[Ffmpeg-devel-irc] ffmpeg-devel.log.20150417

[burek]

[00:36:22 CEST] <jamrial> michaelni: are you ok with applying the proper version of the dcadec patch?
[00:50:44 CEST] <michaelni> i have no opinion on the pkg config stuff, also i would avoid "proper implying the other being not so proper" wording the whole debate is heated enough already
[00:53:46 CEST] <jamrial> i said proper because it was the one oked by the author of the wrapper
[00:53:57 CEST] <jamrial> but ok, i'll keep that in mind
[00:55:27 CEST] <michaelni> thx
[01:12:47 CEST] <cone-630> ffmpeg 03James Almer 07master:8254011b2173: configure: remove old libdcadec check
[05:24:31 CEST] <cone-355> ffmpeg 03Vignesh Venkatasubramanian 07master:1c37848f9029: webmdashenc: Add better error handling
[12:56:35 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07master:38f67260684a: avcodec/atrac3plusdsp: fix on stack alignment
[14:34:46 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07master:8768f8f4b929: avcodec/mpegaudiodec_template: use double to build csa tables
[14:44:17 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.2:1c14b09caf90: avcodec/atrac3plusdsp: fix on stack alignment
[15:06:07 CEST] <rcombs> hmm, trying to work out how best to export whether or not a TrueHD stream has an ATMOS sub stream
[15:06:36 CEST] <rcombs> could be a profile, but I don't know if that really makes sense
[15:10:14 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07n2.2.15:HEAD: avcodec/mpegaudiodec_template: use double to build csa tables
[15:11:25 CEST] <rcombs> or stick it in metadata
[15:12:39 CEST] <Daemon404> rcombs, http://www.reactiongifs.com/r/btt.gif 
[15:12:44 CEST] <Daemon404> since were thinking of places to stick it.
[15:12:57 CEST] Action: Daemon404 runs
[15:45:47 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:a376ef4a17ed: avcodec/msrledec: restructure msrle_decode_pal4() based on the line number instead of the pixel pointer
[15:45:48 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:2f67222780be: avutil/pca: Check for av_malloc* failures
[15:45:49 CEST] <cone-263> ffmpeg 03Rainer Hochecker 07release/2.5:071eb56a6a50: h264: avoid unnecessary calls to get_format
[15:45:50 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:52ba406b94d6: avcodec/h264: Fail for invalid mixed IDR / non IDR frames in slice threading mode
[15:45:51 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:76b289bcf2c5: avcodec/h264_refs: Do not set reference to things which dont exist
[15:45:52 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:043cb40becb2: ffmpeg: Fix extradata allocation
[15:45:53 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:8968de6c6193: avformat/utils: avoid discarded streams in av_find_default_stream_index()
[15:45:54 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:48f616ceeecd: avcodec/h264: Fix race between slices where one overwrites data from the next
[15:45:55 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:7fa861dfe06e: avcodec/h264: finish previous slices before switching to single thread mode
[15:45:56 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:f96fdb46b773: avcodec/h264_slice: Dont reset mb_aff_frame per slice
[15:45:57 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:fe2df122b4c3: avcodec/h264: reset the counts in the correct context
[15:45:58 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:3e3042496120: avcodec/aacdec: Fix storing state before PCE decode
[15:45:59 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:9333ee7c0db1: avcodec/h264: Be more tolerant to changing pps id between slices
[15:46:00 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:6f02d93c0fbf: avcodec/h264_ps: Move truncation check from VUI to SPS
[15:46:01 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:8c0fd44317b0: avcodec/h264: Do not fail with randomly truncated VUIs
[15:46:02 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:74fb9df48b12: swresample: Allow reinitialization without ever setting channel counts
[15:46:03 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:84fedd3be768: swresample: Allow reinitialization without ever setting channel layouts (cherry picked from commit 80a28c7509a11114e1aea5b208d56c6646d69c07)
[15:46:04 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:07ea57c5bbbe: swresample: Check channel layouts and channels against each other and print human readable error messages
[15:46:05 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:41b15817ce1e: swresample/swresample-test: Randomly wipe out channel counts
[15:46:06 CEST] <cone-263> ffmpeg 03Andreas Cadhalpun 07release/2.5:9a241d95efa1: msrledec: use signed pixel_ptr in msrle_decode_pal4
[15:46:07 CEST] <cone-263> ffmpeg 03Andreas Cadhalpun 07release/2.5:621f1a2e6364: aacdec: consistently use avctx for logging in decode_eld_specific_config
[15:46:08 CEST] <cone-263> ffmpeg 03Andreas Cadhalpun 07release/2.5:d846d3e88efd: aasc: return correct buffer size from aasc_decode_frame
[15:46:09 CEST] <cone-263> ffmpeg 03Andreas Cadhalpun 07release/2.5:1d0f9c92c550: aacpsy: avoid psy_band->threshold becoming NaN
[15:46:10 CEST] <cone-263> ffmpeg 03Andreas Cadhalpun 07release/2.5:b420354a8b66: ac3: validate end in ff_ac3_bit_alloc_calc_mask
[15:46:11 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:0683ef4b50c4: avcodec/atrac3plusdsp: fix on stack alignment
[16:15:55 CEST] <wm4> michaelni: CBR seeking actually breaks gapless... again
[16:15:59 CEST] <wm4> probably fixable, but still
[16:16:09 CEST] <wm4> rcombs: what do you need to strip it for?
[16:16:27 CEST] <rcombs> wm4: save space; handle dumb decoders
[16:17:51 CEST] <wm4> AFAIK it's designed to be skipped by unaware decoders; ffmpeg did not do this correctly until recently though
[16:18:52 CEST] <rcombs> dingdingding
[16:19:16 CEST] <rcombs> I have to support some old ffmpegs for hysterical raisins
[16:26:03 CEST] <wm4> ah, lol
[16:26:14 CEST] <wm4> I suppose in theory a bsf could be added
[16:26:43 CEST] <saste> rcombs, any news about the segment patch?
[16:27:11 CEST] <saste> I want you to provide some info about the use case in the commit log
[16:35:54 CEST] <rcombs> saste: yes, saw, just need to get around to doing it
[16:40:36 CEST] <Daemon404> g 42
[16:48:56 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07release/2.5:8a59dbbc68f7: Update or 2.5.6
[16:49:42 CEST] <rcombs> 2.5.6 \o/
[17:00:41 CEST] <wm4> michaelni: so due to codec internals, the mp3 decoder (like many other DCT based decoders, apparently) do not produce output for half a frame or so; shouldn't this silence be skipped?
[17:01:56 CEST] <kierank> wm4: that's a matter for debate iirc
[17:02:14 CEST] <kierank> and whole papers written about it
[17:03:01 CEST] <wm4> well, the gapless stuff cuts it away
[17:03:13 CEST] Action: Daemon404 has bad flashbacks to opus integration
[17:03:23 CEST] <wm4> if you start from the beginning (this appears to be explicitly designed this way in ffmpeg's mp3 code)
[17:03:41 CEST] <wm4> but if you seek in the middle, you get some silence at the start
[17:04:09 CEST] <Daemon404> i dont know why you would use lavc for music albums in the first place...
[17:04:13 CEST] <Daemon404> but thats just me
[17:04:37 CEST] <wm4> because I don't want to use hundreds of broken decoder libs?
[17:04:43 CEST] <wm4> which don't solve this anyway
[17:04:46 CEST] <Daemon404> well i wouldnt use mpv for music
[17:05:10 CEST] <wm4> most wouldn't use it for video either
[17:05:45 CEST] <Daemon404> heh
[17:07:53 CEST] <wm4> Daemon404: every audio player uses some popular mp3 lib like libmpg123 or libmad, and then hacks their own "container" handling around it
[17:07:59 CEST] <wm4> well, open sores audio player
[17:08:44 CEST] <Daemon404> meh
[17:21:57 CEST] <iive> libmad is commercially popular.
[17:47:51 CEST] <cone-263> ffmpeg 03Shivraj Patil 07master:35a7170e69b3: Makefile: Add support for MSA (MIPS-SIMD-Arch)
[17:47:52 CEST] <cone-263> ffmpeg 03Shivraj Patil 07master:4efc0e6451fa: avcodec/mips: MSA (MIPS-SIMD-Arch) optimizations for HEVC horizontal and vertical mc functions
[18:16:35 CEST] <cone-263> ffmpeg 03wm4 07master:92eef26e67ea: avformat/mp3dec: fix gapless audio when seeking in CBR mode
[18:48:11 CEST] <cone-263> ffmpeg 03wm4 07master:2a0aa318f813: fate: gapless: fix a typo in comment
[18:48:12 CEST] <cone-263> ffmpeg 03wm4 07master:49d5c24aa150: fate: gapless: test seeking to a specific position
[18:48:13 CEST] <cone-263> ffmpeg 03wm4 07master:12da8bc38dca: fate: gapless: test mp3 both with and without TOC
[19:33:34 CEST] <fluffrabbit> Is VP9 encoding speed temporal degradation a bug?
[19:36:50 CEST] <fluffrabbit> almost feels like a memory leak or something, and I know that this is not a new issue with FFMPEG, as it has been seen with other codecs
[19:49:04 CEST] <fluffrabbit> I've caught other bugs like this before so really I'd just like somebody else to test and confirm what I'm saying.
[19:50:09 CEST] <fluffrabbit> VP9 only got documentation on the FFMPEG wiki very recently so IDK if you're still considering it "beta" or whatever but it's a codec I've been watching for a long time.
[19:51:38 CEST] <fluffrabbit> Try this: ffmpeg -i in.mp4 -c:v libvpx-vp9 -pix_fmt yuv420p -crf 10 -b:v 0 out.mkv
[19:52:17 CEST] <fluffrabbit> preferably with a video > 1 minute long
[19:52:54 CEST] <fluffrabbit> watch the encoding speed drop throughout
[20:15:14 CEST] <wm4> eh, the jpeg probe check is duplicated between img2dec.c and rawdec.c
[20:15:32 CEST] <wm4> the one in rawdec.c could probably be killed completely
[20:15:58 CEST] <wm4> ubitux: for what subtitle things did you want me to add tests for?
[20:17:58 CEST] <Compn> wm4 : unless you use jpeg raw streams , jpeg webcams, etc, then sure you can remove "duplicated" code...
[20:18:05 CEST] <Compn> or you could just factor it 
[20:20:08 CEST] <wm4> it looks redundant
[20:20:57 CEST] <cone-263> ffmpeg 03Diego Biurrun 07master:247aa7af7d81: avisynth: Simplify shared library name construction
[20:20:58 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07master:67ac31d1bec2: Merge commit '247aa7af7d8197247c181e3fbfe8d93d75e41b29'
[20:34:49 CEST] <cone-263> ffmpeg 03Diego Biurrun 07master:1336bb06c9fb: configure: Simplify avisynth check
[20:34:50 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07master:04fbab147745: Merge commit '1336bb06c9fbf9a14765e9f78616f2aad4f3a45a'
[20:45:55 CEST] <BBB> fluffrabbit: does that occur when you encode directly with vpxenc?
[20:57:37 CEST] <fluffrabbit> sorry, I'm occupied for the next hour or so
[21:37:43 CEST] <cone-263> ffmpeg 03Michael Niedermayer 07master:3acadc89f495: tests/tiny_psnr: Make the search range extend both sides from the specified shift value
[22:15:25 CEST] <fluffrabbit|away> Got a busy day today. I might test with vpxenc when I have the time.
[22:16:26 CEST] <fluffrabbit|away> Seems likely that it is an issue somewhere in ffmpeg's implementation of libvpx, but whether it's in libvpx itself has yet to be seen.
[22:17:44 CEST] <fluffrabbit|away> here I use the word "implementation" loosely as I'm not super clear on how all the little cogs fit together, the make and model of the gears and the grease in between
[00:00:00 CEST] --- Sat Apr 18 2015