palettes, sws supports multiple fixed palette 8bit output formats
[12:50] <ubitux> ok
[12:52] <cone-975> ffmpeg.git 03Anton Khirnov 07master:8809c974a3fb: id3v2: constify the 'tag' parameter to special metadata parsing callback
[12:52] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:06ec2f1d3218: Merge commit '8809c974a3fb51f96e498a5556a4a5bbacc581ce'
[12:52] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:7eef45a76f82: avformat/id3v2: Fix incompatible pointer type warnings
[13:10] <cone-975> ffmpeg.git 03Anton Khirnov 07master:932788be5af8: id3v2: add names to the parameters of ID3v2EMFunc.read
[13:10] <cone-975> ffmpeg.git 03Anton Khirnov 07master:54bc15d5ebfd: id3v2: fix reading v2.2 attached pictures
[13:10] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:24c76bae5f47: Merge commit '932788be5af8dee062c77851b573ea47dd6d047a'
[13:10] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:eb215be9ce74: Merge commit '54bc15d5ebfd07fd468743ba29f709ea19e840b9'
[13:37] <akira4> ubitux, the link to the complete repo : https://github.com/akira4/FFmpeg. Is this better?
[13:43] <ubitux> yes!
[13:43] <ubitux> you probably want to work in a branch ` master
[13:43] <ubitux> but yeah, will do.
[13:45] <cone-975> ffmpeg.git 03Vittorio Giovara 07master:014b6b416fec: vp8: improve memory allocation checks
[13:46] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:9fb5a91c6693: Merge commit '014b6b416fec89777cb9cff61bcf7896eaf7cf39'
[13:54] <cone-975> ffmpeg.git 03Vittorio Giovara 07master:8a9641a652ed: bsf: check memory allocations
[13:54] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:c5eb725fd33e: Merge commit '8a9641a652ed1546fedfda22584f79d3d423096e'
[14:06] <cone-975> ffmpeg.git 03Vittorio Giovara 07master:402fb5550e36: opus_parser: drop unused variable
[14:06] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:2bc841815c9c: Merge commit '402fb5550e36dd994b13941ef5499f9087afd345'
[14:20] <Daemon404> j-b, mostly because it is a) not implemented properly b) might encourage use of an unfinished spec
[14:27] <ubitux> damn, our gif decoder doesn't output in pal8 
[14:27] <Daemon404> i always thought pal8 was stupid as a "Colorspace"
[14:28] <ubitux> i'm not sure to see the problem
[14:31] <Daemon404> its not a colorspace IMO
[14:31] <Daemon404> it should just be rgb.
[14:31] <Daemon404> (the kludges in swscale for pal is ... meh)
[14:33] <ubitux> and every time you need to transmit the palette you recompute it from rgb32?
[14:35] <Daemon404> it'd probably be better as metadata
[14:35] <Daemon404> of course it was hammered to fit as a "Colorspace" instead
[14:36] <Daemon404> (it is also utterly trivial to computer an existing palette as well)
[14:40] <ubitux> sounds simpler to just have it in a plane
[14:41] <ubitux> i don't see the benefit of using a... string to transmit that
[14:41] <Daemon404> i dont count decoding to a palette as actually decoding
[14:41] <Daemon404> since you still have to decode it to use it in any reasonable way
[14:41] <Daemon404> (also, who said string?)
[14:43] <ubitux> metadata are string
[14:43] <ubitux> or maybe you thought "side data" or something
[14:43] <ubitux> but there is really no benefit for that indirection
[14:43] <ubitux> it won't solve anything to have it stored differently
[14:44] <Daemon404> not in the current libav* molds, no
[14:44] <Daemon404> because when all you have are circle holes, every peg will be made into a circle
[14:44] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:def69553e745: h264_cabac: Break infinite loops
[14:44] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:330e892c083e: Merge commit 'def69553e745e007925147b2e7e4b32c57d08fdf'
[14:46] <wm4> Daemon404: my favorite is that formats other than pal8 can require a palette
[14:46] <wm4> even if the palette is not used
[14:46] <wm4> (wat)
[14:46] <nevcairiel> how is it required if its not used
[14:47] <Daemon404> well im currently trialing !swscale for some use cases
[14:47] <Daemon404> ;)
[14:47] <wm4> nevcairiel: it will crash when... copying the AVFrame
[14:48] <wm4> simply because the pixfmts in questions are declared as having a palette
[15:03] <ubitux> (wm4: fyi ./ffmpeg -i ~/fate-samples/fli/fli-engines.fli -ss 3 -vf palettedbg -frames:v 1 -y pal.png ’ http://b.pkh.me/pal.png)
[15:03] <wm4> ah
[15:04] <ubitux> what does pseudopal actually means?
[15:04] <ubitux> i'm disappointed we removed pal8 from the gif decoder :'(
[15:04] <ubitux> but i guess that's because of transparency
[15:07] <cone-975> ffmpeg.git 03Martin Storsjö 07master:024e5a2d5ff8: rtmppkt: Repeat the full 32 bit timestamp for chunking continuation packets
[15:07] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:6e69bf4a7eb5: Merge commit '024e5a2d5ff8a94adce48abb15ce2fb471f9d18e'
[15:08] <nevcairiel>  * The pixel format is "pseudo-paletted". This means that FFmpeg treats it as
[15:08] <nevcairiel>  * paletted internally, but the palette is generated by the decoder and is not
[15:08] <nevcairiel>  * stored in the file.
[15:08] <nevcairiel> there ^
[15:09] <nevcairiel> not sure if that actually applies to any codec
[15:09] <nevcairiel> apparently some combination of raw files
[15:09] <wm4> it's also not sure what the heck it means
[15:10] <Daemon404> i dont understand its point
[15:10] Action: ubitux neither
[15:10] <iive> probably Y8
[15:11] <iive> instead of been treated as YUV with a single plane/channel, it is treated as Pal8 where each index is a single gray gradation.
[15:12] <wm4> iive: there are other pseudopal formats
[15:13] <wm4> like rgb8
[15:13] <wm4> or bgr4_byte (why does this even exist)
[15:13] <iive> cga?
[15:16] <wm4> what?
[15:16] <wm4> that thing from 50 years ago?
[15:18] <iive> yes, 34 years old.
[15:22] <Mavrik> hmm, any pointers on how to blend a progressive AVFrame onto an interlaced AVFrame?
[15:23] <nevcairiel> if your progressive image isnt moving, then just do it, if it is moving, interlace it first? :D
[15:25] <Mavrik> basically I'm doing a filter and even though my progressive image isn't moving I'm seeing it "dance around" on interlaced content :)
[15:25] <Mavrik> Hrmf, I should look at the libass filter
[15:26] <wm4> so you want to render subtitles on an interlaced frame?
[15:27] <Mavrik> Yes, I'm doing my own custom filter to grab subtitle images (which I have in AVFrame) to the video
[15:27] <Mavrik> currently I'm using some old blending code from 1.0 days I guess, which I should probably replace
[15:28] <wm4> I doubt the libass filter does anything special for interlaced, but I don't know
[15:29] <nevcairiel> a static/not-moving interlaced image should be the same as a progressive image, interlaced only gets special when there is changes between two fields (ie. the image moves), unless i'm missing something right now
[15:31] <Mavrik> it's entirely possible I'm messing something else now, I was just more wondering if libavfilter or libswscale already have AVFrame blending code somewhere implemented
[15:31] <Mavrik> (this is my broken result: https://dl.dropboxusercontent.com/u/2425715/interl.png )
[15:32] <nevcairiel> those subtitle artifacts dont like what I expected
[15:35] <Mavrik> the other option is that I'm messing something up when decoding them (they're coming from DVB and I'm porting old code to 2.5)
[15:35] <Mavrik> anyway, I'll look into it, thanks anyway
[16:35] <cone-975> ffmpeg.git 03agathah 07master:72c61c272c52: avcodec/nvenc: fix b frame settings
[16:36] <cone-975> ffmpeg.git 03Clément BSsch 07master:f5326039938c: avfilter: add showpalette filter
[16:40] <cone-975> ffmpeg.git 03Clément BSsch 07master:a83398c8d2c9: doc: fix showpalette position
[18:02] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:b0be6d8385c5: avformat/id3v2: remove unused variable
[18:02] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:46b0fce627d6: avutil/pixfmt: Document the duplicated AV_PIX_FMT_GBRAP* formats
[18:02] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:ec9935b288af: avutil/pixfmt: Support the AV_PIX_FMT_GBRAP* formats with AV_HAVE_INCOMPATIBLE_LIBAV_ABI
[18:03] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:5f964c6027da: avcodec/flac_parser: Use sizeof(variable) instead of sizeof(type)
[18:03] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:6c3cb02a742f: avcodec/flac_parser: Check for av_malloc() failure
[18:09] <wm4> michaelni: do you know a single user of AV_HAVE_INCOMPATIBLE_LIBAV_ABI?
[18:11] <michaelni> no, i was thinking about removing it already
[18:12] <wm4> that would be a nice cleanup
[18:29] <ubitux> dropping abi compatibility?
[18:29] <ubitux> so no more tags etc for codec ids?
[18:31] <wm4> there are two parts to this
[18:31] <wm4> 1. not breaking ffmpeg's ABI if libav adds something
[18:31] <wm4> and 2. having a perfectly Libav-compatible ABI
[18:31] <wm4> this is about 2.
[18:31] <wm4> for 1. you still need stupid hacks
[18:33] <ubitux> mmh yeah indeed my bad
[18:56] <nevcairiel> i dont really view the the fourcc codec id tags as a problem or anything
[21:29] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:e7d85186d639: avcodec/parser: Check that the parser return code is valid
[23:43] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:71bd6ea0fada: avcodec/flashsv2enc: use av_realloc_array()
[23:43] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:5989a2b24f36: avcodec/g722enc: Use av_mallocz_array()
[23:43] <cone-975> ffmpeg.git 03Michael Niedermayer 07master:2c22042aa072: avcodec/g2meet: Use av_mallocz_array() for framebuffer allocation
[00:00] --- Fri Jan 16 2015