December 2007 - ffmpeg-devel - lists.ffmpeg.org

Bug in av_open_input_stream() since r11071 (ByteIOContext change) [3rd try to send]
by stefan＠lucke.in-berlin.de 22 Dec '07

22 Dec '07

Old version in libavformat/utils.c from r11007 looks like: int av_open_input_stream(AVFormatContext **ic_ptr, ? ? ? ? ? ? ? ? ? ? ? ? ?ByteIOContext *pb, const char *filename, ? ? ? ? ? ? ? ? ? ? ? ? ?AVInputFormat *fmt, AVFormatParameters *ap) { ? ? int err; ? ? AVFormatContext *ic; ? ? AVFormatParameters default_ap; ? ? if(!ap){ ? ? ? ? ap=&default_ap; ? ? ? ? memset(ap, 0, sizeof(default_ap)); ? ? } ? ? if(!ap->prealloced_context) ? ? ? ? ic = av_alloc_format_context(); ? ? else ? ? ? ? ic = *ic_ptr; ? ? if (!ic) { ? ? ? ? err = AVERROR(ENOMEM); ? ? ? ? goto fail; ? ? } ? ? ic->iformat = fmt; ? ? if (pb) ? ? ? ? ic->pb = *pb; This instruction was a structure-copy if an address of a ByteIOContext is supplied. ic's own ByteIOContext was initialized in av_alloc_format_context(); In current svn version (r11278), supplied parameter pointer gets ic's pb independent of it's value. line 377: ic->pb = pb; It may be NULL in case current format has AVFMT_NOFILE set. Attached patch fixes this issue. Stefan Lucke

2 2

[ffmpeg-devel] Cross compiling to windows CE
by robertogarridomartin＠gmail.com 22 Dec '07

22 Dec '07

Hi there! I'm trying to build ffmpeg on the Windows CE platform. I'm following some threads in the Iternet, but none of them seem to have been solved... I've downloaded the CeGCC compiler and I put it in my mSYS directory. When I try to configure ffmpeg with this command: ./configure --cross-compile \ --cross-prefix=/opt/cegcc/arm-wince-cegcc/bin/ --arch=armv4l \ --enable-memalign-hack --disable-ipv6 --prefix=/usr/local/wince \ --log=yes --disable-encoders --disable-decoders --disable-parsers \ --disable-muxers --disable-demuxers --enable-parser=h263 \ --enable-encoder=h263 --enable-decoder=h263 --enable-muxer=h263 \ --enable-demuxer=h263 --disable-ffserver --disable-ffplay \ --disable-ffmpeg --disable-network I got the next message: /opt/cegcc/arm-wince-cegcc/bin/gcc is unable to create an executable file. C compiler test failed. If you think configure made a mistake, make sure you are using the latest version from SVN. If the latest version fails, report the problem to the ffmpeg-devel at mplayerhq.hu mailing list or IRC #ffmpeg on irc.freenode.net . Include the log file "config.err" produced by configure as this will help solving the problem. Anyone knows anything about it? Thanks a lot in advance, Roberto

3 5

Possible violation of FFmpeg's license by Transgaming's Cider
by flameeyes＠gmail.com 22 Dec '07

22 Dec '07

Hi, I stumbled through this yesterday, it might be of interest to you guys, but I'm not a lawyer, so it might be all right and just being me thinking negatively. I bought a copy of Command & Conquer 3 for Mac; the game has been ported to Mac by Transgaming (of Cedega fame), through their Cider toolkit (which is probably just based on Cedega itself, although I admit I don't know enough of that). On the game's manual, there is a list of copyright holders, many Free Software projects are listed, including libpng, libjpeg, cg toolkit, dmalloc, SDL, bstring, iniparser, Squish of course ReWind and Wine, but also FFmpeg: "The Cider libquartz.dylib component includes portions of ffmpeg, Copyright (C) 2000-2006 Fabrice Bellard, et al." I do have libquartz.dylib on the DVD, but I can't find any copy of the LGPL license nor the corresponding source code on either the DVD or the manual, nor I find a written offer to send me the source code. The only reference to source code is at the end of the page: "Source code to the LGPL components is available via CVS access through: http:/transgaming.org/cvs/". I'm not sure if that is enough to be compliant with LGPL license, as if I recall correctly you need to provide the source code, or a written offer, together with the binary form. I wonder if I should also submit this to the other projects involved. -- Diego "Flameeyes" Petten? http://farragut.flameeyes.is-a-geek.org/

6 5

[PATCH] don't set is_streamed when it's not
by aurel＠gnuage.org 22 Dec '07

22 Dec '07

Hi, Currently ffserver uses url_open_buf() and url_open_dyn_buf() and then set is_streamed in the resulting ByteIOContext. This seems plain wrong. Buffers are really not streamed. Attached patch avoid this. OK ? Aurel

7 18

ASF EOF detection
by design4mmedia＠gmail.com 22 Dec '07

22 Dec '07

When end of ASF file is reached, only AVERROR_IO returned, with no indication of EOF in ByteIOContext.eof_reached. Is the mod given below the correct place to add this detection: static int asf_read_packet(AVFormatContext *s, AVPacket *pkt) ............ asf->packet_pos= url_ftell(&s->pb); if (asf->data_object_size != (uint64_t)-1 && (asf->packet_pos - asf->data_object_offset >= asf->data_object_size)) { pb->eof_reached = 1; return AVERROR_IO; /* Do not exceed the size of the data object */ } Thanks, Vinod.

2 2

Bug in av_open_input_stream() since r11071 (ByteIOContext change) [2nd try to send]
by stefan＠lucke.in-berlin.de 22 Dec '07

22 Dec '07

old version from r11007 looks like: int av_open_input_stream(AVFormatContext **ic_ptr, ByteIOContext *pb, const char *filename, AVInputFormat *fmt, AVFormatParameters *ap) { int err; AVFormatContext *ic; AVFormatParameters default_ap; if(!ap){ ap=&default_ap; memset(ap, 0, sizeof(default_ap)); } if(!ap->prealloced_context) ic = av_alloc_format_context(); else ic = *ic_ptr; if (!ic) { err = AVERROR(ENOMEM); goto fail; } ic->iformat = fmt; if (pb) ic->pb = *pb; This instruction was a structure-copy if an address of a ByteIOContext is supplied. ic's own ByteIOContext was initialized in av_alloc_format_context(); In current svn version (r11278), supplied parameter pointer gets ic's pb despite of it's value. line 377: ic->pb = pb; It may be NULL in case current format has AVFMT_NOFILE set. Attached patch fixes this issue. Stefan Lucke

1 0

MMX version for put_no_rnd_h264_chroma_mc8_c
by christophe.gisquet＠free.fr 21 Dec '07

21 Dec '07

Hello, from the opreport log concerning Robotica_720.wmv as seen in the previous thread, it was noticeable that put_no_rnd_h264_chroma_mc8_c was taking around 10%. I observed the same phenomenon on another "WMV HD" sample (Amazing Caves 720p). However, the few AVC samples from mphq I tested didn't seem to use it, so it probably of lesser importance for this codec. So here is a patch to provide a MMX implementation of it. I also had a MMX2 implementation (using pshufw where useful), but the performance difference of the functions (~2%, measured with START/STOP_TIMER) wasn't worth increasing object size. I'm fine with the license (LGPL 2.1 or any version) of h264dsp_mmx.c where most of the code is located. Best regards, -- Christophe GISQUET

7 19

Bug in av_open_input_stream() since r11071 (ByteIOContext change)
by stefan＠lucke.in-berlin.de 21 Dec '07

21 Dec '07

old version from r11007 looks like: int av_open_input_stream(AVFormatContext **ic_ptr, ByteIOContext *pb, const char *filename, AVInputFormat *fmt, AVFormatParameters *ap) { int err; AVFormatContext *ic; AVFormatParameters default_ap; if(!ap){ ap=&default_ap; memset(ap, 0, sizeof(default_ap)); } if(!ap->prealloced_context) ic = av_alloc_format_context(); else ic = *ic_ptr; if (!ic) { err = AVERROR(ENOMEM); goto fail; } ic->iformat = fmt; if (pb) ic->pb = *pb; This instruction was a structure-copy if an address of a ByteIOContext is supplied. ic's own ByteIOContext was initialized in av_alloc_format_context(); In current svn version (r11278), supplied parameter pointer gets ic's pb despite of it's value. line 377: ic->pb = pb; It may be NULL in case current format has AVFMT_NOFILE set. Attached patch fixes this issue. Stefan Lucke

1 0

[PATCH] Fixes compile error on my AMD 64 running linux
by iorrus＠yahoo.co.uk 21 Dec '07

21 Dec '07

Hello, FFMpeg won't compile for me without this patch. I found this on http://groups.google.com/group/lucky.freebsd.ports/browse_thread/thread/884… I get the following errors. {standard input}:45598: Error: ff_h264_lps_range(%eax,%esi,2)' is not a valid 6 4 bit base/index expression {standard input}:45611: Error: `ff_h264_norm_shift(%esi)' is not a valid 64 bit base/index expression {standard input}:45613: Error: `ff_h264_mlps_state+128(%eax)' is not a valid 64 bit base/index expression {standard input}:45625: Error: `-1(%ebx)' is not a valid 64 bit base/index expre ssion {standard input}:45628: Error: `ff_h264_norm_shift(%ecx)' is not a valid 64 bit base/index expression {standard input}:45650: Error: `ff_h264_lps_range(%eax,%esi,2)' is not a valid 6 4 bit base/index expression {standard input}:45663: Error: `ff_h264_norm_shift(%esi)' is not a valid 64 bit base/index expression {standard input}:45665: Error: `ff_h264_mlps_state+128(%eax)' is not a valid 64 bit base/index expression {standard input}:45677: Error: `-1(%ebx)' is not a valid 64 bit base/index expre ssion {standard input}:45680: Error: `ff_h264_norm_shift(%ecx)' is not a valid 64 bit base/index expression {standard input}:45891: Error: `ff_h264_lps_range(%edx,%esi,2)' is not a valid 6 4 bit base/index expression {standard input}:45904: Error: `ff_h264_norm_shift(%esi)' is not a valid 64 bit base/index expression {standard input}:45906: Error: `ff_h264_mlps_state+128(%edx)' is not a valid 64 bit base/index expression {standard input}:45918: Error: `-1(%ebx)' is not a valid 64 bit base/index expre ssion {standard input}:45921: Error: `ff_h264_norm_shift(%ecx)' is not a valid 64 bit base/index expression {standard input}:45930: Error: `last_coeff_flag_offset_8x8(%edi)' is not a valid 64 bit base/index expression {standard input}:45935: Error: `ff_h264_lps_range(%edx,%esi,2)' is not a valid 6 4 bit base/index expression {standard input}:45948: Error: `ff_h264_norm_shift(%esi)' is not a valid 64 bit base/index expression {standard input}:45950: Error: `ff_h264_mlps_state+128(%edx)' is not a valid 64 bit base/index expression {standard input}:45962: Error: `-1(%ebx)' is not a valid 64 bit base/index expre ssion {standard input}:45965: Error: `ff_h264_norm_shift(%ecx)' is not a valid 64 bit base/index expression make: *** [h264.o] Error 1 ___________________________________________________________ Support the World Aids Awareness campaign this month with Yahoo! For Good http://uk.promotions.yahoo.com/forgood/

2 1

Detecting the size of a video before encoding it...
by ferrangil＠gmail.com 21 Dec '07

21 Dec '07

Hi all, I'm encoding lots of videos for a website, using Drupal, and I'm encoding each video to the original resolution (then, the flash player fits the video...) The problem is that I want to use the -s param just when the original video size is, lets say, 1280 pixels, but not when it's 320x240.... The idea is to just change the resolution for the really big videos, like limiting the video size to 848 or 720 px... I'm not sure if it's the correct mailing list. Any suggestion will be welcomed! Thanks ferran

2 1