[Ffmpeg-devel-irc] ffmpeg-devel.log.20130107

burek burek021 at gmail.com
Tue Jan 8 02:05:02 CET 2013


[11:27] <cone-628> ffmpeg.git 03Paul B Mahol 07master:7f7f31bfcb7d: iff: support seeking with maud
[11:28] <durandal_1707> i think that iff demuxer should be split into ~5 demuxers
[11:31] <pross-au> yep, with iff.h or something for shared stuff
[11:54] <cone-628> ffmpeg.git 03Paul B Mahol 07master:251602369576: 012v: remove double ; and return correct error code if ff_get_buffer() fails
[12:01] <ubitux> should we add some NULL pointer check in functions like avcodec_close() or avformat_close_input()?
[12:01] <ubitux> it will be handy for error handling
[12:02] <durandal_1707> checks on what?
[12:03] <ubitux> check if the pointer is NULL
[12:03] <ubitux> and if so, do nothing (instead of dereference :p)
[12:04] <durandal_1707> on what pointers?
[12:06] <pross-au> all of 'em?
[12:06] <ubitux> the passed pointer?
[12:07] <ubitux> for avformat_close_input, the dereferenced one
[12:07] <cone-628> ffmpeg.git 03Peter Ross 07master:f2dc158b0a15: frmdec: use AV_PIX_FMT_xxx
[12:07] <cone-628> ffmpeg.git 03Peter Ross 07master:8d4c817c4a48: frmdec: terminate frm_pix_fmt_tags
[12:07] <cone-628> ffmpeg.git 03Peter Ross 07master:6cff56f0baba: frmdec: tag 2 is AV_PIX_FMT_RGB0
[12:17] <durandal_1707> pross-au: gonna split mvc decoders into separate files?
[12:21] <durandal_1707> when news will be posted?
[12:22] <pross-au> on my list of thiks todo
[12:25] <kierank> durandal_1707: mvc decoder?
[12:26] <durandal_1707> kierank: something  irrelevant for broadcating
[12:26] <kierank> what mvc
[12:26] <kierank> i have interests outside of broadcasting you know...
[12:27] <pross-au> Not broadcasting
[12:27] <durandal_1707> Silicon Graphics Motion Video Compressor 1 & 2
[12:27] <pross-au> Is an old SGI format, from the 90s
[12:28] <pross-au> there is also mvc3 :D
[12:29] <kierank> ok, i thought it was nokia mvc
[12:29] <pross-au> whats that?
[12:30] <Skyler_> H.264 proposal format, from the H.26L days
[12:31] <durandal_1707> pross-au: gonna do mvc3 ?
[12:36] <pross-au> dont have any samples :(
[12:36] <pross-au> it has a fancier name too, i forget
[12:36] <pross-au> so durandal_1707, you would sleep better a night if there was mvc1.c and mvc2.c ?
[12:38] <durandal_1707> currently if one of them is disabled code that not gonna be used is also build
[12:39] <durandal_1707> i'm just for splitting because there is nothing big shared betwen them
[12:40] <pross-au> okay
[12:40] <pross-au> i'll priotise that ahead of the WBMP decoder.
[12:40] <durandal_1707> you can keep then in same file if the ifdefs are correctly positioned ...
[12:41] <durandal_1707> no need to rush ...
[12:41] <durandal_1707> it is just minor thing
[12:46] <cone-628> ffmpeg.git 03Paul B Mahol 07master:2a89081cadff: Add EVRCA and SMV codec id
[13:16] <cone-628> ffmpeg.git 03Peter Ross 07master:508836932fcb: wtvdec: demux thumbnail picture to AVStream.attached_pic
[13:16] <cone-628> ffmpeg.git 03Peter Ross 07master:361ecb617624: wtvdec: ignore WM/MediaThumbType metadata entry
[13:16] <cone-628> ffmpeg.git 03Peter Ross 07master:801b636633ea: wtvenc: mux thumbnail picture
[13:16] <cone-628> ffmpeg.git 03Peter Ross 07master:8b4842c2e2cc: wtvdec: warn about truncated files
[13:51] <durandal_1707> pross-au: libsiren can decode audio in vivo 2 version?
[14:04] <cone-628> ffmpeg.git 03Luca Barbato 07master:89b51b570daa: oggdec: free the ogg streams on read_header failure
[14:04] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:2db8660f094a: Merge commit '89b51b570daa80e6e3790fcd449fe61fc5574e07'
[14:09] <Compn> durandal_1707 : i think thats correct, yes
[14:13] <pross-au> durandal_1707: the 'vivo siren' and 'msn siren' packets can both be decoded with libsiren, but they use different initialisation parameters. i guess codec_tag can be used to differentiate them.
[14:15] <Compn> is someone working on porting libsiren to ffmpeg ?
[14:21] <cone-628> ffmpeg.git 03Luca Barbato 07master:d894f74762bc: oggdec: make sure the private parse data is cleaned up
[14:21] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:1700be585593: Merge remote-tracking branch 'qatar/master'
[14:21] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:7a6beedd3fcd: oggparsevorbis: fix vorbis_cleanup return type
[15:04] <ubitux> pross-au: wtv-demux makes valgrind complain
[15:05] <ubitux> (memleak)
[15:27] <pross-au> Compn: libsiren port: i am *not* working on it.
[15:27] <pross-au> ubitux: thanks
[15:29] <durandal_1707> Compn: no plan to work on siren, after finishing evrc will do wavpack encoder
[15:52] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:878e5bd6f555: mpeg12: Support decoding dimensions that are a multiple of 4096
[15:52] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:bcc4ef1a415c: mpeg12enc: Correctly mask dimensions
[15:52] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:7fb87bc5f24b: mpeg12enc: check dimension validity
[16:17] <durandal_1707> wtf is Fire Flower ?
[16:22] <Compn> ?
[16:23] <ubitux> mario?
[16:24] <durandal_1707> that is name of 1.1 release
[16:27] <ubitux> yes i noticed
[16:28] <ubitux> to hell the peaceful names, let's start fighting with fire flowers
[16:28] <wm4> reminds me of playing supertux...
[16:35] <ubitux> nitpick question
[16:35] <ubitux> should subtitles be marked with a "bitmap based" flag, or a "text based" flag?
[16:35] <ubitux> (using a "text based" prop will require to mark more codecs)
[16:36] <durandal_1707> more patches
[16:36] <wm4> ubitux: it'd be best if they had just a type field?
[16:36] <wm4> and please not per rect
[16:37] <ubitux> there is already a type field per rect, and a "format" field in the AVSubtitle
[16:37] <wm4> why so many types
[16:37] <ubitux> it's set in a really ugly way currently
[16:37] <ubitux> sub->format = sub->num_rects && sub->rects[0]->ass;
[16:37] <ubitux> but it sucks and i want to change that
[16:38] <ubitux> wm4: think of a subtitles format/codecs supporting text & images
[16:38] <ubitux> i can give you at least one
[16:38] <ubitux> we don't support that feature though
[16:39] <ubitux> yet.
[16:39] <wm4> mixing them is a recipe for pain, I'm not sure if that is really the correct way of handling it
[16:40] <wm4> you should probably decode them as images always
[16:40] <wm4> what formats support this? kate?
[16:40] <ubitux> jacosub
[16:40] <ubitux> :D
[16:40] <ubitux> and maybe sami
[16:41] <wm4> does anyone use these formats/features?
[16:41] <ubitux> no one use jacosub nowadays
[16:41] <wm4> you know, in theory, ASS allows starting external programs and such
[16:41] <ubitux> some sami files can be found, maybe some are using it to overlay a logo or something
[16:41] <ubitux> but i doubt it
[16:41] <ubitux> wm4: yeah that's not important, i think we can safely assume either text-based or bitmap-based subs
[16:42] <ubitux> the question is more, should we consider bitmap or text subtitles the default
[16:42] <ubitux> i'd better go for text subtitles by default
[16:42] <ubitux> and consider bitmap one the "exception" (and thus mark codec as such)
[16:43] <ubitux> (that also means less work for me)
[16:43] <ubitux> but i was asking for opinions anyway
[16:44] <wm4> well, it would make sense to decode them as image
[16:44] <wm4> in fact, I wonder why you don't just provide "decoding" of text subtitles to images directly
[16:45] <ubitux> that would mean a rasterization of all the text in the decoder
[16:45] <ubitux> wm4: because you can't transcode if you do that
[16:45] <wm4> that's a separate thing, though
[16:45] <ubitux> the interesting thing about decoding is that it provides a common way of representing subtitles styles
[16:46] <ubitux> so you can use it in all the encoders
[16:46] <ubitux> and thus do some subtitles conversion keeping the markup
[16:46] <ubitux> things you couldn't do with images
[16:47] <wm4> video decoders (sometimes) allow decoding to multiple formats
[16:48] <wm4> it's mostly used to select between hardware vs. software decoding
[16:55] <cone-628> ffmpeg.git 03Peter Ross 07master:a572cc8c7a03: wtvdec: prevent memory leak in get_tag()
[18:35] <Plorkyeran> @ubitux> no one use jacosub nowadays <-- one person does
[18:36] <Plorkyeran> clearly you should design your software for the sake of one person that will never use it
[18:38] <ubitux> < Plorkyeran> @ubitux> no one use jacosub nowadays <-- one person does // really? who?
[18:38] <Plorkyeran> the central-anime guy
[18:39] <Plorkyeran> he still times on his amiga for some reason
[18:39] <ubitux> wtf :)
[18:39] <beastd> can he code :)
[18:39] <Plorkyeran> I don't think so
[18:40] <ubitux> Plorkyeran: could you tell him i wrote a jacosub decoder in ffmpeg just for him?
[18:40] <durandal_1707> lol
[18:44] <JEEB> woah
[20:35] <ubitux> cbsrobot: are you able to reproduce the ass bug on mac os?
[21:03] <cbsrobot> ubitux: yes (if it's a bug)
[21:04] <ubitux> so do you have any idea what is the cause?
[21:04] <cbsrobot> at least all the libs are not liked to the binary
[21:04] <cbsrobot> with ldd normally you see what is missing and what not
[21:05] <cbsrobot> and with otool I'm not sure how to do it
[21:05] <ubitux> just missing a link to a library?
[21:05] <ubitux> so if a library is missing, it just fails silently? oO
[21:06] <cbsrobot> ubitux: I mean the configure line states clearly that it is buid as a static binary and al the necessary libs are there
[21:06] <cbsrobot> but the output of otool does not mention them
[21:06] <cbsrobot> not sure what is going on
[21:07] <cbsrobot> with my ffmpeg the subtitling works, with the one form http://www.evermeet.cx/ffmpeg/snapshots/ it does not work
[21:07] <ubitux> interesting
[21:07] <ubitux> i just checked the code, we check for any init failure
[21:08] <cbsrobot> if you can give me a hint where to look &.
[21:08] <ubitux> so if libass was failing for some reason (like not being able to load a rendering lib) it would have error out
[21:08] <cbsrobot> at configure - yes
[21:10] <ubitux> cbsrobot: i meant as runtime right now
[21:10] <cbsrobot> well that aswell
[21:11] <ubitux> cbsrobot: if it's a static build, maybe there is some kind of conflicts with the libraries
[21:12] <ubitux> like libass calling the functions of freetype inside the binary instead of the one of the system or something
[21:12] <ubitux> mmh wait
[21:13] <ubitux> the dynamic libs are not found in your first example
[21:13] <ubitux> while they should, it's not a full-static build
[21:20] <cbsrobot> how would you do that ?
[21:21] <ubitux> do what?
[21:21] <ubitux> btw, does drawtext filter works?
[21:22] <cbsrobot> building a non full-static build ?
[21:22] <ubitux> by full-static i mean external libs embedded
[21:22] <cbsrobot> let me test
[21:22] <ubitux> --enable-static (which is the default) only embed our libs into the programs
[21:23] <ubitux> (’ ./ffplay -f lavfi -i color,drawtext=text=hello:fontcolor=white:fontsize=30)
[21:24] <ubitux> even better: ./ffplay -f lavfi -i 'color,drawtext=text=hello:fontcolor=white:fontsize=30:x=t*10'
[21:24] <ubitux> :-°
[21:27] <cbsrobot> ubitux: I only have ffmpeg, no ffplay
[21:27] <cbsrobot> but it works
[21:27] <ubitux> ok
[21:30] <ubitux> cbsrobot: just to make sure, if you use -v verbose, do you see the "Change happened at time ..." messages?
[21:30] <ubitux> hum, -v debug sorry
[21:32] <cbsrobot> ubitux: I have quickly to get afk
[21:32] <cbsrobot> I ping you back
[21:42] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:f65daf577af2: rv10: always check direct mode interpolation times.
[21:42] <cone-628> ffmpeg.git 03Michael Niedermayer 07master:0780fe27404c: rmdec: Limit videobufsize to remaining amount of data
[22:10] <cbsrobot> ubitux: I found: Fontconfig error: Cannot load default config file
[22:11] <cbsrobot> on the one I downloaded
[22:11] <ubitux> oh!
[22:11] <cbsrobot> didnt see it before
[22:11] <ubitux> but it doesn't error out?
[22:11] <cbsrobot> no
[22:11] <cbsrobot> well its red
[22:13] <cone-628> ffmpeg.git 03Clément BSsch 07master:dbf0a9057271: lavc: fix forgotten "que" reference.
[22:13] <cbsrobot> not sure if it matters, but my ffmpeg shows : [Parsed_subtitles_0 @ 0x7fcb89c19640] Raster: FreeType 2.4.7
[22:13] <cbsrobot> the one failing: [Parsed_subtitles_0 @ 0x107b15740] Raster: FreeType 2.4.10
[22:13] <cbsrobot> FrBidi is the same version
[22:14] <ubitux> it doesn't seem to appear in https://ffmpeg.org/trac/ffmpeg/attachment/ticket/2100/ffmpeg-20130102-200734.log
[22:22] <cbsrobot> with the failing ffmpeg I cannot see "Change happened at time &."
[22:26] <ubitux> i wonder if ass_render_frame doesn't return NULL in these cases
[22:57] <cbsrobot> ubitux: something else I shoud check ?
[22:57] <ubitux> i don't know&
[22:57] <cbsrobot> me neighter atm
[22:57] <ubitux> :(
[00:00] --- Tue Jan  8 2013


More information about the Ffmpeg-devel-irc mailing list