[00:15] <ubitux> saste: you forgot to update fate after the time parsing patch [00:15] <saste> ubitux, how is it affected? [00:15] <ubitux> -12:34 -> error [00:15] <ubitux> +12:34 -> +754000000 [00:16] <ubitux> parseutils [00:16] <ubitux> fate is staining toward yellow [00:18] <gnafu> Ew, "yellow stains." [00:19] <cone-345> ffmpeg.git 03Stefano Sabatini 07master:e95637841c0a: tests: fix parseutils test after 12a269a5229d3a37be0743fc9655f743ebc44b6e [00:25] <ubitux> btw [00:26] <ubitux> is 30:12.123 supported? :) [00:27] <ubitux> mmh seems it is [00:27] <ubitux> great [00:28] <cone-345> ffmpeg.git 03Clément BSsch 07master:0ac71f9a320d: fate/subtitles: add character encoding conversion test. [00:42] <saste> uhm... dynamic expression evaluation in volume filter [00:48] <saste> how should I handle NaN? [00:48] <saste> NaN = 0? [01:39] <Compn> saste : like if a pointer returns NaN ? [01:39] Action: ubitux is lost with the format negociation thing [01:39] Action: Compn no clue :) [01:40] <Compn> ubitux : audio or video ? [01:40] <ubitux> audio [01:40] <ubitux> Compn: NaN out of eval() i guess [01:40] <ubitux> saste: NaN = 1 i'd say (unchanged output) [01:41] <wm4> ubitux: libavfilter format negotiation? [01:41] <ubitux> wm4: yes [01:41] <wm4> fun stuff, I guess... [01:41] <ubitux> saste: i don't need to set anything for the output? [01:41] <ubitux> removing the format thing from query_formats in ebur128 seems to solve stuff [01:41] <ubitux> (for the output) [01:42] <Compn> michaelni : i didnt get reply from laurent (fenrir) about dxva2 patch [02:05] <cone-345> ffmpeg.git 03Matt Wolenetz 07release/1.1:5bed920971c5: Fix Win64 AVX h264_deblock by not using redzone on Win64 [02:30] <michaelni> Compn, did you (or someone else) test that dxva2 patch ? [02:32] <Compn> i cant find anyone to test it , either [03:40] <cone-345> ffmpeg.git 03Clément BSsch 07fatal: ambiguous argument 'refs/tags/n1.1.3': unknown revision or path not in the working tree. [03:40] <cone-345> Use '--' to separate paths from revisions [03:40] <cone-345> refs/tags/n1.1.3:HEAD: fate/subtitles: add character encoding conversion test. [03:48] <michaelni> av_interleaved_write_frame(): No space left on device [03:48] <michaelni> once again one of my fate boxes ate up all its space ... [03:49] <Compn> michaelni : it would be nice if someone came up with a list of cards that support dxva2 as well [03:49] <Compn> at least then we can try to find someone [03:49] <Compn> with the right card [03:50] <michaelni> yes [03:51] <Compn> i guess my card supports dxva 2.0 [03:51] <Compn> radeon 5830 [03:51] <Compn> i could test it [04:00] <michaelni> if you find bugs in it then the author could fix them and improve the patch, so its surely a good idea to test [04:02] <Compn> michaelni : still, someone has to review it [04:02] <Compn> michaelni : what about including that vid.stab filter ? [04:03] <Compn> that the author has asked if he should work on including a wrapper for a 3rd party dll or include the whole thing in libavfilter ? [04:03] <Compn> or its up to saste ? [04:05] <michaelni> did the author state exactly what he would prefer ? [04:06] <michaelni> did someone object ? [04:11] <Compn> michaelni : theres a lot of 'ffmpeg policy this ' and 'ffmpeg policy that' [04:11] <Compn> i dont see any actual objections [04:12] <Compn> everyones waiting for the great dictator to make his judgement ;) [04:12] <Compn> so itd be cool to have your opinion if you could write a few words on the subj [04:12] <Compn> or your official decree [04:12] <Compn> your royal proclamation [04:15] <Compn> i think he stated that his filter is faster (by a factor of 3-4) and has more features than ours [04:15] <Compn> are we going to add in every video filter until we port them? or keep them seperate until we port them ? are we going to port every filter? so many qustions [04:20] <Compn> thanks for replying :) [04:30] <tonsofpcs> still no answer about dts? [04:34] <michaelni> tonsofpcs, dts=pts when theres no reordering (or the file is broken or the code is buggy) [04:37] <tonsofpcs> michaelni - why is this done? [10:38] <cahit> i have used swresample instead of patching aacdec for S16 sample format, now i get audio values like: 186 2 0 0 75 255 0 0 221 254 0 0 243 254 0 0 173 0 0 0 234 1 0 0 82 255 0 0 80 255 0 0 87 5 0 0 182 8 0 0 153 5 0 0 160 2 0 0 101 3 0 0 [10:38] <cahit> lots of zeros inbetween [10:39] <nevcairiel> looks like stereo with one silent channel [10:40] <cahit> yes, but the audio is stereo [10:40] <cahit> i init resample context like this: [10:40] <cahit> SwrContext* resampleCtx = swr_alloc_set_opts(NULL,av_get_default_channel_layout(pCodecCtx->channels), AV_SAMPLE_FMT_S16, pCodecCtx->sample_rate, av_get_default_channel_layout(pCodecCtx->channels), pCodecCtx->sample_fmt,pCodecCtx->sample_rate, 0, 0); [10:40] <cahit> it should be safe as i don't change channels etc.. [10:44] <cahit> i think i give the input correct: int dataSize = swr_convert(resampleCtx, &resampledOut, AVCODEC_MAX_AUDIO_FRAME_SIZE,((const uint8_t**) &aligned_samples), length / (pCodecCtx->channels * 2)); [10:44] <cahit> do you have any ideas related to what i am missing here? [10:45] <cahit> (if i print audio values before resample, i don't get zeroes) [10:45] <nevcairiel> the decoder probably outputs planar audio, and you're only providing the first plane [10:46] <nevcairiel> the "in" parameter to swr_convert should simply be AVFrame->data [10:46] <cahit> ov! [10:49] <cahit> so i should give it const uint8_t *in[] { a,b } [10:50] <nevcairiel> since you obviously use avcodec for decoding, why not simply use the "data" member of the AVFrame you get from the decode function, and be done with it? :) [10:50] <nevcairiel> dont need to reshuffle data [10:50] <nevcairiel> or pointers [10:50] <nevcairiel> but yes, it needs an "array" of pointers to the different planes [10:52] <cahit> i am using avcodec_decode_audio3 function and manually expanding the data pointer until all the data in that av_read_frame packet. at the end of every read_frame loop i push decoded data to another buffer. [10:52] <cahit> so after avcodec_decode_audio3 i only have a single uint pointer [10:54] <nevcairiel> i would suggest to do the S16 conversion inside this loop and push the final S16 data into your buffer [10:54] <nevcairiel> would probably simplify it a bit [10:57] <cahit> and i think i have to use avcodec_decode_audio4, or i have to reverse current audio3 response to generate other other plane from returned data... [10:58] <nevcairiel> using the latest version is always a good idea, the older ones will go away eventually [11:00] <cahit> thank you for your help, it's a lot cleaner now [12:28] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:3c14c82b7e01: avfilter: Silence warning: passing argument 3 of av_image_copy from incompatible pointer type [12:28] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:cfcab4c50747: vf_overlay: silence warning: X may be used uninitialized in this function [12:29] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:23c9180c001c: aacdec: Fix warning: initialization from incompatible pointer type [12:31] <durandal_1707> michaelni: uploaded gif.tar to incoming (for animation gif coverage) [12:38] <michaelni> durandal_1707, uploaded [13:28] <durandal_1707> michaelni: uploaded exr.tar.bz2 to incoming (for some exr coverage) [13:57] <michaelni> durandal_1707, uploaded [14:50] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:5270cb39ba7c: bmp: Fix warning X may be used uninitialized in this function [14:50] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:c10c2aed47da: h264: Silence warning: comparison of distinct pointer types lacks a cast [14:52] <durandal_1707> "libav is more or less only the more API-stable version of ffmpeg. " [15:10] <ubitux> durandal_1707: lol [15:11] <durandal_1707> ubitux: it is one of comments on debian community on g+ [15:11] <cahit> int dataSize = swr_convert(resampleCtx, &resampledOut, AVCODEC_MAX_AUDIO_FRAME_SIZE,(const uint8_t**)decoded_frame.extended_data, (decoded_frame.linesize[0]+decoded_frame.linesize[1]) / (pCodecCtx->channels * 2)); [15:11] <cahit> i am doing something wrong (: [15:12] <cahit> (trying to convert planar pcm to s16 for wav) [15:12] <nevcairiel> isnt there a decoded_frame.samples or something for the number of samples in a frame [15:13] <nevcairiel> besides, the number of samples is defined by one plane, you dont have to add both [15:15] <cahit> and when i'm pushing it to queue, i do this: av_fifo_generic_write(fifoPlayback,(uint8_t *)resampledOut,dataSize*4 [15:16] <cahit> i didn't understand the conversion, so i found out that i have to multiply with 4 by chance. try&fail [15:20] <durandal_1707> cahit: did you explored other usages of swr_convert() in source code? [15:21] <cahit> yes, i actually tried different parts from swresample-test.c [15:30] <cone-978> ffmpeg.git 03Paul B Mahol 07master:df63e0c8bbf3: doc/filters: add forgotten sentence for blend filter examples [15:30] <cahit> i think i have to multiply by 4 as i have devided the length with pCodecCtx->channels * 2, right? [15:30] <durandal_1707> cahit: have you read documentation? [15:31] <cahit> i really did, i think i am not good at understanding differences these days :-/ [15:32] <durandal_1707> https://ffmpeg.org/doxygen/trunk/group__lswr.html#gaa5bb6cab830146efa8c760fa... [15:33] <durandal_1707> so 1 sample in s16 case is 2 bytes [15:35] <cahit> i see [15:36] <cahit> thank you for your patience, really.. i got it now. [16:40] <wm4> durandal_1707: looking at this, I wonder what "Convertion will run directly without copying whenever possible." is supposed to mean? [17:01] <durandal_1707> wm4: ? [17:02] <wm4> durandal_1707: doesn't the API copy the sample data from one place to another? [17:02] <durandal_1707> wm4: what API? [17:02] <wm4> swr [17:05] <durandal_1707> michaelni, this sure can be better worded [17:07] <durandal_1707> wm4: i think that means extra copying, but you could try pass in = out and see what happens [17:16] <michaelni> durandal_1707, feel free to change any wording to be better [17:17] <michaelni> iam not the best at wording things ... [17:19] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:02ac3398eb52: rtmpproto: Check APP_MAX_LENGTH [17:29] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:da8ef5ac2f1b: rtmpproto: increase APP_MAX_LENGTH [17:45] Action: durandal_1707 hate potato programmers [19:33] <cone-978> ffmpeg.git 03Michael Niedermayer 07master:73fce258b774: vf_mp: Set pseudo pal [22:39] <cone-267> ffmpeg.git 03Michael Niedermayer 07release/1.1:1f9073f41be1: vf_mp: Set pseudo pal [23:16] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:285485ac5f89: matroskaenc: fix cue tracknum off by 1 error [23:16] <cone-267> ffmpeg.git 03Michael Niedermayer 07master:82d79289db03: avformat: Allocate duration_error separately [00:00] --- Sun Feb 24 2013