January 2008 - ffmpeg-devel - lists.ffmpeg.org

New project using ffmpeg
by diego＠biurrun.de 23 Feb '08

23 Feb '08

On Fri, Jan 04, 2008 at 12:42:08PM +0100, Cezary Falba wrote: > On 23 Dec 2007, at 22:14, Diego Biurrun wrote: > >> On Wed, Dec 19, 2007 at 12:25:52PM +0000, Cezary Falba wrote: >>> >>> On 18 Dec 2007, at 17:34, Diego Biurrun wrote: >>> >>>> On Tue, Dec 18, 2007 at 11:43:36AM +0000, Cezary Falba wrote: >>>>> >>>>> On 18 Dec 2007, at 08:20, Diego Biurrun wrote: >>>>>> Also, you claim that your software is available under the GPL, >>>>>> but when I download the source code, there is no sign of the GPL >>>>>> anywhere and all files are marked as copyrighted by you or Saffron >>>>>> Digital or lack any copyright or license information. This makes >>>>>> them effectively non- free. >>>>>> >>>>>> Also, there is the following statement on your website: >>>>>> >>>>>> By downloading our software you accept the GPL licence which can be >>>>>> found here. >>>>>> >>>>>> This makes me think that you still haven't read the GPL, since it >>>>>> contains a paragraph about the need to accept the license that this >>>>>> conflicts with. >>>>>> >>>>>> Releasing your software under a license you have not read is >>>>>> tantamount to buying a house/car/whatever without reading the >>>>>> contract. I'll contact you in private with some excellent Brooklyn >>>>>> bridges that I have for sale at premium conditions... >>>>> >>>>> Thanks a lot for checking a site and all advices. >>>>> I just fixed a website and added a license to both sources and >>>>> installers. >>>>> It will be highly appreciated if you could re-check that page again >>>>> when you have few minutes. >>>>> There is no rush as we still have time before we officially release it >>>>> to public. >>>> >>>> I have trouble testing the Mac version since it now seems to require OS >>>> X 10.5, but I'm still on 10.4. Your website claims that only 10.4 is >>>> required.. >>>> >>>> You have added LGPL copyright headers to your source files, but the >>>> license text you include is the GPL. This looks like an oversight from >>>> you. Also, your copyright headers only contain two of the three >>>> paragraphs commonly used for this purpose. >>>> >>>> Are you using FFmpeg in GPL mode? >>>> >>>> I'm being told that the Windows installer of your program is totally >>>> broken and installs stuff into C: instead of the path it claims to >>>> install to. It also displays a license different from the (L)GPL. >>>> >>>> The Windows source package contains a __MACOSX directory, weird. >>>> >>>> I still find it awkward that the first mentions of FFmpeg, AviSynth and >>>> Perl on the above website are not links to homepages. Instead you hide >>>> those links at the bottom of the page. >>>> >>>> So it's getting better, but you're still not at 100%. >>> >>> Good to hear from you that we are get better with that. >>> I just made some tweaks to installers and website. >>> I also replaced license files with proper LGPL... >> >> Well, we are moving sideways, rather than forward. Here is a list of >> issues, some of which are old: >> >> - The Windows source code link gives "403 Permission denied". >> >> - The OS X installer claims to work on system 10.4.6 or later. When I >> try to install on 10.4.11 I get a message that "10.4.6 or later is >> required", well I thought that 11 > 6, but oh well ... >> >> - You mark your source files as LGPL 2.1 or later, but the license file >> you include and link to is the LGPL version 3. What is it going to >> be? >> >> - You do not provide FFmpeg sources for download. A pointer to our >> homepage is not enough. You need to give people the exact source you >> are using. >> >> - In the paragraph >> Saffron Master Encoder is our in house encoding utility made on top >> of open source projects such as: FFmpeg, AviSynth and Perl that >> prepares your source video to be used with our mobile delivery >> platform. >> the words FFmpeg, AviSynth and Perl should be clickable links that >> lead to each project's homepage. You have put links below that >> paragraph, but this feels awkward and FFmpeg is missing. >> This is a minor issue, but it would be nice if you could fix it. >> >> >> Overall I have to say this is starting to feel a bit like free >> consulting work for your company. I'm under the impression that you >> still have not read the licenses or its FAQs yourself. Instead you have >> me go over the results of your tries to comply with the license. I'm >> not seeing you put in the necessary effort to amend the situation. This >> is frustrating. > > I've fixed OS X installer, replaced licenses in both installers and source > codes and fixed web site. Now there are links in paragraph you pointed. The OS X installer is not fixed, it still refuses to run on my 10.4.11 system, claiming that "10.4.6 or later is required". Anyway, others tell me that your Windows installer does display the LGPL now. However, there is no way to uninstall your program. While this is entirely your problem, don't expect us to recommend your program to anyone this way. > I've also added information about getting FFmpeg source from SVN. This is not enough. I thought I was clear about this in the paragraph above. You need to provide complete FFmpeg source code including any modifications you did. Sending people to our homepage is not enough. When I download the Mac OS X source code and try to build it in xcode, the build fails because the ffmpeg executable is missing. > And I'm very sorry you got that feeling. I fully understand and apologize. > I don't want to bother you every time I do some change to the site but me > and my bosses really want to be all right with FFmpeg and other guys > whose work we use in this project. I'll be honest and blunt now: I do not want your apologies, I want you to work harder on this and not use me as free consulting labor. I pointed out 4 issues (plus one minor) above, you fixed two. If you and your bosses really want to be all right with us and use open source code while respecting copyrights and not breaking the law, you need to take this much more seriously. In one of my mails I asked you whether you used FFmpeg in GPL mode or not. You never answered my question. I looked myself, this is the output of the Windows version you use: FFmpeg version SVN-r8797, Copyright (c) 2000-2007 Fabrice Bellard, et al. configuration: --enable-gpl --enable-pp --enable-swscaler --enable-pthreads --enable-liba52 --enable-avisynth --enable-libdts --enable-libfaac --enable-libfaad --enable-libgsm --enable-libmp3lame --enable-libnut --enable-libogg --enable-libtheora --enable-libvorbis --enable-x264 --enable-xvid --enable-amr_nb --enable-amr_wb --cpu=i686 --enable-memalign-hack --extra-ldflags=-static So you are using it in GPL mode, but what is more important is that you also compile with libamr support, which is non-free and not redistributable. As long as you keep distributing this binary you are violating the copyright of 3GPP. I'm going to make you a numbered list of issues and questions, maybe this way you will find it easier to address all of them. 0. You are using libamr. That's entirely your problem, but you are in breach of copyright. 1. You are still not providing FFmpeg source code. You need to provide the exact source code you are using, including any modifications. Pointing at our homepage or Subversion repo is *not* enough. 2. If you continue using FFmpeg in GPL mode, the license of your program becomes the GPL. You have to update your licenses accordingly. 3. The OS X installer is broken, at least on OS X 10.4.11, I have no way to verify your claims that you fixed its other issues. 4. Have you read the LGPL? The GPL? The GPL FAQ? best regards Diego P.S.: Please keep the CC, restoring it everytime is getting annoying.

2 8

[RFC] [PATCH] Make ModeAlphabet constant in vp3.c
by flameeyes＠gmail.com 17 Feb '08

17 Feb '08

The attached patch makes ModeAlphabet constant, by removing the custom mode from the top, and using a static variable array inside the unpack_modes instead for custom modes. This way the big part of ModeAlphabet can be moved into .rodata. I admit I'm not sure if this is good myself (hence the RFC).

10 39

[PATCH] Fix for lowres decoding support on ARM
by siarhei.siamashka＠gmail.com 16 Feb '08

16 Feb '08

Hello, Video decoding with 'lowres' option enabled is currently broken for ARM. A patch to fix this problem is attached. Apparently lowres decoding is also broken when using 'alpha', 'bfin', 'mlib', 'ps2', 'sh4' and 'sparc' optimizations (every platform except for 'i386' and 'ppc'). Also the existence of this problem shows that lowres decoding is not properly covered by regression tests.

3 6

[PATCH] Merge some computations in C code for VC-1 inverse transforms
by christophe.gisquet＠free.fr 16 Feb '08

16 Feb '08

Hi, M. Niedermayer highlighted some sub-optimality in mail http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2008-January/040591.html These sub-optimalities can also be seen in the C version of those inverse transforms. Testing on IA32, it seems that: - performing the butterfly a,b->a-b,a+b with the no-additional-register trick never works - adding the bias constant at the first butterflies hardly ever works (unless there's register pressure from the looks of where it works best) - merging some computation (in the 4x1 or 1x4 parts) is always a win None of them works with any stage of the 8x8 inverse transform. Here are some benches with *_TIMER: bef. after 4x4: 2329 2100 4x8: 4872 4448 8x4: 4926 4267 So very roughly, a 10% improvement (YMMV). Best regards, Christophe GISQUET

4 7

[RFC] replace some static with asm_visibility or so
by Reimar.Doeffinger＠stud.uni-karlsruhe.de 16 Feb '08

16 Feb '08

Hello, as you may know current code does not compile with ICC and assembler due to it optimizing away all the constants. I wonder if you'd find it acceptable to replace all the "static" attributes by a define, e.g. "asm_visibility" that would be static by default but empty for any compilers that cannot handle it (actually, I'd be just happy with having it defined to static always, a one line change seems ok to me to compile with a different compiler, just not the current situation). Greetings, Reimar D?ffinger

19 105

Behaviour of liba52 decoder
by tjordan＠macrosystem.de 14 Feb '08

14 Feb '08

Hello, when feeding complete AC3 frames to the liba52-decoder-implementation one has to call avcodec_decode_audio2 always two times, because the first call fills the input buffer and then returns, whereas the second call decodes. This is because the loop does while (buf_size > 0) { and buf_size is 0 after copying the data to the input buffer. this is in general not a problem for ffmpeg.c, as it retries, but at least imho a not so lucky concept for ffmpeg users just calling avcodec_decode_audio2. Why not giving one frame when we can give one 1:1? any comments? -- Regards, Thorsten

10 38

[PATCH] DVCPRO HD [1/3] dynamic DV macroblock lookup
by dmaas＠maasdigital.com 13 Feb '08

13 Feb '08

This is a series of 3 incremental patches that add DVCPRO HD support to ffmpeg. Patch 1 of 3 removes the look-up tables used by the DV codec to determine macroblock locations, and replaces them with functions that computes the macroblock locations at runtime. I feel this is a necessary step before adding DVCPRO HD support, because the look-up tables that would be required for the HD format would be very large. DVCPRO HD has about four times as many macroblocks as NTSC/PAL formats, and the coordinate numbers won't fit in 16-bit entries anymore, so the table would have to grow even more than that. I cannot detect any measurable speed change with this patch compared to the current look-up table code. If anything it may be slightly faster due to smaller cache footprint. (apologies for compressing the patch; it's 100KB uncompressed due to the removal of large look-up tables) Regards, Dan

3 10

transcoding on nvidia tesla
by dan＠novapulsar.com 12 Feb '08

12 Feb '08

Hi, How would one go about porting or using the ffmpeg application on nvidias tesla? http://www.nvidia.com/object/tesla_computing_solutions.html http://www2.pny.com/MarketingPromotions/Tesla/teslaHPC.aspx http://www.nvidia.com/object/tesla_tech_specs.html I believe a successful deployment of such a capability can be revolutionary. Tesla supports these OS's * MicrosoftR WindowsR XP (32-bit) (Available on Tesla C870 and Tesla D870 only) * LinuxR (64-bit and 32-bit) * Red Hat Enterprise Linux 4 and 5 * SUSE 10.1, 10.2 and 10.3 (coming soon) Thanks, Dan

9 30

h264-parser splits before PPS thus muxing to mp4 fails
by tjordan＠macrosystem.de 11 Feb '08

11 Feb '08

Hello, i encountered a problem while remuxing h264 material from TS to MP4. Video stream is copied, Audiostream transformed from AC3 to AAC, but only video matters here. The problem is that the input reader uses h264_parser to split the data, which exits just after SPS is found (h264_parser.c line 129 ff) but the MP4 muxer wants SPS and PPS, thus the assertion in mov_write_avcc_tag fails (movenc.c line 478). In these streams SPS comes first, then PPS. My question is if it would be sensible to split only after having SPS and PPS in the parser. If a stream has SPS there but no PPS this would be of course problematic, but the question is if that could happen. I guess it depends what the h264 spec says but you are the experts for that h264 code. What is your opinion? comments welcome... -- Regards, Thorsten

3 9

[PATCH]Fix number of first file in image2 format
by mbardiaux＠mediaxim.be 10 Feb '08

10 Feb '08

Currently, input probes first at zero, but output begins at number 1. Note this is just a fix for the default, I am preparing a patch that will add an AVOption for the start of the sequence (which will save those having scripts relying on the present behaviour!) -- Michel Bardiaux R&D Director T +32 [0] 2 790 29 41 F +32 [0] 2 790 29 02 E mailto:mbardiaux at mediaxim.be Mediaxim NV/SA Vorstlaan 191 Boulevard du Souverain Brussel 1160 Bruxelles http://www.mediaxim.com/

2 3