[Ffmpeg-devel-irc] ffmpeg-devel.log.20140406

burek burek021 at gmail.com
Mon Apr 7 02:05:02 CEST 2014

[01:41] <cone-788> ffmpeg.git 03Michael Niedermayer 07master:1fec720afa49: avformat/cinedec: ask for sample when biBitCount isnt supported
[01:41] <cone-788> ffmpeg.git 03Michael Niedermayer 07master:419800acc46a: examples/avio_reading: fix null dereference on error
[01:41] <cone-788> ffmpeg.git 03Michael Niedermayer 07master:b978391ed5fd: avformat/oggparsevorbis: remove non functional null check
[03:07] <BBB> ubitux: I sort of expected that thread to devolve into that sort of rubbish :(
[03:40] <cone-788> ffmpeg.git 03Carl Eugen Hoyos 07master:59c1023ef544: Consider aspect ratio when writing tiff dpi.
[06:18] <cone-788> ffmpeg.git 03Michael Niedermayer 07master:f87b3d552394: avcodec/pthread_frame: fix missing unlock on error
[06:18] <cone-788> ffmpeg.git 03Michael Niedermayer 07master:754f84663e8b: avcodec/msrle: check return code for success before use
[13:34] <nevcairiel> is there an easy way to scan an assembled function if it uses instructions of a certain simd function set? ie. I want to verify a sse function doesn't accidentally use sse2 functions
[13:53] <J_Darnley> I don't know an existing solution but...
[13:54] <J_Darnley> you could grep a plain-text dump of the instructions for certain instructions
[16:02] <cone-625> ffmpeg.git 03Thilo Borgmann 07master:33462542cf47: configure: Add workaround for qtkit detection using ICC compiler.
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:edf4406c4dda: avformat/hls: add support for alternative renditions
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:cc1162d7751c: avformat/hls: always return EOF for unneeded playlist streams
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:da7759b3579d: avformat/hls: add support for byte-ranged segments
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:f7c7fc13698a: avformat/hls: add some logging messages
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:1cb8d986b8d4: avformat/hls: split read_from_url() out of read_data()
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:01b184e68d86: avformat/hls: parse ID3 timestamps for elementary audio streams
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:61e1a702783c: avformat/hls: track seeking on a per-playlist basis
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:4e85202bcc26: avformat/hls: improve segment selection when restarting list reception
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:d549b0910c34: avformat/hls: avoid unnecessary segment retrievals
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:8fd6875c83d5: avformat/hls: do not care about stream start timestamps
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:6b4b73e75da9: avformat/hls: do not use sequence numbers for packet ordering
[16:56] <cone-625> ffmpeg.git 03Anssi Hannula 07master:05ce529a596c: avformat/hls: factor identical playlist allocations out of parse_playlist
[16:57] <cone-625> ffmpeg.git 03Anssi Hannula 07master:ad701326b430: avformat/hls: open playlists immediately when AVDISCARD_ALL is dropped
[16:57] <cone-625> ffmpeg.git 03Anssi Hannula 07master:9aa0606e87a2: avformat/hls: properly take stream_index into account when seeking
[16:57] <cone-625> ffmpeg.git 03Anssi Hannula 07master:f29cb45bf2cf: avformat/hls: flush the subdemuxer when seeking
[16:57] <cone-625> ffmpeg.git 03Vladlen Y. Koshelev 07master:cdbd9e218d44: avformat/hls: support for seeking on event playlists
[17:14] <nevcairiel> so I finally found the offending instruction, and my disasm claims its a movq .. is movq sse2?
[17:14] <nevcairiel> 6A2F2851  movq        mmword ptr [eax+ecx],xmm2  
[17:14] <nevcairiel> specifically
[18:25] <cone-625> ffmpeg.git 03Michael Niedermayer 07master:28f8114bd240: avcodec/vp8: fix checking vp7_feature_value_size
[18:25] <cone-625> ffmpeg.git 03Michael Niedermayer 07master:c1b24ca762cb: avformat/oggdec: check for av_packet_new_side_data() failure
[18:25] <cone-625> ffmpeg.git 03Michael Niedermayer 07master:4ccd7cb45b9a: avformat/oggdec/ogg_read_packet(): factorize failure code path
[18:28] <michaelni> nevcairiel, movq with xmm should be sse2
[18:28] <nevcairiel> it appears so
[18:28] <michaelni> 0x6a is a push though not a movq
[18:29] <nevcairiel> thats just the binary offset
[18:29] <nevcairiel> not the instruction
[18:29] <michaelni> ahh ok thought it was intended to be the intrsuctions bytes
[18:33] <nevcairiel> I sent a patch to the ML to correct it
[18:44] <cone-625> ffmpeg.git 03Hendrik Leppkes 07master:fc7e02f0ff34: dcadsp: fix SSE code to not use SSE2 instructions.
[19:29] <ubitux> why isn't movq defined on movh for avx2?
[19:30] <ubitux> (since movq xm0, blah seems to work)
[19:31] <ubitux> (it translates to vmovq)
[19:34] <ubitux> mmh i guess that's because movh means "half" and thus for ymm it would mean dq
[19:38] <ubitux> i don't get why movdq[au] (or mov[au]) would be 256 load with ymm though
[19:40] <cone-625> ffmpeg.git 03Reimar Döffinger 07master:57d10a28f41b: pulseaudio: move NULL check up before dereference.
[19:47] <ubitux> side note: today libav decides to NIH fontconfig support in drawtext, with a different configure option ofc
[19:49] Action: ubitux wonders if the answer will be "i'm not following all our downstreams"
[19:49] <nevcairiel> of course it will be
[19:49] <nevcairiel> its a handy excuse
[20:01] <kierank> ubitux: that was why i moved to ffmpeg
[20:01] <kierank> drawtext and friends
[20:30] <BBB> nevcairiel: I've used qemu for that in the past
[20:30] <BBB> nevcairiel: that actually worked
[20:30] <BBB> nevcairiel: but it's a little ... clunky
[20:32] <BBB> nevcairiel: also if move is all there is, see has an analog for that and I think x86inc.asm knows it?
[22:05] <cone-625> ffmpeg.git 03Reimar Döffinger 07master:a0568ed25440: rtspdec: check for strchr returning NULL.
[22:36] <cone-625> ffmpeg.git 03Paul B Mahol 07master:a420ccd4f2a0: LucasArts SMUSH SANM video decoder
[22:36] <cone-625> ffmpeg.git 03Michael Niedermayer 07master:881f4e8d3440: Merge commit 'a420ccd4f2a011887451a7d5e1bebba4fd7c40e2'
[23:15] <cone-625> ffmpeg.git 03Paul B Mahol 07master:6f273093e54c: LucasArts SMUSH VIMA audio decoder
[23:15] <cone-625> ffmpeg.git 03Michael Niedermayer 07master:787be6d07496: Merge commit '6f273093e54cba130f3ffde3d6433e74baa4ad89'
[23:37] <cone-625> ffmpeg.git 03Paul B Mahol 07master:eeadcdfd1a6f: LucasArts SMUSH demuxer
[23:37] <cone-625> ffmpeg.git 03Michael Niedermayer 07master:5297003aa7a0: Merge commit 'eeadcdfd1a6f3089b6bf6e194d6ece8d3f113123'
[23:46] <J_D> Wow, someone went on a lucasarts bender.
[23:47] <nevcairiel> fwiw, those existed in ffmpeg for quire a while already
[23:56] <cone-625> ffmpeg.git 03Paul B Mahol 07master:a027d2a8731f: fate: add tests for SANM and VIMA
[23:56] <cone-625> ffmpeg.git 03Michael Niedermayer 07master:9e7f1474dd5f: Merge remote-tracking branch 'qatar/master'
[00:00] --- Mon Apr  7 2014

More information about the Ffmpeg-devel-irc mailing list