December 2009 - ffmpeg-devel - lists.ffmpeg.org

[PATCH 4/6] RTMP output: add header compression for output
by piratfm＠gmail.com 04 Dec '09

04 Dec '09

It works with all tested servers. Also one todo. Now 1byte hdr. following 12byte hdr. can't be created.

2 5

[RFC] Request for pixdesc API review
by stefano.sabatini-lala＠poste.it 04 Dec '09

04 Dec '09

Hi, we recently got completed all the definitions for the pixel format descriptors in libavutil/pixdesc.{h,c}. pixdescs API is the designated replacement for the hackish PixFmtInfo system defined in libavcodec/imgconvert.c, and it may be used, once completed, at least by lavfi and lsws code. Since now mainly just Micheal and me worked on it, before pixdescs API can be published, we need, as suggested by Michael, a review of the new API from some other developer, especially to get sure that it is clear enough and we are not missing pieces. See the thread: http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/99224 In particular I plan to move it back to lavc and start to replace the functions using PixFmtInfo with corresponding functions using pixdescs, then once this will be done I plan to move it again to lavu and enable it. Comments are very welcome. Regards. -- FFmpeg = Fierce and Faboulous Magical Pitiless Erratic God

4 32

Fix decoding Xvid stream problem
by konsunwin＠gmail.com 04 Dec '09

04 Dec '09

Hi,everyone I use libavcodec to decode Xvid encoded stream, but sometime error, output as: [mpeg4 @ 0x8bae990]hmm, seems the headers are not complete, trying to guess time_increment_bits [mpeg4 @ 0x8bae990]my guess is 1 bits ;) [mpeg4 @ 0x8bae990]looks like this file was encoded with (divx4/(old)xvid/opendivx) -> forcing low_delay flag but when output "my guess is 5 bits", it can be rightly decoded, so I modified the code as follows: --- h263dec_ref.c 2009-03-02 13:18:33.000000000 +0800 +++ h263dec.c 2009-12-04 12:10:39.000000000 +0800 @@ -62,7 +62,7 @@ break; case CODEC_ID_MPEG4: s->decode_mb= ff_mpeg4_decode_mb; - s->time_increment_bits = 4; /* default value for broken headers */ + s->time_increment_bits = 5; /* default value for broken headers */ s->h263_pred = 1; s->low_delay = 0; //default, might be overriden in the vol header during header parsing break; --- h263_ref.c 2009-02-25 00:12:47.000000000 +0800 +++ h263.c 2009-12-04 12:19:23.000000000 +0800 @@ -5878,7 +5878,7 @@ time_incr++; check_marker(gb, "before time_increment"); - +/* if(s->time_increment_bits==0 || !(show_bits(gb, s->time_increment_bits+1)&1)){ av_log(s->avctx, AV_LOG_ERROR, "hmm, seems the headers are not complete, trying to guess time_increment_bits\n"); @@ -5888,7 +5888,7 @@ av_log(s->avctx, AV_LOG_ERROR, "my guess is %d bits ;)\n",s->time_increment_bits); } - +*/ if(IS_3IV1) time_increment= get_bits1(gb); //FIXME investigate further else time_increment= get_bits(gb, s->time_increment_bits); Are there any other better advices ?

2 1

[PATCH] use lzw compression in gif encoder
by baptiste.coudurier＠gmail.com 03 Dec '09

03 Dec '09

Hi guys, Finally. The func ptr in the lzw encoder is the least worse solution I could find. -- Baptiste COUDURIER Key fingerprint 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA FFmpeg maintainer http://www.ffmpeg.org

4 5

[PATCH] Fix linesize computation for NV formats
by stefano.sabatini-lala＠poste.it 03 Dec '09

03 Dec '09

Hi, for NV12 and NV21 data[1] contains both the U and V components, so it needs to be multiplied by 2. Alternatively we could have: linesize[1] = width which is equivalent. I couldn't find any NV12 samples and libswscale cannot issue that in output, so I didn't tested it. Regards. -- FFmpeg = Faithless and Faithful Mystic Puristic Exploitable Game

3 20

Is av_log_level supposed to be exported?
by ramiro.polla＠gmail.com 03 Dec '09

03 Dec '09

Hi, av_log_level from libavutil/log.c is only used in that file. It is not declared in libavutil/log.h. It's meant to be used with the av_log_get_level() and av_log_set_level() functions. Should it be made static in log.c, or declared in libavutil/log.h so users can access it directly? Ramiro Polla

3 11

[PATCH 3/6] RTMP output: change channel for connect and create_stream
by piratfm＠gmail.com 03 Dec '09

03 Dec '09

2009/12/2 Kostya <kostya.shishkov at gmail.com>: > Part 7 - several things mixed - I'm not against changing invoke channel > but why? I'm just look in network dumps, and player using system channel for connecting and creating stream, also fme using this channel for releasing and publishing stream. Video channel player using to sending play command.

2 1

[BUG]+[Workaround] Video concat example from faq broken (yuv4mpegpipe broken?)
by abodera＠gmail.com 03 Dec '09

03 Dec '09

Hey everyone! Example taken directly from http://ffmpeg.org/faq.html. The third example for "How can I join video files?" does not work. Tried under different compilations of ffmpeg under x86_64. For example on ffmpeg rev. 20520 using the EXACT copy+paste of the example from FAQ, with the exact same file formats and names, on Linux, always results in a video file with only the first segment (input1.flv) of video. The container reports it has the length of input1+input2 but seeking to later parts can not be performed because there is no data (eof in vlc). I have found out it is related to yuv4mpegpipe which fails to work when concatenated. For example when disregarding pipes and performing the following sequentially: ffmpeg -i input1.flv -an -f yuv4mpegpipe input1.yuv ffmpeg -i input2.flv -an -f yuv4mpegpipe input2.yuv cat input1.yuv input2.yuv > input_combined.yuv ffmpeg -f yuv4mpegpipe -i input_combined result.flv It would also result with broken flv file, with only the first segment (input1.yuv) - this suggests that it is not related to pipes, but rather the yuv4mpegpipe codec. A simple workaround: use rawvideo instead! The only caveat is that you need to specify the exact video dimensions in the last ffmpeg invocation (because rawvideo does not encapsulate that). Here's an example script of video concat using rawvideo: #!/bin/bash mkfifo temp1.a mkfifo temp1.v mkfifo temp2.a mkfifo temp2.v mkfifo all.a mkfifo all.v dims=`ffmpeg -i $1 2>&1 | grep -Po 'Stream.*?Video.*?(\d{3,4}x\d{3,4})' | grep -Po '\d+x\d+'` ffmpeg -i $1 -vn -f u16le -acodec pcm_s16le -ac 2 -ar 44100 - > temp1.a < /dev/null & ffmpeg -i $2 -vn -f u16le -acodec pcm_s16le -ac 2 -ar 44100 - > temp2.a < /dev/null & ffmpeg -i $1 -an -f rawvideo - > temp1.v < /dev/null & ffmpeg -i $2 -an -f rawvideo - > temp2.v < /dev/null & cat temp1.a temp2.a > all.a & cat temp1.v temp2.v > all.v & ffmpeg -f u16le -acodec pcm_s16le -ac 2 -ar 44100 -i all.a \ -f rawvideo -s $dims -i all.v \ -b 128k -y $3 rm temp[12].[av] all.[av] Best regards, Arthur. -- __ /.)\ +48 695 600 936 \(./ abodera at gmail.com

1 0

[PATCH] Add -fpre option.
by ramiro.polla＠gmail.com 03 Dec '09

03 Dec '09

Hi, msg7668 in issue659 spots another case where the preset files won't open a valid path. Also "../" doesn't work for example. Attached patch makes it the user's job to pass -fpre to open a preset file and FFmpeg will just pass it on to open(). Ramiro Polla

3 9

[BUG, PATCH] Revert broken use of ff_put_string() in vorbis_enc.c
by pmhahn＠pmhahn.de 02 Dec '09

02 Dec '09

Revert commit 20095: Don't use ff_put_string in vorbis encoder. ff_put_string() is declared in bistream.c, which uses put_bits() without BITSTREAM_WRITER_LE being defined. vorbis_enc.c on the other hand #defines BITSTREAM_WRITER_LE and thus must not use this function. Broken files contain the string "vorsib" instead of "vorbis". --- libavcodec/vorbis_enc.c (Revision 20639) +++ libavcodec/vorbis_enc.c (Arbeitskopie) @@ -530,7 +530,14 @@ // identification header init_put_bits(&pb, p, buffer_len); put_bits(&pb, 8, 1); //magic - ff_put_string(&pb, "vorbis", 0); + /* + * Do not use + * bitsream.c:void ff_put_string(PutBitContext * pbc, const char *s, int terminate_string) + * because it is Big-Endian which is incompatible with + * vorbis_enc.c:#define BITSTREAM_WRITER_LE + */ + for (i = 0; "vorbis"[i]; i++) + put_bits(&pb, 8, "vorbis"[i]); put_bits32(&pb, 0); // version put_bits(&pb, 8, venc->channels); put_bits32(&pb, venc->sample_rate); @@ -549,7 +556,8 @@ // comment header init_put_bits(&pb, p, buffer_len); put_bits(&pb, 8, 3); //magic - ff_put_string(&pb, "vorbis", 0); + for (i = 0; "vorbis"[i]; i++) + put_bits(&pb, 8, "vorbis"[i]); put_bits32(&pb, 0); // vendor length TODO put_bits32(&pb, 0); // amount of comments put_bits(&pb, 1, 1); // framing @@ -562,7 +570,8 @@ // setup header init_put_bits(&pb, p, buffer_len); put_bits(&pb, 8, 5); //magic - ff_put_string(&pb, "vorbis", 0); + for (i = 0; "vorbis"[i]; i++) + put_bits(&pb, 8, "vorbis"[i]); // codebooks put_bits(&pb, 8, venc->ncodebooks - 1); BYtE Philipp PS: I'm not subscribed, please cc: me on replies. -- / / (_)__ __ ____ __ Philipp Hahn / /__/ / _ \/ // /\ \/ / /____/_/_//_/\_,_/ /_/\_\ pmhahn at titan.lahn.de

4 3