[Ffmpeg-devel-irc] ffmpeg-devel.log.20131221

burek burek021 at gmail.com
Sun Dec 22 02:05:03 CET 2013

[00:05] <cone-64> ffmpeg.git 03Paul B Mahol 07master:e518cb863edc: jvdec: Fix memory leak of jv->frames
[00:05] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:b9f33ad32077: Merge commit 'e518cb863edc931888ccca6cad86f73ca7430cef'
[00:13] <cone-64> ffmpeg.git 03Luca Barbato 07master:15739a9bd19a: jvdec: Do not feed the decoder with known wrong data
[00:13] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:0c16f6449bb5: Merge commit '15739a9bd19a7d47ad8afb25348c684a3bdd6ef2'
[00:20] <cone-64> ffmpeg.git 03Luca Barbato 07master:f412b2c9f3a7: flv: Use the correct type to hold the file offset
[00:20] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:6fc9f5e0ff77: Merge commit 'f412b2c9f3a7add0ab8021262ec4bad249347e30'
[00:31] <cone-64> ffmpeg.git 03Gildas Cocherel 07master:2a41826bea38: lavc: add HEVC profiles names
[00:31] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:d6b7ea06f194: Merge commit '2a41826bea3833895dc06939831b7f35ca1f597e'
[00:33] <cone-64> ffmpeg.git 03Gildas Cocherel 07master:321cb8b048c7: hevc: store profile and level in AVCodecContext
[00:33] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:2fb9e77ac2da: Merge commit '321cb8b048c7a8c0d2950f9278f36d912a886fa0'
[00:51] <cone-64> ffmpeg.git 03Guillaume Martres 07master:67bbaed5c498: hevc: don't check for errors in PTL code
[00:51] <cone-64> ffmpeg.git 03Gildas Cocherel 07master:2d18aaa14bc2: hevc: refactor Profile Tier Level
[00:51] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:e00281b6f507: Merge commit '2d18aaa14bc2233b0647d2c5abb72e0bc828c2a8'
[00:53] <cone-64> ffmpeg.git 03Gildas Cocherel 07master:259cddb13798: hevc_ps: fix indentation
[00:54] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:3cd36b07dd67: Merge commit '259cddb137988a25607f091c2414158ff81ca7d6'
[01:02] <cone-64> ffmpeg.git 03Guillaume Martres 07master:78d0b7f36e39: hevc: warn when an unknown profile is used
[01:02] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:9567162237d2: Merge commit '78d0b7f36e39bb4745613e312024ddd508e2f0fb'
[01:20] <cone-64> ffmpeg.git 03Mickaël Raulet 07master:5c3fa74b94fd: hevc: support luma bit depth != chroma bit depth for PCM coding units
[01:20] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:88839718b58c: Merge commit '5c3fa74b94fdedfd63b606f36920d7fbbc069b72'
[01:29] <cone-64> ffmpeg.git 03Martin Storsjö 07master:d307e408d4a9: arm: Don't clobber callee saved registers in scalarproduct
[01:29] <cone-64> ffmpeg.git 03Michael Niedermayer 07master:70d6ce783cc5: Merge remote-tracking branch 'qatar/master'
[02:22] <cone-64> ffmpeg.git 03Derek Buitenhuis 07release/1.1:1a5a6ac01b0a: pthread: Fix deadlock during thread initialization
[02:22] <cone-64> ffmpeg.git 03Michael Niedermayer 07release/1.1:5bd291e265b9: avfilter/ff_insert_pad: fix order of operations
[02:22] <cone-64> ffmpeg.git 03Michael Niedermayer 07release/1.1:b589b8a7042f: Merge commit '1a5a6ac01b0ad2cf3d2128372ea41f3c1cfc2d3f' into release/1.1
[02:41] <cone-64> ffmpeg.git 03Ben Jackson 07release/1.1:311583e77982: pthread: Avoid spurious wakeups
[02:41] <cone-64> ffmpeg.git 03Diego Biurrun 07release/1.1:a3f8c6a42759: x86: ac3dsp: Remove 3dnow version of ff_ac3_extract_exponents
[02:41] <cone-64> ffmpeg.git 03Anton Khirnov 07release/1.1:56eded8bc7bc: mpeg4videodec: split initializing static tables into a separate function
[02:41] <cone-64> ffmpeg.git 03Michael Niedermayer 07release/1.1:2d7f139155dd: Merge commit '56eded8bc7bccdf14245bae3a45b0fecf9d9d122' into release/1.1
[02:49] <cone-64> ffmpeg.git 03Anton Khirnov 07release/1.1:bd405475ceb3: mpeg4video_parser: init mpeg4 static tables.
[02:49] <cone-64> ffmpeg.git 03Anton Khirnov 07release/1.1:718a2ddcb898: h264/mpegvideo: do not provide pixel formats for hwaccels that are not compiled in
[02:49] <cone-64> ffmpeg.git 03Reinhard Tartler 07release/1.1:3f7d89034bfe: alsdec: check block length
[02:49] <cone-64> ffmpeg.git 03Martin Storsjö 07release/1.1:bdb975ab699a: arm: Don't clobber callee saved registers in scalarproduct
[02:49] <cone-64> ffmpeg.git 03Michael Niedermayer 07release/1.1:50ff83e3afa7: Merge remote-tracking branch 'qatar/release/9' into release/1.1
[07:44] <clever> Too many buffered pts
[07:44] <clever> what typicaly causes an error like this?
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/0.10:594b8436084c: arm: Don't clobber callee saved registers in scalarproduct
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/0.11:69f724a538ef: arm: Don't clobber callee saved registers in scalarproduct
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/0.6:a94d12b69027: arm: Don't clobber callee saved registers in scalarproduct
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/0.7:cce6bdc75839: arm: Don't clobber callee saved registers in scalarproduct
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/0.8:8925c44db1a2: arm: Don't clobber callee saved registers in scalarproduct
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/0.9:320df1c7c0fe: arm: Don't clobber callee saved registers in scalarproduct
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/1.0:a4c804d0a2ca: arm: Don't clobber callee saved registers in scalarproduct
[10:01] <cone-45> ffmpeg.git 03Martin Storsjö 07release/1.2:3e65caf5bc3e: arm: Don't clobber callee saved registers in scalarproduct
[10:02] <cone-45> ffmpeg.git 03Martin Storsjö 07release/2.0:62f05d6309b8: arm: Don't clobber callee saved registers in scalarproduct
[10:02] <cone-45> ffmpeg.git 03Martin Storsjö 07release/2.1:0d32483a11b0: arm: Don't clobber callee saved registers in scalarproduct
[11:41] <michaelni> clever, where is that error comming from ?
[11:44] <michaelni> ubitux/saste, there are 4 patches for doc/examples on ffmpeg-dev, do you have time/want to review them or should i ?
[11:44] <michaelni> "83663 N   1214 19:46 Ilya Basin      (1.7K) [FFmpeg-devel] [PATCH] fix 2 memleaks in doc/examples/muxing.c (and replies to it)"
[11:54] <wm4> clever: is that a mplayer error?
[12:00] <cone-45> ffmpeg.git 03Peter Ross 07master:8baaa924bd42: wtvdec: when testing for truncated files, actually check the .wtv file size
[12:12] <ubitux> michaelni: i almost never touched that example :p
[12:13] <michaelni> ok, can you help "83792     1219 17:18 JULIAN GARDNER  (3.5K) Re: [FFmpeg-devel] [FFmpeg-user] BUG in use of extradata and extradata_size with dvb subtitles and teletext" ?
[12:13] <michaelni> iam not completely sure what its about
[12:19] <michaelni> ubitux, ill look at the examples
[12:19] <michaelni> example patches
[12:21] <ubitux> yes i planed to look at this one
[12:23] <michaelni> ok, thx
[12:24] <wm4> btw. are the decoding example fine now wrt. to the frame_alloc shit?
[12:24] <wm4> since elenril managed to unify these
[12:25] <ubitux> the regression is fixed?
[12:33] <cone-45> ffmpeg.git 03Ilya Basin 07master:b6714fa077d9: examples/muxing: fix av_frame_free() not called when got_packet is false
[12:33] <cone-45> ffmpeg.git 03Ilya Basin 07master:7d1d0b3ecf98: examples/muxing: use S16 sample_fmt for resample src regardless of codec sample_fmt
[12:33] <cone-45> ffmpeg.git 03Ilya Basin 07master:d1b8e01ef1fe: examples/muxing: fix memleaks in resampler
[13:48] <cone-45> ffmpeg.git 03Michael Niedermayer 07master:263105deebbd: avcodec/vqavideo: check destination size on all exit pathes
[13:48] <cone-45> ffmpeg.git 03Michael Niedermayer 07master:03acb035d129: avcodec/vqavideo: clear uninitialized parts of the decode buffer
[14:10] <cone-45> ffmpeg.git 03Michael Niedermayer 07master:587e3c3426be: avformat/electronicarts: use 64bit variable for avio_tell() result
[14:10] <cone-45> ffmpeg.git 03Michael Niedermayer 07master:a07d4bc80a7b: avformat/gxfenc: dont use a int intermediate to store in a unsigned int table
[14:13] <saste> is there any use for softpulldown?
[14:19] <saste> or in other words, isn't telecine already a valid substitute for it?
[14:34] <saste> ubitux: ^^?
[14:53] <ubitux> saste: dunno? :P
[17:21] <cone-45> ffmpeg.git 03Michael Niedermayer 07master:55328d5f7246: avformat/electronicarts: detect endianness by comparing sizes instead of using an arbitrary threshold
[19:25] <cone-45> ffmpeg.git 03Michael Niedermayer 07master:ea1e630c47e7: vcodec/rv34: use av_mallocz() to allocate tables
[19:44] <clever> wm4: yeah, i found the error in mplayer source
[19:44] <clever> wm4: i just cant figure out why its happening for this pix fmt
[19:45] <clever> i'm trying to modify my AVHWAccel to output AV_PIX_FMT_RGB565BE
[19:45] <clever> but it refuses to open or query the vo module
[19:46] <wm4> it normally means that you're putting in more packets than you get frames out (for a certain codec delay)
[19:47] <clever> and it not even trying to open a vo module, so no frames are going to the output
[19:49] <clever> wm4: http://privatepaste.com/9cfb1bb465 it also ignores -frames 3
[19:50] <wm4> I wouldn't use mplayer to test this
[19:50] <wm4> because mplayer is full of crap
[19:52] <clever> i can pull the code back out and try it on ffplay, but do you know how well ffmpeg/ffplay handle rgb565 internally?
[19:52] <clever> trying to avoid pixel conversion, so i just set the hw decoder to create 565 directly
[19:52] <wm4> ffplay will just convert it to yuv420p I think
[19:53] <nevcairiel> 565 is a rather terrible format, why do you want it anyway
[19:53] <clever> there are no working vo modules that accept yuv420
[19:53] <clever> so if i pick yuv420, it has to convert in software
[19:53] <clever> and most of the working modules are converting yuv420 into 565 in software
[19:54] <nevcairiel> the decoder has to work in yuv420p internally, so there is a conversion somewhere, are you certain its not also done in software by the hw driver? :p
[19:55] <clever> ive yet to get ffmpeg to play with rgb565, so i dont know
[19:55] <clever> but acording to the docs, it should be all in hardware
[19:56] <clever> i would stick to yuv420, but the compositor is only giving out green, it ignores the u/v channels completely
[19:56] <nevcairiel> sounds like you might've send the information wrong
[19:57] <clever> i was getting various ghosting artifacts when i gave ffmpeg bad yuv420, the colors where all over the place
[19:57] <clever> but this time, i'm not even getting that, just green
[19:58] <clever> http://gallery.earthtools.ca/index.py/paste/DSCF0558.JPG
[19:59] <ubitux> sounds like zero-ed luma plane
[19:59] <ubitux> oh well, zero-ed chroma
[19:59] <clever> i tried doing a memset on the chrome planes with 128, but it does the same thing
[20:00] <ubitux> can we see the code?
[20:01] <clever> one sec
[20:01] <clever> http://privatepaste.com/3c802f9886 is the vo module
[20:01] <clever> createFrame allocates a buffer in the cpu and sets up the 3 plane pointers
[20:02] <clever> updateFrame copies it to the gpu
[20:02] <clever> dpFlip changes which resource is being displayed
[20:02] <ubitux> ah, mplayer stuff
[20:02] Action: ubitux goes away
[20:03] <clever> ubitux: how easy is it to add a vo module to ffplay?
[20:03] <ubitux> ffplay is sdl only so far
[20:03] <clever> now you see why i went with mplayer, lol
[20:04] <clever> the sdl libs are software only, -vo x11 works better
[20:04] <ubitux> there is an accelerated video output driver in the rpi?
[20:05] <clever> thats what i just made
[20:05] <ubitux> just like vdpau, there is an api for hw decoding, and an api for hw rendering?
[20:05] <clever> the gpu uses dispmanx to composite up to 20 layers into the final video output
[20:05] <clever> for each scan line, it will fetch the relevant pixel data and convert it on the fly, from all layers that are in that line
[20:06] <clever> with hw scalling, alpha, and pixel conversion
[20:08] <clever> i wrote the hw decode and hw rendering as seperate modules, both working with yuv420, so it can easily work with other codecs
[22:04] <clever> ubitux: ok, i finaly got SOME color, but its totaly scrambled
[22:04] <clever> the source size is specified with width << 16, height << 16, and if i remove the shifting, i get various colored lines
[22:15] <ubitux> BBB: WIP, probably not as side tracked as the first attempt: https://github.com/ubitux/FFmpeg/compare/vp9-lpf
[22:16] <ubitux> only the fm/flat8in/flat8out check are done& now starts the "fun" :s
[22:16] <ubitux> of course, it might not even actually work
[22:17] <ubitux> i'm pretty sure it doesn't but well :)
[22:26] <BBB> simd usually doesn't work on a first attempt
[22:26] <BBB> ubitux: nice work!
[22:27] <BBB> I have a working 8x8 sub-16x16 idct, and a non-working 4x4 sub-16x16
[22:27] <ubitux> :)
[22:27] <BBB> or maybe we should call it 16x16 sub-8x8 and 16x16 sub-4x4
[22:27] <BBB> I'm debugging the sub4x4 but have had some busy work days lately
[22:28] <ubitux> same, just started working on lpf only today
[22:52] <BBB> well that was easy, 4x4 works also now
[22:52] <BBB> silly typo
[22:52] <BBB> (as is always the case)
[22:52] <BBB> 2x2 still to be done, also no speed measurements yet
[22:52] <BBB> then on to 32x32 again, I can feel it hating me
[22:55] <ubitux> :)
[23:37] <cone-372> ffmpeg.git 03Michael Niedermayer 07master:b797a00f6346: avformat/rmdec: zero string destination before use
[00:00] --- Sun Dec 22 2013

More information about the Ffmpeg-devel-irc mailing list