[Ffmpeg-devel-irc] ffmpeg-devel.log.20140706

[burek]

[00:24] <Timothy_Gu> michaelni: so im reading doc/swresample.txt. What do you mean by "special converter"?
[00:29] <michaelni> something that combines several steps, and is faster then doing the individual steps seperate, but which only supports a commonly occuring subset of possibilities
[00:38] <jamrial> every other yasm patch breaks msvc...
[00:40] <cone-486> ffmpeg.git 03Timothy Gu 07master:baef3ba727a3: avcodec/old_codec_ids: workaround doxygen inclusion bug
[00:40] <Timothy_Gu> michaelni: can you give an example?
[00:53] <michaelni> for example if input and output fornats match memcpy() is used to bypass the normal path, or if no resamping/rematrixing is needed then the sample format can be converted without going through a planar intermediate. One also could imagine some special case handling for common cases like the samplerates/channel configs/sample formats used by dvd/ bluray/flash/prores whatever in principle
[00:56] <Timothy_Gu> ok, thx
[01:16] <cone-486> ffmpeg.git 03Timothy Gu 07master:b27555a3a05f: avutil/samplefmt: improve doxygen
[01:30] <Daemon404> [23:38] <@jamrial> every other yasm patch breaks msvc... <-- aligment fun?
[01:55] <jamrial> maybe
[01:57] <jamrial> mmh, seems to also fail in one x86 linux fate slot with SEGV
[01:57] <jamrial> maybe that mova should be movu
[01:57] <cone-486> ffmpeg.git 03Michael Niedermayer 07master:26b9f5bcdb69: avcodec/Makefile: add libutvideo.h to SKIPHEADERS
[01:57] <cone-486> ffmpeg.git 03Timothy Gu 07master:689e02808d09: avutil/audio_fifo: split into a separate doxy module
[01:58] <jamrial> nevcairiel: could you test the above? see if that fixes it on msvc
[04:06] <cone-486> ffmpeg.git 03Michael Niedermayer 07master:02729f7a5afe: avcodec/mjpegenc_common: Restore removed copyright statements
[04:06] <cone-486> ffmpeg.git 03Timothy Gu 07master:37715b4594dd: swresample: split option table to a separate file
[05:17] <cone-486> ffmpeg.git 03Michael Niedermayer 07master:a53c5d454ee6: avformat/smoothstreamingenc: fix memleak
[14:40] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:11e4890eeb82: avfilter/unsharp_opencl: use av_mallocz_array()
[14:40] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:d90ad5b2bb22: avfilter/vf_vignette: Use av_malloc_array()
[14:40] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:e592bf3d2ff7: avfilter/vsrc_cellauto: Use av_mallocz_array()
[19:24] <cone-54> ffmpeg.git 03Martin Storsjö 07master:865461099e06: librtmp: Don't free the temp url at the end of rtmp_open
[19:24] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:1c8c4b8a89b9: Merge commit '865461099e062de5a3a109c2a5be98004c11d8bd'
[19:40] <cone-54> ffmpeg.git 03Nidhi Makhijani 07master:77fc7b767267: vc1test: Return proper error codes
[19:40] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:d95a04547a32: Merge commit '77fc7b76726719746ca45df6c4a62c41abad506e'
[19:55] <cone-54> ffmpeg.git 03Nidhi Makhijani 07master:02a7a5e330ec: vc1test: Check malloc call
[19:55] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:3a29af4efc60: Merge commit '02a7a5e330ec4373a761f63c292220a1de13b968'
[20:10] <cone-54> ffmpeg.git 03Nidhi Makhijani 07master:373a6dda5422: cljr: split decoder and encoder
[20:10] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:225c5cc81535: Merge commit '373a6dda5422186bc057297342a9e559a564595e'
[20:16] <wm4> why does ffprobe have no -ss option
[20:16] <wm4> is there a pseudo muxer that just dumps incoming information?
[20:20] <saste> wm4, it has it
[20:20] <saste> -read_intervals
[20:24] <wm4> saste: thanks
[20:26] <cone-54> ffmpeg.git 03Nidhi Makhijani 07master:b0633f83f277: paf: split audio and video decoder
[20:26] <cone-54> ffmpeg.git 03Michael Niedermayer 07master:0d1994ee3cf8: Merge commit 'b0633f83f277c05bf1f617a99c7aedd2db8306e3'
[20:30] <wm4> where do files uploaded with http://upload.ffmpeg.org/upload/ end up?
[20:31] <j-b> wm4: in incoming/
[20:31] <wm4> it's not there
[20:31] <wm4> maybe the upload just didn't work or so
[20:32] <j-b> wm4: http://streams.videolan.org/incoming/ 
[20:32] <wm4> yes, I looked there
[20:32] <cone-54> ffmpeg.git 03James Almer 07master:195f7bd23de7: x86/svq1enc: use unaligned mov on SSE2
[20:34] <j-b> I need to code something with a progress bar
[20:44] <Timothy__Gu> wm4: is gjdfgh you?
[20:45] <wm4> Timothy__Gu: yes
[20:45] <Timothy__Gu> why dont you just use wm4?
[20:45] <Timothy__Gu> thats enough anonimousity
[20:45] <wm4> it's just another throw away account, so I didn't pay attention to that
[20:46] <Timothy__Gu> old: http://ffmpeg.org/doxygen/2.2/group__preproc__misc.html
[20:46] <Timothy__Gu> new: http://ffmpeg.org/doxygen/trunk/group__preproc__misc.html
[20:47] <wm4> heh
[20:47] <Timothy__Gu> it's hard to imagine that removing an include can cause so much difference
[20:54] <Timothy__Gu> michaelni: libav's cherry picking my recent patches, but without the @par -> @li change (because the change makes their doxygen ugly with customized CSS). Just use the patch already committed to ffmpeg (i.e. with the @par -> @li).
[20:56] <michaelni> ok
[20:56] <cone-54> ffmpeg.git 03Timothy Gu 07master:02ca9efdcd55: avutil: actually install hash.h
[20:56] <cone-54> ffmpeg.git 03Timothy Gu 07master:0c5838821172: swresample: grammar/capitalization fixes
[21:07] <cone-54> ffmpeg.git 03Timothy Gu 07master:c0d9b026f911: swresample: group all the option constants in a section in doxy
[21:07] <cone-54> ffmpeg.git 03Timothy Gu 07master:fc71434e8464: swresample: add SwrDitherType doxy
[21:07] <cone-54> ffmpeg.git 03Timothy Gu 07master:77c5f546e761: swresample: add SwrContext doxy
[21:15] <ejl> anyone have the patience to explain how to submit a patch to someone who's completely stupid about using git?
[21:15] <ejl> a.k.a. me
[21:16] <wm4> ejl: git format-patch HEAD~1 creates a patch for the most recent commit, which you then could attach to a mail
[21:16] <wm4> that's the "5 seconds explanation" to it
[21:16] <Compn> ejl : can make a regular diff -u if you dont like git :)
[21:16] <wm4> Compn: shut up
[21:16] <Compn> bwahaha
[21:16] <Compn> wm4 : stop being prejudice against anti-git devels
[21:17] <wm4> ejl: git can also send mails directly with git send-email, but I guess that's advanced usage
[21:17] <ejl> I'm not anti-git, I just haven't used it
[21:26] <ejl> the process is create branch --> add files --> commit --> format patch?
[21:26] <j-b> yes
[21:26] <j-b> but you do not need to create a branch
[21:26] <ejl> hmm
[21:27] <wm4> it'll probably cause less of a mess to use a branch
[21:27] <Timothy_Gu> ejl: and you don't need add: you just `git commit -a`
[21:28] <wm4> ejl: be sure to look at the patch when you're done, to make sure no unrelated changes got in
[21:28] <ejl> yeah I know
[21:28] <ejl> the change is trivial, me getting this wrong is the risky part :(
[21:29] <Timothy_Gu> so what i do is: git checkout -b fun-feature; text-editor; git commit -a; git format-patch master
[21:29] <Timothy_Gu> git send-email *.patch
[21:44] <wm4> Timothy_Gu: I didn't know git send-email can take patch files; you can make it send patches "directly" without creating patch files too
[21:45] <Timothy_Gu> wm4: really? i only know that it takes patch files...
[21:58] <ubitux> Timothy_Gu: git send-email master
[21:58] <Timothy_Gu> oh, thanks
[22:03] <ejl> just to be sure, does this look correct?  http://pastebin.com/aqLYUNji
[22:05] <cone-54> ffmpeg.git 03Timothy Gu 07master:2711b4708a20: swresample: improve Doxygen introduction
[22:18] <Timothy_Gu> ejl: the formatting is good, but if possible please use your real name. (git config --global user.name Your Name)
[22:19] <ejl> ah
[22:19] <Timothy_Gu> ejl: HOWEVER, we are deprecating the YUVJ IDs, so you also need to add avctx->color_range = AVCOL_RANGE_JPEG;
[22:21] <Timothy_Gu> for encoder, leave the original id there, but add the J id. in the code, check if color_range is JPEG, and return if not.
[22:23] <wm4> <Timothy_Gu> ejl: HOWEVER, we are deprecating the YUVJ IDs <- and been doing that for years :/
[22:24] <Timothy_Gu> wm4: this time it's for real. I assume you saw koda's patches.
[22:24] <wm4> yes, but there's still libavfilter left, which is the real problem
[22:24] <Timothy_Gu> oh well...
[22:26] <Timothy_Gu> but they can always ask those OPW girls to do them
[22:28] <cone-54> ffmpeg.git 03Timothy Gu 07master:81f47e272d66: swresample: better doxy for configuration-returning functions
[22:28] <cone-54> ffmpeg.git 03Timothy Gu 07master:064945b3aa7c: swresample: organize functions into doxy groups
[22:31] <ejl> so does that mean transcoding JPEG YUV to a format that takes MPEG YUV or vice versa will fatal error unless -color_range is used?
[22:35] <ejl> oh nm, guess that's what UNSPECIFIED is for
[22:37] <JEEB> calling them MPEG and JPEG YCbCr is really derpy, just call them full and limited range :P
[22:37] <JEEB> since it's the same YCbCr
[22:37] <JEEB> just the range is what is different
[22:39] <Compn> is this a naming bikeshed ? :)
[22:42] <ejl> the worth of a programmer is measured by the amount of time spent naming things, and the length of his hard tabs
[22:43] <JEEB> nah, I'm not bikeshedding really :P
[22:44] <JEEB> call them as you want, and swscale calls them like that IIRC. I just dislike them being specified to JPEG and friends
[22:50] <ejl> mmm, actually the color_range usage is still kind of odd.  Like, there are cases like ljpeg_encode_init which assume == MPEG is invalid, which only works via the assumption that UNSPECIFIED and the correct one are the only other options
[22:50] <ejl> which is true, currently, but... eh
[23:02] <ejl> well... actually I'm still not sure I get the default case here.  Like, if someone transcodes an MPEG-range input to an output that only accepts JPEG, and color_range is unspecified, what indicates that the conversion must be done?
[23:34] <Timothy_Gu> ejl: the input decoder should in theory set the color_range to a value, but I'm pretty sure not all decoders do that.
[23:37] <ejl> right, it sets the color range that the decoder is inputting, but how does it know the color range that it needs to convert to for the encode?
[23:38] <ejl> like if I transcode an MPEG frame to JPEG, color_range on the MPEG codec is MPEG, but the color range on the JPEG codec is unspecified, so it doesn't convert the range and the output will be wrong.
[23:38] <cone-54> ffmpeg.git 03Eric Lasota 07master:8be23d424fee: avcodec/roqvideo: use JPEG color range
[23:39] <michaelni> ejl, encoder range is known from AVCodec.pix_fmts
[23:40] <michaelni> also please send a patch which sets color_range correctly for the decoder
[23:40] <ejl> yeah I'm doing that, I'm just trying to make sure I understand this
[23:43] <Timothy_Gu> michaelni: did you check if FATE breaks? It should ...
[23:43] <michaelni> i updated fate
[23:44] <Timothy_Gu> ok
[23:44] <ejl> i.e. given: <@Timothy_Gu> for encoder, leave the original id there, but add the J id. in the code, check if color_range is JPEG, and return if not.
[23:44] <ejl> if YUV444P is left in the encode format list, then transcoding from MPEG to that will cause bad color, because it has no way of knowing that it needs to convert the range, right?
[23:45] <michaelni> yes
[23:51] <ejl> ok.  I can add that too, if that's the intended behavior, but it seems like codecs should declare what color ranges they support if the J IDs are going to be removed.
[00:00] --- Mon Jul  7 2014