August 2010 - ffmpeg-devel - lists.ffmpeg.org

[PATCH] fixup mpegts_get_pcr to be more reliable
by elupus＠ecce.se 14 Sep '11

14 Sep '11

Hi, Here are a few patches i created while trying to figure out why a file "locked" up in mpegts_get_pcr on seek. I don't seem to learn that patches to the mpegts demuxer have a tendency to get lost in the void. Anyway... 0001: The code for backwords pcr searching could never be reached, so remove it. 0002: Respect the maximum stream position that the function is allowed to process. 0003: Make sure the return value of url_seek is respected 0004: Reindent after the above 0005: Make mpegts_get_pcr resync to stream if the position it is asked to look for a timestamp isn't on a mpegts syncword. ---- The real issue I found out later is that this particlar stream doesn't contain any packets with pid = pcr_pid, so it reads the whole file to attempt to find a starting timestamp for av_gen_search. Are the pid's for the pcr supposed to be listed in PMT? I haven't checked this yet, but i suspect the given pid isn't available in there either. Maybe it should be reset to -1 if the pid isn't found? Regards Joakim

2 1

[PATCH] Screen frame grabbing for Win32 platform (bump)
by christophe.gisquet＠gmail.com 21 Jul '11

21 Jul '11

Hello, now for an even older thread bump: https://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2007-February/022360.html I've just adapted it to the new system and just left in the cursor capture in. It should be noted the frame grabbing itself is very slow and nothing much can be done about it, except going a _completely_ different way. Best regards, Christophe

11 27

[PATCH] lavfi test pattern generator
by stefano.sabatini-lala＠poste.it 26 Jun '11

26 Jun '11

Hi all, useful for testing, it is based on the test pattern generator by Nicolas George: http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2007-October/036838.html basically I just stripped the audio signal generator and adapted few things to the lavfi framework, so most credit goes to Nicolas. And it would plenty cool to find some way to get ffmpeg/ffplay take as input such a generator, but I still can't see how to do that. What I can do now is to take as input /dev/null, then overlay the test_pattern over it, but this is rather ugly. Regards. -- FFmpeg = Freak and Fierce Mysterious Proud Erratic Gadget

7 24

[FWD] [PATCH] immediate decoding of dvd stills
by nicola_sabbi＠fastwebnet.it 24 Jun '11

24 Jun '11

Some time ago someone posted this small patch to permit decoding dvd still images without setting LOW_DELAY; unfortunately it seems the patch was forgotten Is it ok to commit? If not, what's the right way to proceed?

5 17

[PATCH] encoder for adobe's flash ScreenVideo2 codec
by joshuawarner32＠gmail.com 07 May '11

07 May '11

Hi all, I've developed an encoder for Adobe's Flash ScreenVideo2 format, which is stored in flv files. My encoder currently only supports a large subset of the format. The only player that supports this codec (so far) is Adobe Flash Player itself, but ScreenVideo2 makes dramatic improvement in file size over ScreenVideo (currently in ffmpeg as flashsv) - and should be very useful for uploading screencasts, etc. Most options (block size, etc) now just fall back on defaults because I couldn't find a general algorithm that produced consistantly better results than these. All the code is in place to be able to change these parameters dynamically, so future improvements there should be easy. The patch is attached. The codec is listed as flashsv2 in ffmpeg. Joshua Warner

13 41

[Toy] MS Video 1 encoder
by kostya.shishkov＠gmail.com 04 May '11

04 May '11

Here's a result of sleepless day :). With a small patch to ELBG it was possible to implement simple 15-bit MSVC encoder. At least it can encode more or less recognizable video. Since I have no intentions nor time to polish it, feel free to use it as you like - finish, rework into RPZA encoder, add link to this post to "Interesting patches" on Multimedia Wiki and forget it, whatever. (I should be working on AAC encoder anyway)

5 5

-newvideo stream order regression ?
by chris.stones＠gmail.com 02 Apr '11

02 Apr '11

# ffmpeg-SVN-r20373 -i INPUT_A (OPTIONS_A) OUTPUT -i /INPUTB (OPTIONS B) -newvideo would place INPUT_A with OPTIONS_A at stream index 0, and INPUT_B with OPTIONS_B at stream index 1. # ffmpeg-SVN-r22815 -i INPUT_A (OPTIONS_A) OUTPUT -i /INPUTB (OPTIONS B) -newvideo would place INPUT_B with OPTIONS_A at stream index 0, and INPUT_A with OPTIONS_B at stream index 1. The documentation doesnt imply stream orders, but the old behaviour makes alot more sence to me. Please confirm that this change was not intentional. I will investigate. Thanks. Chris.

2 1

Michael's ffmpeg.c patch - Favor streams with more packets if the user did not specify what she wants.
by chris.stones＠gmail.com 02 Apr '11

02 Apr '11

This patch http://git.ffmpeg.org/?p=ffmpeg;a=commitdiff;h=920cfe58c3137df7df8070bf3c3e… fixes an old issue ( https://roundup.ffmpeg.org/issue1156 ) But changes the behavior of the command line options, breaks my software ( and therefore probably other FFmpeg reliant software ). And in my opinion, makes the command line options more difficult to understand and error-prone. Take a look at the following command line, it encodes one video as Mpeg4, and one as SNOW. each with different qscales. ~/workspace/ffmpeg/ffmpeg -i /tmp/2snut_wJ6sAf/rgb%06d.png -strict -2 -vcodec snow -pix_fmt yuv420p -qscale 2.0 -f nut -y ~/test1.nut -i /tmp/2snut_wJ6sAf/a%06d.png -strict -2 -vcodec mpeg4 -pix_fmt yuv420p -qscale 6.0 -newvideo Before the above patch, it was easy to see that rgb%06d.png would be encoded with the SNOW codec, at quantize scale 2, and video a%06d.png would be encoded as an mpeg4, with a 6.0 quantize scale. However, after the patch, because order is determined by number of packets, it is impossible for the user to know what is going to be encoded with what settings / which codec. In the case of my sample media ( both videos are identical in frame number and resolution, but a%06d.png is much smaller. ) the a%06d.png video is encoded with snow, qscale 2. and a%06d.png is an mpeg4 at qscale 6. in other words, every script, everywhere is going to need to add "-map 0.0 -map 1.0 .... -map n.0" to the end of every "-newvideo" option. If this patch is not going to be un-done, then I think the documentation needs updating, as the implicit order of "-newvideo" nolonger is nolonger sane ? But thats just my opinion.. Any thoughts ???? THANKS Chris.

4 5

Fwd: framebuffer device demuxer
by giliarde＠gmail.com 12 Mar '11

12 Mar '11

Hi, patch attached. Giliard wrote: > Dear developer, > > Recently I added a extension in the FFmpeg for grab the screen content > of devices which use framebuffer ( see http://www.directfb.org/ ). > I implemented this as a framebuffer device demuxer in the directory > "libavdevice". > I needed modify the files: libavdevice/alldevices.c, > libavdevice/Makefile and configure. > All similar to x11grab. > > If you has interest for included in the FFmpeg source code official, the > attachment contain the files modified and implemented. > First, thanks a lot for your work. Would it be possible for you to send a patch ? (diff -wup) of your tree and send it attached in ffmpeg-devel mailing list ? Thanks a lot. -- Baptiste COUDURIER GnuPG Key Id: 0x5C1ABAAA Key fingerprint 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA FFmpeg maintainer http://www.ffmpeg.org

9 39

Small memleak when use AAC codec.
by igor＠plusmobile.ru 11 Mar '11

11 Mar '11

I found small memory leak when using aac. This leak is 32 bytes per loop without MEMALIGN_HACK, and 48 bytes with MEMALIGN. Simple code: #include <iostream> extern "C" { #include "libavformat/avformat.h" #include "libavcodec/avcodec.h" } AVOutputFormat *OutFormat; AVFormatContext *FormatCtx; AVCodecContext *VideoCodecCtx, *AudioCodecCtx; AVCodec *VideoCodec, *AudioCodec; AVStream *VideoStream, *AudioStream; int main (int argc, char * const argv[]) { av_register_all(); while(1) { FormatCtx = avformat_alloc_context(); OutFormat = guess_format(NULL, "test.3gp", NULL); FormatCtx->oformat = OutFormat; strncpy(FormatCtx->filename, "test.3gp", sizeof(FormatCtx->filename)); VideoStream = av_new_stream(FormatCtx, 0); VideoCodecCtx = VideoStream->codec; VideoCodecCtx->codec_id = CODEC_ID_H263; VideoCodecCtx->codec_type = CODEC_TYPE_VIDEO; VideoCodecCtx->bit_rate = 16000; VideoCodecCtx->width = 176; VideoCodecCtx->height = 144; VideoCodecCtx->time_base.den = 10; VideoCodecCtx->time_base.num = 1; VideoCodecCtx->pix_fmt = PIX_FMT_YUV420P; VideoCodecCtx->flags |= CODEC_FLAG_QSCALE; VideoCodecCtx->gop_size = 20; VideoCodecCtx->global_quality = (float)5 * (float)FF_QP2LAMBDA; VideoStream->quality = (float)5; VideoStream->r_frame_rate.den = 10; VideoStream->r_frame_rate.num = 1; VideoCodecCtx->flags |= CODEC_FLAG_GLOBAL_HEADER; VideoCodec = avcodec_find_encoder(VideoCodecCtx->codec_id); if(avcodec_open(VideoCodecCtx, VideoCodec) < 0) { printf("Fail open video\n"); return 1; } AudioStream = av_new_stream(FormatCtx, 1); AudioCodecCtx = AudioStream->codec; AudioCodecCtx->codec_id = CODEC_ID_AAC; AudioCodecCtx->codec_type = CODEC_TYPE_AUDIO; AudioCodecCtx->bit_rate = 32000; AudioCodecCtx->sample_rate = 8000; AudioCodecCtx->channels = 1; AudioCodecCtx->time_base= (AVRational){1, 8000}; AudioCodecCtx->sample_fmt = SAMPLE_FMT_S16; AudioCodecCtx->flags |= CODEC_FLAG_GLOBAL_HEADER; AudioCodec = avcodec_find_encoder(AudioCodecCtx->codec_id); if(avcodec_open(AudioCodecCtx, AudioCodec)) { printf("Fail open audio\n"); return 2; } url_fopen(&FormatCtx->pb, "test.3gp",URL_WRONLY); av_write_header(FormatCtx); AVStream *st; av_write_trailer(FormatCtx); url_fclose(FormatCtx->pb); avcodec_close(VideoStream->codec); avcodec_close(AudioStream->codec); for(uint32_t i=0;i< FormatCtx->nb_streams;i++) { st = FormatCtx->streams[i]; if (st->parser) { av_parser_close(st->parser); av_free_packet(&st->cur_pkt); } av_metadata_free(&st->metadata); av_free(st->index_entries); av_free(st->codec->extradata); av_free(st->codec); av_free(st->priv_data); av_free(st); } av_free(FormatCtx); usleep(3000); } return 0; } I tried to understand exactly where the memory is lost, and Founded that the functions av_cold void ff_psy_preprocess_end(struct FFPsyPreprocessContext *ctx) don't free context, so if add av_free(ctx); mem leak goes away. Patch attached.

2 1