July 2005 - ffmpeg-devel - lists.ffmpeg.org

jpeg to mpeg
by willem＠stucky-solutions.net 01 Aug '05

01 Aug '05

the last 3 week I have been struggling to convert my jpeg files from zone minder to mpeg movies, can any one help me please

3 2

Fix for overflow issue in mpegvideo.c
by boehme＠inb.uni-luebeck.de 01 Aug '05

01 Aug '05

Hi, the attached patch fixes an overflow issue I was getting for (overconservatively) large buffer sizes and large picture heights, i.e. buf_size=2.5 MB and end_y=1024. I'm aware this buf_size is _much_ too large (being lazy, I allocate enough buffer for a complete uncompressed image so I _know_ it will be enough)... OTOH, ffmpeg's (static) buffer size (256 KB) is already within an order of magnitude of the "critical" size. Comments, please... Martin -- Martin B?hme Inst. f. Neuro- and Bioinformatics Ratzeburger Allee 160, D-23538 Luebeck Phone: +49 451 500 5514 Fax: +49 451 500 5502 boehme at inb.uni-luebeck.de

5 9

Subject: -acodec 'copy' fails
by bastian.suess＠planet-interkom.de 31 Jul '05

31 Jul '05

Hi, I am trying to convert a multi-language VOB-file (from DVD) to a single-language VOB-file (only german). => ffmpeg crashes => my command line: > ffmpeg -i HARRY_POTTER_PHILOSOPHERS_STON1-1.vob \ -vcodec copy -acodec copy -f vob harryff13.vob -map 0.0 -map 0.3 => Output: ffmpeg version CVS, build 4759, Copyright (c) 2000-2004 Fabrice Bellard configuration: built on Jul 28 2005 20:06:50, gcc: 3.3.3 (SuSE Linux) Input #0, mpeg, from 'HARRY_POTTER_PHILOSOPHERS_STON1-1.vob': Duration: 00:42:39.7, start: 0.052000, bitrate: 6711 kb/s Stream #0.0[0x1e0]: Video: mpeg2video, yuv420p, 720x576, 25.00 fps, 7500 kb/s Stream #0.1[0x80]: Audio: 0x0000 Stream #0.2[0x82]: Audio: 0x0000 Stream #0.3[0x81]: Audio: 0x0000 Output #0, vob, to 'harryff13a.vob': Stream #0.0: Video: mpeg2video, 720x576, 25.00 fps, q=2-31, 7500 kb/s Stream #0.1: Audio: ac3 Stream mapping: Stream #0.0 -> #0.0 Stream #0.3 -> #0.1 Gleitkomma-Ausnahme (german translation of floating point execption) =>I found, that 1) The call-path " main() => av_encode() => av_write_header() => av_frac_init() " causes in av_frac_init() that st->time_base.num and st->codec->sample_rate are both zero ! So an Division by zero occures. 2) Something similar happens in " main() => av_encode() => output_packet() " ist->st->codec->sample_rate and ist->st->codec->channels are also both zero. Simple patching the code to avoid the division by zero keeps the progam running, but it works only with audio-channel-mapping to 0.1. Does everybody knows how to fix it the rigt way? I guess anywhere around Line 1700 in ffmpeg.c : "/* if stream_copy is selected..." Thanks!

2 1

Re: MPEG and ClosedCaption
by gamester17＠xboxmediacenter.com 31 Jul '05

31 Jul '05

Foushee, Carey-P55376 wrote: > Is it possible to pull out the closed caption text using FFMpeg? No (but it would be nice if we could), if you want to code a patch see: CC (Closed Captions) software decoding for hearing impaired: http://etvcookbook.org/reference/referenceplatform.html http://www.koders.com/c++/fidD5E1B8D838372D90B37658DE9CF5FAC65A378E45.aspx http://www.koders.com/c/fid120D7F025E56C431D7435F78CCE4D27F43AC21B5.aspx http://www.mplayerhq.hu/DOCS/HTML/en/subosd.html CC (Closed Captions) pass-through for hearing impaired: http://sourceforge.net/projects/betvout/ More info on Closed Captions (CC, line 21 data): http://www.robson.org/capfaq/technical.html FYI; A couple of Closed Caption FAQ's: Q: Are there captions on all DVDs? A: Standards were created very late for DVDs, so there are several different ways to put captions on them. The most common approaches are to either embed Line-21- style captions in what are called the "user bits" or to use "subpictures" to display captions that look like subtitles. Your DVD player must support the same method that was used to encode the DVD, although many DVDs use both of these methods. When using subtitles, DVDs support captioning in multiple languages. Q: What's the difference between subtitles and captions? A: Generally speaking, subtitles are intended for hearing audiences and captions for deaf audiences. Subtitles may translate the dialog into a different language, but rarely show all of the audio. For example, captions show sound effects (e.g., "phone ringing" and "footsteps"), while subtitles don't. Q: Why does my DVD show subtitles and captions simultaneously? A: In an effort to work with as much equipment as possible, some DVDs are encoded with both subtitles and closed captions. If your TV is set to show captions, and you turn on the DVD subtitles, you'll see the captions over top of the subtitles. If you're getting both, you'll have to either turn off subtitling on the DVD player, or turn off captions on the television set. Q: Can you use captions for anything else? A: Closed Captions pass-through signals are also used by third-party external CC readers such as TV Guardian and Cursefree TV to censor/filter audio/speach DVD's in TV (can auto mute the audio on programed CC signals), this does not work on normal subtitles. PS! I believe Closed Captions (CC, or line 21 data) is only broadly used in the U.S.A., is a simple feature that is included in most US (NTSC) DVD-Video movies (VOB files), and some DVD player pass-through that signal to a TV that supports CC the TV will decode and display them on demand. For the hearing impaired CC are a must to fully enjoy DVD movies as they display text of audio that going on that only the hearing can hear and the normal subtitles do not show. When using subtitles, DVDs support captioning in multiple languages. Regards / Gamester17

1 0

SIGFPE Arithmetic exception in h263.c
by dario＠sinistro.net 31 Jul '05

31 Jul '05

Another fatal error has occurred (which seems that the bitstream is corrupted in anyhow), but now in h263.c (read previous post on SIGSEGV in h264.c). Program received signal SIGFPE, Arithmetic exception. [Switching to thread 1292.0x220] 0x656ec78d in __udivmoddi4 () (gdb) bt #0 0x656ec78d in __udivmoddi4 () #1 0x656ec678 in __divdi3 () #2 0x6556c83b in decode_vop_header (s=0x3d209d0, gb=0x3d22544) at h263.c:5854 #3 0x65554c99 in ff_h263_decode_frame (avctx=0xe74680, data=0x4e1fd70, data_size=0x4e1fd6c, buf=0x625fdd0 "", buf_size=16906) at h263dec.c:493 #4 0x6548380a in avcodec_decode_video (avctx=0xe74680, picture=0x4e1fd70, got_picture_ptr=0x4e1fd6c, buf=0x625fdd0 "", buf_size=16906) at utils.c:747 #5 0x0061042f in ?? () #6 0x00660378 in ?? () #7 0x00661146 in ?? () #8 0x0065a885 in ?? () #9 0x00626e76 in ?? () #10 0x00626b3c in ?? () #11 0x7c80b50b in $R000000 () (gdb) list h263.c:5854 5849 s->pb_field_time= ( ROUNDED_DIV(s->time, s->t_frame) 5850 - ROUNDED_DIV(s->last_non_b_time - s->pp_time, s->t_frame))*2; 5851 } 5852 //av_log(s->avctx, AV_LOG_DEBUG, "last nonb %Ld last_base %d time %Ld pp %d pb %d t %d ppf %d pbf %d\n", s->last_non_b_time, s->last_time_base , s->time, s->pp_time, s->pb_time, s->t_frame, s->pp_field_time, s->pb_field_time); 5853 5854 s->current_picture_ptr->pts= (s->time + s->avctx->time_base.num/2) / s->avctx->time_base.num; 5855 if(s->avctx->debug&FF_DEBUG_PTS) 5856 av_log(s->avctx, AV_LOG_DEBUG, "MPEG4 PTS: %Ld\n", s->current_picture_ptr->pts); 5857 5858 check_marker(gb, "before vop_coded"); (gdb) disass $pc-32 $pc+32 Dump of assembler code from 0x656ec76d to 0x656ec7ad: 0x656ec76d <__udivmoddi4+125>: rorb 0x4589e875(%ecx) 0x656ec773 <__udivmoddi4+131>: in (%dx),%al 0x656ec774 <__udivmoddi4+132>: mov 0xffffffec(%ebp),%edx 0x656ec777 <__udivmoddi4+135>: mov 0xffffffe8(%ebp),%eax 0x656ec77a <__udivmoddi4+138>: add $0x2c,%esp 0x656ec77d <__udivmoddi4+141>: pop %ebx 0x656ec77e <__udivmoddi4+142>: pop %esi 0x656ec77f <__udivmoddi4+143>: pop %edi 0x656ec780 <__udivmoddi4+144>: pop %ebp 0x656ec781 <__udivmoddi4+145>: ret 0x656ec782 <__udivmoddi4+146>: test %eax,%eax 0x656ec784 <__udivmoddi4+148>: jne 0x656ec791 <__udivmoddi4+161> 0x656ec786 <__udivmoddi4+150>: mov $0x1,%eax 0x656ec78b <__udivmoddi4+155>: xor %edx,%edx 0x656ec78d <__udivmoddi4+157>: div %esi 0x656ec78f <__udivmoddi4+159>: mov %eax,%esi 0x656ec791 <__udivmoddi4+161>: mov %edi,%eax 0x656ec793 <__udivmoddi4+163>: mov 0xffffffdc(%ebp),%edx 0x656ec796 <__udivmoddi4+166>: div %esi 0x656ec798 <__udivmoddi4+168>: mov %eax,0xffffffd0(%ebp) 0x656ec79b <__udivmoddi4+171>: mov 0xffffffd8(%ebp),%eax 0x656ec79e <__udivmoddi4+174>: div %esi ---Type <return> to continue, or q <return> to quit--- 0x656ec7a0 <__udivmoddi4+176>: mov %edx,0xffffffd8(%ebp) 0x656ec7a3 <__udivmoddi4+179>: mov %eax,0xffffffd4(%ebp) 0x656ec7a6 <__udivmoddi4+182>: jmp 0x656ec746 <__udivmoddi4+86> 0x656ec7a8 <__udivmoddi4+184>: cmp %ebx,0xffffffdc(%ebp) 0x656ec7ab <__udivmoddi4+187>: jbe 0x656ec7d0 <__udivmoddi4+224> End of assembler dump. (gdb) info all-registers eax 0x1 1 ecx 0x9 9 edx 0x0 0 ebx 0x0 0 esp 0x4e1f96c 0x4e1f96c ebp 0x4e1f9a4 0x4e1f9a4 esi 0x0 0 edi 0x0 0 eip 0x656ec78d 0x656ec78d eflags 0x10246 66118 cs 0x1b 27 ss 0x23 35 ds 0x23 35 es 0x23 35 fs 0x3b 59 gs 0x0 0 st0 -nan(0x8061816380667e65) (raw 0xffff8061816380667e65) st1 -nan(0x0817e7f81) (raw 0xffff00000000817e7f81) st2 -nan(0x080808080) (raw 0xffff0000000080808080) st3 -nan(0x08081807e) (raw 0xffff000000008081807e) st4 -nan(0x80647f6480648164) (raw 0xffff80647f6480648164) st5 -nan(0x810080007e007d) (raw 0xffff00810080007e007d) st6 -nan(0x7e007e007e007e) (raw 0xffff007e007e007e007e) ---Type <return> to continue, or q <return> to quit--- st7 1 (raw 0x3fff8000000000000000) fctrl 0xffff027f -64897 fstat 0xffff0000 -65536 ftag 0xffffffff -1 fiseg 0x1b 27 fioff 0x7483216f 1954750831 foseg 0xffff0023 -65501 fooff 0x3d25c20 64117792 fop 0x4cc 1228 (gdb) info frame Stack level 0, frame at 0x4e1f9a4: eip = 0x656ec78d in __udivmoddi4; saved eip 0x656ec678 called by frame at 0x4e1fa04 Arglist at 0x4e1f9a4, args: Locals at 0x4e1f9a4, Previous frame's sp is 0x0 Saved registers: ebx at 0x4e1f998, ebp at 0x4e1f9a4, esi at 0x4e1f99c, edi at 0x4e1f9a0, eip at 0x4e1f9a8 Is there any hint for me to find out what could be going on? i.e. corrupted packet, packet loss, missing vopheader, anythiing! ;-) Thanks a lot again, Dario Andrade Executive Director DATSCOM Mobile +55.21.9453.5005 Office +55.21.2141.9525

3 3

[PATCH] Changing the behaviour of make with default target
by eclipse7＠gmx.net 31 Jul '05

31 Jul '05

Hi, problems with linking the shared libs with m$ toolchain on windows keep coming up again and again. Maybe I am wrong, as I didn't get feedback yet. I for sure know it worked for me, as i had to use libav* with m$ toolchain one time not so long ago. I think part of the problem is that when configured with --enable-shared the static libs and the shared libs get built when using make with default target. The problem on windows is that you use shared libs with an import lib. So what happens normally is that the avcodec.lib is built which is the just static lib and the dll and a def file and, with proper environment setup including m$ lib.exe, also the import lib is built (which is named avcodec.lib and overwrites the static lib). But in case the environment isn't setup properly (no lib.exe in path) the import lib doens't get build and people seem to confuse it with the static lib which doesn't work properly when used with m$ tools. Despite the problem that most of them most probably didn't want to use the static version but the shared one. Make doesn't error out, because the rule is written to silently ignore an error at that place. Maybe because it assumes that many people don't use m$ tools, which would make sense to me. So maybe it would be better to just build the shared lib only by default when configured with --enable-shared. Then ppl would notice there is something missing. This is what the attached patch does. There is still the problem with filename clash tho, but i don't think it will still be that problematic as it is now. People knowing what they are doing will still be able to build both libs. Also a different approach would be to print a big fat warning for m$ devel tool users if the import lib could not be generated. Comments welcome... Alex (beastd)

2 3

ffmpeg crashes on windows
by tomillo1235＠yahoo.es 30 Jul '05

30 Jul '05

Hi, We are building project in VC .net 2003, and we have problems while running the code on Windows. We have called a function avcodec_close() which is causing the error. When we execute our code, the error which we are getting is access violation, no output. We have compiled the .dll and .lib with MinGW and MSys: Commands used: ./configure --enable-memalign-hack --enable-shared --disable-strip make We have renamed libgcc.a as libgcc.lib and linked to our proyect in order to avoid linking problems. also EMULATE_INTTYPES is #defined. We are compiling the last cvs version. Can you suggest solution on this problem Thanks in advance Juan Ramos ______________________________________________ Renovamos el Correo Yahoo! Nuevos servicios, m?s seguridad http://correo.yahoo.es

2 1

Re: I want to resize getted frame from movie file.
by ffmpeg＠gmail.com 30 Jul '05

30 Jul '05

Dear: > > I want to watch fixed display size from movie file. (eg. 640x480) > I'm trying to use img_resample function. > I don't know what i do before calling img_resample_init function. > something following line. > > if(packet.stream_index == m_nVideoStream) > { > // decoding next frame. > bytesDecoded = avcodec_decode_video(m_pCodecCtx, m_pFrame, &frameFinished, > packet.data, packet.size); > > if(frameFinished) > { > img_convert((AVPicture *)m_pFrameRGB, PIX_FMT_RGB24, (AVPicture*)m_pFrame, > m_pCodecCtx->pix_fmt, m_pCodecCtx->width, m_pCodecCtx->height); // m_pFrame > has YUV frame > ImgReSampleContext *s; > s = img_resample_init( 640, 480, width, height); // initialize resample you'd better use img_resample_full_init(...) to initialize the resample context ! > img_resample(s, m_pic, (AVPicture *)m_pFrameRGB); // error occur How do you initialize the "m_pic"? you should fill "m_pic" with yuv420(640x480) layout buffer firstly There is an important note that img_resample(...) is only happy with YUV420 pix format,but not RGB > img_resample_close(s); > Jeon,Chang-Min > Thanks. Take care!

1 0

MPEG and ClosedCaption
by Carey.Foushee＠gdc4s.com 29 Jul '05

29 Jul '05

Is it possible to pull out the closed caption text using FFMpeg?

1 0

Segmentation fault on h264.c
by dario＠ip.tv 29 Jul '05

29 Jul '05

Hi, Unfortunatelly I don't have a saved bitstream somewhere neither I am working with ffmpeg utility, instead this came from a live decoding using lavc api directly. The exact line that it happened was (marked with an arrow): if(cur->reference == 0) cur->reference = 1; for(i=0; h->delayed_pic[i]; i++) if(h->delayed_pic[i]->key_frame || h->delayed_pic[i]->poc==0) <------- (*) cross_idr = 1; out = h->delayed_pic[0]; for(i=1; h->delayed_pic[i] && !h->delayed_pic[i]->key_frame; i++) It does happen on high latency (perhaps high packet loss) transmissions. So my guess would be that it is referencing an invalid 'delayed_pic[i]" member. Here are the extra info: Program received signal SIGSEGV, Segmentation fault. [Switching to thread 1688.0xac8] decode_frame (avctx=0xe56c60, data=0x4fffd70, data_size=0x4fffd6c, buf=0x61dfde4 "", buf_size=381) at h264.c:7478 7478 if(h->delayed_pic[i]->key_frame || h->delayed_pic[i]->poc==0) (gdb) info line Line 7478 of "h264.c" starts at address 0x655a04e6 <decode_frame+470> and ends at 0x655a04f7 <decode_frame+487>. (gdb) bt #0 decode_frame (avctx=0xe56c60, data=0x4fffd70, data_size=0x4fffd6c, buf=0x61dfde4 "", buf_size=381) at h264.c:7478 #1 0x6548380a in avcodec_decode_video (avctx=0xe56c60, picture=0x4fffd70, got_picture_ptr=0x4fffd6c, buf=0x61dfde4 "", buf_size=381) at utils.c:747 #2 0x0061042f in ?? () #3 0x00660378 in ?? () #4 0x00661146 in ?? () #5 0x0065a885 in ?? () #6 0x00626e76 in ?? () #7 0x00626b3c in ?? () #8 0x7c80b50b in $R000000 () (gdb) disass $pc-32 $pc+32 Dump of assembler code from 0x655a04c6 to 0x655a0506: 0x655a04c6 <decode_frame+438>: out %al,(%dx) 0x655a04c7 <decode_frame+439>: add %eax,(%eax) 0x655a04c9 <decode_frame+441>: inc %esi 0x655a04ca <decode_frame+442>: mov %esi,0xffffffd8(%ebp) 0x655a04cd <decode_frame+445>: test %ecx,%ecx 0x655a04cf <decode_frame+447>: jne 0x655a04d8 <decode_frame+456> 0x655a04d1 <decode_frame+449>: movl $0x1,0x50(%eax) 0x655a04d8 <decode_frame+456>: mov 0x1ee18(%ebx),%eax 0x655a04de <decode_frame+462>: xor %ecx,%ecx 0x655a04e0 <decode_frame+464>: test %eax,%eax 0x655a04e2 <decode_frame+466>: je 0x655a050a <decode_frame+506> 0x655a04e4 <decode_frame+468>: mov %eax,%esi 0x655a04e6 <decode_frame+470>: mov 0x30(%esi),%edx 0x655a04e9 <decode_frame+473>: test %edx,%edx 0x655a04eb <decode_frame+475>: jne 0x655a04f7 <decode_frame+487> 0x655a04ed <decode_frame+477>: mov 0xe4(%esi),%edi 0x655a04f3 <decode_frame+483>: test %edi,%edi 0x655a04f5 <decode_frame+485>: jne 0x655a04fe <decode_frame+494> 0x655a04f7 <decode_frame+487>: movl $0x1,0xffffffd4(%ebp) 0x655a04fe <decode_frame+494>: inc %ecx 0x655a04ff <decode_frame+495>: mov 0x1ee18(%ebx,%ecx,4),%esi End of assembler dump. (gdb) info all-registers eax 0x3e11860 65083488 ecx 0x11 17 edx 0x0 0 ebx 0x41f5b50 69163856 esp 0x4fffa9c 0x4fffa9c ebp 0x4fffad4 0x4fffad4 esi 0x1 1 edi 0x6 6 eip 0x655a04e6 0x655a04e6 eflags 0x210202 2163202 cs 0x1b 27 ss 0x23 35 ds 0x23 35 es 0x23 35 fs 0x3b 59 gs 0x0 0 st0 -nan(0x8181818181818181) (raw 0xffff8181818181818181) st1 -nan(0x8181818181818181) (raw 0xffff8181818181818181) st2 -nan(0x818181818080807f) (raw 0xffff818181818080807f) st3 -nan(0x81008100810081) (raw 0xffff0081008100810081) st4 0 (raw 0xffff0000000000000000) st5 0 (raw 0xffff0000000000000000) st6 0 (raw 0xffff0000000000000000) ---Type <return> to continue, or q <return> to quit--- st7 0 (raw 0xffff0000000000000000) fctrl 0xffff027f -64897 fstat 0xffff0000 -65536 ftag 0xffffffff -1 fiseg 0x8 8 fioff 0xbfa333f4 -1079823372 foseg 0xffff0010 -65520 fooff 0xee77ebd0 -294130736 fop 0x475 1141 (gdb) info frame Stack level 0, frame at 0x4fffad4: eip = 0x655a04e6 in decode_frame (h264.c:7478); saved eip 0x6548380a called by frame at 0x4fffb14 source language c. Arglist at 0x4fffad4, args: avctx=0xe56c60, data=0x4fffd70, data_size=0x4fffd6c, buf=0x61dfde4 "", buf_size=381 Locals at 0x4fffad4, Previous frame's sp is 0x0 Saved registers: ebx at 0x4fffac8, ebp at 0x4fffad4, esi at 0x4fffacc, edi at 0x4fffad0, eip at 0x4fffad8 (gdb) info locals out = (Picture *) 0xcccccccc cur = (Picture *) 0x3e12eb0 prev = (Picture *) 0x3e12d18 out_idx = 0 pics = 16 cross_idr = 0 dropped_frame = 0 s = (MpegEncContext *) 0x41f5b50 buf_index = 381 buf_index = 381 (gdb) print i $1 = 1 (gdb) print h $2 = (H264Context *) 0x3e11860 (gdb) print h->delayed_pic[1] $3 = (Picture *) 0x80808080 (INVALID!) (gdb) print h->delayed_pic[0] $4 = (Picture *) 0x80808080 (gdb) print h->delayed_pic $5 = {0x80808080, 0x80808080, 0x80808080, 0x7f7f7f7f, 0x7f7f7f7f, 0x7f7f7f7f, 0x81818181, 0x81818181, 0x797c8181, 0x84756f78, 0x80807f82, 0x786c5c7a, 0x7a7b7b7b, 0x7c7b7a79, 0x7e7e7e7d, 0x7d7d7e7e} (gdb) print cur $6 = (Picture *) 0x3e12eb0 (gdb) print cur->reference $7 = 3 (gdb) print s->current_picture_ptr $8 = (Picture *) 0x3e12eb0 (gdb) print h->delayed_output_pic $9 = (Picture *) 0x7e7d7d7c That segmentation fault is happenning very often. Thanks a lot, -- Dario Andrade Executive Director DATSCOM Mobile +55.21.9453.5005 Office +55.21.2141.9525

1 0