[Ffmpeg-devel-irc] ffmpeg-devel.log.20130928

[burek]

[10:57] <cone-168> ffmpeg.git 03Anton Khirnov 07master:38e15df1489d: avframe: note that linesize is not the usable data size
[10:57] <cone-168> ffmpeg.git 03Michael Niedermayer 07master:945c7e399af2: Merge commit '38e15df1489d86c016515223ee693e7d0326c56a'
[11:10] <cone-168> ffmpeg.git 03Anton Khirnov 07master:0767bfd1994c: lavfi: allow user-provided execute() callbacks
[11:10] <cone-168> ffmpeg.git 03Michael Niedermayer 07master:7381d31f2267: Merge remote-tracking branch 'qatar/master'
[11:35] <cone-168> ffmpeg.git 03Ronald S. Bultje 07master:c341f734e5f9: Convert multiplier for MV from int to ptrdiff_t.
[13:05] <BBB> michaelni: will you also apply the emu_edge patch, or do I need to resend that?
[13:06] <BBB> thanks btw :) almost ready for vp9 now
[13:08] <michaelni> BBB i wanted after some more testing
[13:26] <BBB> michaelni: ok thanks
[13:30] <wm4> so we'll get hevc and vp9 at about the same time? interesting...
[13:32] <Compn> wm4 : it'll be more interesting to see which one the anime people use
[13:33] <Compn> probably depends how many bits each codec can max out 
[13:33] <wm4> why the anime people?
[13:33] <wm4> because they like breaking everything all the time?
[13:33] <Compn> because they often choose something that requires a lot of new code in mplayer :P
[13:33] <Compn> ordered chapters, 10bit, etc
[13:33] <Compn> ass subs
[13:33] <durandal_1707> mplayer have none of that
[13:34] <wm4> mplayer had 10 bit support before mplayer2
[13:34] <Compn> 10bit should work too...
[13:34] <Compn> er i mean ass 
[13:34] <wm4> mplayer didn't fix most code assuming that 1 Y pixel is 1 byte, though
[13:34] Action: Compn needs to wake up
[13:38] <michaelni> BBB seems iam getting failures on arm (fate-vp8-test-vector-003 and others), ill double check that it works before 
[13:39] <michaelni> its mismatchig checksums no crash so far
[13:56] <BBB> michaelni: vp8?
[13:56] <BBB> michaelni: may be the same issue the c code had
[13:56] <michaelni> seems 6 vp8 tests fail (no other failures)
[13:57] <michaelni> vectors 3 4 and 7
[14:01] <BBB> that's bilinear again
[14:01] <BBB> is that neon or armv6?
[14:01] <BBB> it's neon, I see the code is using dst_stride for src again
[14:02] <BBB> same issue C code had
[14:02] <BBB> let me see if I can fix that...
[14:33] <BBB> michaelni: http://pastebin.com/B9Z1gixE
[14:55] <michaelni_> BBB seems segfaulting
[15:02] <michaelni_> BBB http://pastebin.com/Js6bvK7L
[15:14] <BBB> michaelni_: I have the idea that that disassembly isn't quite accurate, can you disassemble all of ff_put_vp8_bilin16_h_neon?
[15:15] <BBB> the start offset is probably not at an instruction boundary
[15:17] <BBB> oh I probably have to update the stack offsets
[15:20] <BBB> michaelni_: http://pastebin.com/2NhTQRvL
[15:52] <cone-323> ffmpeg.git 03Michael Bradshaw 07master:f46a3e3d6ec4: libopenjpeg: fix encoding of odd sized subsampled images
[15:58] <michaelni_> BBB new patch crashes randomly, gdb doesnt provide very usefull baktrace (only ??) and disassembly says Cannot access memory at address 0xf7f0a70
[16:00] <michaelni_> ill retest with make distclean just to be sure
[16:05] <michaelni_> still crashing after distclean
[16:26] <BBB> michaelni_: pff... ok I'll check more
[16:39] <cone-323> ffmpeg.git 03Paul B Mahol 07master:88c8ed18c972: avfilter/dualinput: remove unused variable
[17:39] <cone-323> ffmpeg.git 03Paul B Mahol 07master:07da0854fe98: avformat/vfq: fix deadlock in vqf_read_header()
[18:13] <michaelni_> burek, "<durandal_1707> michaelni: those FFmpeg forums nobody use except noobs (expecting help that never comes) so should be removed from contact page"
[18:14] <burek> well, feel free to remove it, if you think it doesn't do any help
[18:15] <burek> after all, it's a lot easier to criticize than to help, right? :)
[18:15] <burek> (considering i didn't see durandal not once on that forum, helping other people) :)
[18:16] <michaelni> iam not thining anything, just wanted to make sure you see it as its your forum
[18:16] <michaelni> s/thining/thinking/
[18:16] <durandal_1707> burek: that response was caused by recent reply to thread in forum about -af nonexistent and that lavfi should be used ....
[18:16] <burek> i didnt understand this last sentence
[18:17] <burek> (@michaelni)
[18:17] <durandal_1707> thread was from 2012
[18:17] <durandal_1707> and now there is -af
[18:19] <michaelni> burek, i am not reading the forums so i cant say anything about them, i just saw durandal_1707 request and noticed you where not online so i just wanted to forward it
[18:20] <burek> ok, thank you for that michaelni
[18:20] <burek> durandal_1707
[18:20] <burek> if you really want to help people, you should join the forum and help them
[18:20] <burek> i don't see any benefit from removing the link to the forum from the contact page
[18:21] <durandal_1707> it was not request, it was more like request for status change, because i do not see anybody else except burek and llogan help, and they are not on forum 24h
[18:21] <burek> if there was an official ffmpeg forum, i would be more than happy to close mine (and forward the database of posts to the new one, if needed)
[18:22] <burek> durandal_1707, but how would removing the link from the contact page help anyone?
[18:23] <ubitux> wm4: ok so, back on vobsub debugging: (one - major) problem is that mpeg pes hdr reader is being too smart and factorize the packets
[18:24] <ubitux> so sometimes i get one too large packets, and i just can't actually split it to match the vobsub index expected packet size
[18:25] <ubitux> i'm still trying to find a sexy way to deal with this
[18:26] <wm4> maybe that's the reason mplayer's demux_mpg didn't work with ffmpeg's vobsub
[18:37] <durandal_1707> see that fraps report, it take almost 10 days to reply
[18:42] <durandal_1707> michaelni: from here: http://ffmpeg.gusari.org/viewtopic.php?f=16&t=1078 vlc can take up to 31 bits in fraps, while maximal supported looks to be 30
[19:08] <michaelni> increasing FF_HUFFMAN_BITS should solve this, best is probably to make it per codec variable as the increase could slow things down
[19:09] <michaelni> and having it non constant for get_vlc() also would slow things down
[19:12] <ubitux> wm4: btw, you use lavc decoder for dvdsub?
[19:12] <wm4> ubitux: yes
[19:12] <wm4> I still have the old decoder around... for old things
[19:13] <wm4> like the current libav stable release
[19:19] <ubitux> ok
[19:22] <burek> <@durandal_1707> see that fraps report, it take almost 10 days to reply
[19:22] <burek> wouldn't it be more helpful if you would join the forum and help the guy, answering that question
[19:47] <durandal_1707> nope, it wouldn't be more helpful
[19:47] <durandal_1707> michaelni: aren't you saying that to fix this one need to make things slower?
[20:02] <burek> is there a text-only version of histogram filter?
[20:03] <durandal_1707> what text only version would do?
[20:04] <durandal_1707> short answer no
[20:04] <burek> for example if there is a need to calculate the mean value of all the colors in the image
[20:04] <burek> text version would help, giving the distribution in a parsable format, which can be used to calc the mean value
[20:05] <durandal_1707> hmm, no
[20:06] <burek> ok, thanks
[20:21] <cone-323> ffmpeg.git 03Michael Niedermayer 07master:a63b479d1191: avformat/utils: factor find_decoder() out
[20:44] <cone-323> ffmpeg.git 03Michael Niedermayer 07master:2b834a0605d9: avcodec/huffman: Allow specifying nb_bits to ff_huff_build_tree()
[21:05] <burek> just to ask here too, in case people dont read #ffmpeg channel... is there any good page that explains the relation between -maxrate, -minrate and -bufsize, because there are several duplicate questions about it, where people in general lack the understanding of that relation and as a result use them in a wrong way
[21:22] <burek> grr.. how to delete a page.. I accidentally put all the first letters in caps..
[21:30] <ubitux> oh fuck, they're split
[21:31] <ubitux> why on earth those dvdsub are so complex
[21:32] <wm4> because DVDs are a horrible thing
[21:34] <ubitux> well
[21:34] <ubitux> here is i have a subtitle chunk at a given ts
[21:34] <ubitux> but it's not complete, and the rest of that chunk is happening at a LATER timestamp
[21:35] <Daemon404> i dream that one day
[21:35] <Daemon404> linux will flawlessly play a dvd from 1998
[21:35] <ubitux> and you can't start decoding the first one, since the control sequence is in the next packet
[21:35] <ubitux> what. the. hell/
[21:38] <JEEB> that... makes no sense
[21:38] <JEEB> are you supposed to just cache it until you get a control sequence?
[21:39] <Daemon404> i had the dvd spec at one point
[21:39] <Daemon404> lost it though...
[21:39] <wm4> admit it, you burned it in  dark satanic ritual
[21:39] <wm4> +a
[21:39] <ubitux> JEEB: yup
[21:39] <Daemon404> wm4, it was a pdf
[21:39] <ubitux> JEEB: i have no idea why the timestamps are diverging..
[21:39] <Daemon404> so no i didnt burn it... i think
[21:40] <Daemon404> ;)
[21:40] <wm4> Daemon404: I thought that didn't exist
[21:40] <wm4> something about dvd specs only existing in printed form
[21:40] <Daemon404> maybe
[21:40] <Daemon404> i might be on crack.
[21:40] <JEEB> huh
[21:40] <JEEB> indeed
[21:40] <JEEB> there were "PDFs"
[21:40] <JEEB> which were shells that looked like PDFs
[21:40] <JEEB> and then you had a binary plugin to adobe reader
[21:41] <JEEB> that used a USB key to decrypt the book
[21:41] <JEEB> this was only sold from 2007 to 2012 or so? I think
[21:41] <ubitux> through an ie6 activex?
[21:41] <Daemon404> so you could print screen it?
[21:41] <burek> :facepalm:
[21:42] <JEEB> http://www.dvdfllc.co.jp/pdf/Guidelines%20for%20the%20DVD%20Book%20Digital%20Edition%20%28Nov.%202009%29.pdf
[21:42] <JEEB> also I love it how the logo of the DVD FLLC consists of seemingly Comic Sans MS
[21:43] <Daemon404> 'easily portable'
[21:43] <wm4> the fuck
[21:43] <ubitux> Daemon404: usb keys are not heavy
[21:43] <ubitux> ;))
[21:43] <Daemon404> lol
[21:43] <JEEB> :D
[21:44] <JEEB> "** IMPORTANT NOTICE ** The sale of DVD Book Digital Edition was terminated on July 31, 2012."
[21:44] <JEEB> yeah, it was stopped in 2012
[21:45] <JEEB> http://www.dvdfllc.co.jp/format/f_sbsc.html
[21:47] <ubitux> i'm wondering if i should reassemble within dvdsubdec
[22:38] <wm4> oh man what's with those patches to force codecs
[22:38] <wm4> why is this in libavformat and not in ffmpeg.c?
[22:51] <michaelni> av_find_stream_info opens the decoders
[22:59] <wm4> hm I see... I thought there was a separation between demuxers and decoders, but that's not really the case anymore
[23:01] <michaelni> a sad necessity for all the real world formats
[23:03] <michaelni> instead of the passing over AVFormatContext we also could use the AVDIct argument to av_find_stream_info() but it would nt work for streams added after the av_find_stream_info() call that is by the read_packet function used inside av_find_stream_info
[23:06] <cone-323> ffmpeg.git 03Matthieu Bouron 07master:c461265a769e: lavf/mxfdec: add j2k codec to intra only codecs
[23:43] <j-b> Is there a good doc around, describing categories of video filters (deinterlacers, smoothers, sharpeners, ...) ?
[23:53] <ubitux> j-b: filters are not organized into categories, except "audio", "video" and "multimedia" (mixed)
[23:54] <ubitux> and they're described here: http://ffmpeg.org/ffmpeg-filters.html
[23:54] <j-b> no, I am looking for something wider than FFmpeg
[23:54] <ubitux> oh, ok sorry
[23:54] <j-b> a bit more theory, probably
[23:54] <ubitux> there was a nice coursera course a while ago
[23:55] <ubitux> but it was some specific filters though, nothing about interlacing for instance
[23:55] <j-b> ubitux: the best I got was http://www.aquilinestudios.org/avsfilters/index.html
[23:55] <j-b> but that's a bit poor
[23:55] <ubitux> is it for you?
[23:56] <j-b> not really
[23:56] <j-b> I need to teach a bit to students
[23:57] <ubitux> https://www.coursera.org/course/images  this one was quite interesting
[23:58] <ubitux> deinterlacing/telecining needs a dedicated course IMO
[23:58] <ubitux> you can't easily introduce these filters between gamma correction and sharpening
[00:00] --- Sun Sep 29 2013