[Ffmpeg-devel-irc] ffmpeg-devel.log.20140116

[burek]

[00:18] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07master:565102dcac49: Do not set mkv bit_depth to av_get_bytes_per_sample() for G.726.
[00:18] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07master:11329370770e: Do not read mkv audio bit_depth if bits_per_coded_sample is already set.
[00:29] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07release/1.2:d6a705d7787f: Do not set mkv bit_depth to av_get_bytes_per_sample() for G.726.
[00:29] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07release/1.2:30147f14d4b6: Do not read mkv audio bit_depth if bits_per_coded_sample is already set.
[00:29] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07release/2.1:0d45e821dc14: Do not set mkv bit_depth to av_get_bytes_per_sample() for G.726.
[00:29] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07release/2.1:2f4b781e4db5: Do not read mkv audio bit_depth if bits_per_coded_sample is already set.
[02:15] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07master:cb36e441de47: Never favour an audio stream with 0 channels in av_find_best_stream().
[02:15] <cone-668> ffmpeg.git 03Carl Eugen Hoyos 07master:fe9f57592491: Warn if s16be is muxed into flv.
[02:27] <BBB> ubitux: that's fine, patch lgtm, just commit it
[02:47] <cone-668> ffmpeg.git 03Paul B Mahol 07release/1.2:b8b77aefe894: avcodec/ffv1dec: fix format detection
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:da82be0cc00c: ffserver: strip odd chars from html error messages before sending them back
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:45dd7df83b39: avfilter/ff_insert_pad: fix order of operations
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:d805a5171382: avcodec/avpacket/av_packet_split_side_data: ensure that side data padding is initialized
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:252002aec13b: avcodec/utils: add some saftey checks to add_metadata_from_side_data()
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:803ca5c3491a: avutil/opt: initialize ret
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:751e684aae1e: avcodec/jpeglsdec: check err value for ls_get_code_runterm()
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:d0ed672484e4: avformat/thp: fix variable types to avoid overflows
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:971b13752d92: avformat/thp: force moving forward
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:c7027ae7381d: h264: Do not treat the initial frame special in handling of frame gaps
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:dea7f1c62e1a: avcodec: move end zeroing code from av_packet_split_side_data() to avcodec_decode_subtitle2()
[02:47] <cone-668> ffmpeg.git 03Paul B Mahol 07release/1.2:5a8d78ab7f54: avcodec/libopusenc: change default frame duration to 20 ms
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:3488e9e269f8: avcodec/error_resilience: factor er_supported() check out
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:b841869477c9: avcodec/error_resilience: check that er is supported before attempting to read the status of the previous slice
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:5dde8ba59e90: avutil/log: skip IO calls on empty strings
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:ab38b390592d: evrcdec: use memmove() instead of memcpy() when regions can overlap.
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:a92b73da9907: svq3: use memmove to avoid overlap in memcpy.
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:a2186a805442: avcodec/mpegvideo_motion: Check P field references
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:8ded3738d15c: avcodec/c93: Check for block overlap.
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:d1da1c8384c6: avcodec/mpegvideo_motion: Use a field from the current frame if the last is unavailable in DMV & 16x8
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:08a319549a41: avcodec/bitstream: Check bits in ff_init_vlc_sparse()
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:bb6f4667946c: avcodec/bitstream: check codes in ff_init_vlc_sparse()
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:c93501687ca9: xan: Check for overlapping copies
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:6e3697b98552: eamad: Check against minimum supported dimensions
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:2b26f8c6bf76: evrcdec: fix sign error
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:ca7f64e0ac20: shorten: allocate space for padding
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:7855083443e0: dxa: check vectors of 4x4 motion blocks
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:84b100396efb: dxa: check vectors of 2x2 motion blocks
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:3d380ffde9af: avformat/vqf: check number of channels before use.
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:e438fd3be905: avcodec/lcldec: Check that dimensions are a multiple of the subsample factors
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:dee327b0e70d: shorten: fix "off by padding" bug
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:271f5d68b922: avcodec/ansi: Check x/y
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:483b2016bb70: avcodec/ansi: Fix right/bottom end checks
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:a2bea0df0e4a: aacps: correct opdipd code to match spec
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:3d1e4b7ca5d5: avcodec/aacps: fix number of bands used with ipd/opd
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:47faf347a0ed: h264_cavlc: fix assertion failure due to reading too long vlc
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:b9058f58bde9: h264_refs: Check for attempts to assign pictures to short & long.
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:3f2efac0fec1: h264: Move slice_table clean out of frame_start
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:97978b7ae8e9: h264: move the default_ref_list_done check down after its inputs have been written
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:c8b90c7cd526: h264: Always decode MMCOs into temprary array
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:15a736483e3c: h264: check mb_width/height
[02:47] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:db20610c4967: wmaprodec: Fix null pointer dereference in decode_frame()
[02:54] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:884094debae7: Update for 1.2.5
[02:54] <cone-668> ffmpeg.git 03Michael Niedermayer 07release/1.2:5a0a156e88f7: cmdutils: update year
[05:54] <juanmabc> tsup, any func helper to planar to linear(usual) frm->data?
[05:58] <juanmabc> i guess i have to swr_convert ?
[06:48] <juanmabc> i'm so happy it finally works ;DDD
[07:59] <ubitux> BBB: ok, applied :)
[07:59] <cone-48> ffmpeg.git 03Clément BSsch 07master:9cc8fa63ddba: vp9/x86: simplify a few mc inits.
[09:52] <sspiff> ubitux: could you give me a link to that patch you mentioned that could fix the dvb sub bug?
[09:53] <ubitux> it's pushed
[09:53] <ubitux> http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=c917cde9cc52ad1ca89926a617f847bc9861d5a0
[09:54] <plepere> ubitux : what's the thing to debug assembly already ? I can't find it in my history. :/
[09:54] <ubitux> the thing to debug?
[09:55] <ubitux> "gdb"?
[09:55] <plepere> ah thanks. :)
[09:55] <plepere> having trouble on some qpel
[10:01] <ubitux> plepere: "set disassembly-flavor intel" in ~/.gdbinit, then "b ff_hevc_foobar_qpel_ssse3", "r", copy paste http://pastie.org/pastes/8638328/text, and "ni"
[10:01] <ubitux> hf.
[10:01] <ubitux> (it's a real pain, and i'm probably doing it wrong)
[10:02] <ubitux> ('forgot the step "gdb --args ./ffmpeg_g -i foo.hevc -f null -")
[10:13] <sspiff> ubitux: doesn't look like it would fix the out-of-order problem of clear packets, so I'll still have a look next week.
[10:13] <sspiff> The solution I have in mind might also remove the DVBSUB specific code from ffmpeg.c
[10:13] <ubitux> sspiff: the clear packet thing is probably wrong anyway
[10:14] <ubitux> sspiff: see this: http://trac.ffmpeg.org/ticket/2024#comment:8
[10:17] <sspiff> ubitux: yeah I read
[10:17] <sspiff> looking at the DVB subtitle standard, looks like he's right about not having to send a clear packet
[10:18] <sspiff> the page composition segment comes with A) a timeout value and B) can be redefined by a later packet to have 0 regions
[10:25] <sspiff> ubitux: but sending an update is apparently exactly what ffmpeg does, currently
[10:25] <plepere> ubitux, I'm doing it on our own openHEVC project, so it's going to be a bit diffrent, but thanks for the reminders. :)
[10:25] <ubitux> sspiff: i don't know :)
[10:25] <ubitux> sspiff: you're the expert
[10:25] <sspiff> for DVB subs, it sends out two frames, one with the actual sub, and a second with the exact same thing but with num_rects set to 0
[10:26] <ubitux> sspiff: yes but maybe all of them should be cleared
[10:26] <sspiff> ubitux: you mean "not all of them", right?
[10:27] <ubitux> yes sorry
[10:28] <sspiff> from my current point of view, it should be decided in the encoder wether or not they need to be cleared explicitly or not, since some will just expire and need not be cleared explicitly. However, clearing them really means (re)defining them with 0 rectangles, emitting an extra clear shouldn't harm
[10:28] <sspiff> let me read that comment again, it said something about a bug with to many clears, right? :)
[10:30] <ubitux> i have difficulties to understand joolz most of the time so don't ask me :p
[10:34] <sspiff> I think what joolz is suggesting is the same thing I have in mind :)
[10:35] <sspiff> but I'm not entirely sure :)
[10:36] <sspiff> to be honest, his distinction of "sending a clear only when the timeout hasn't expired" is not entirely correct
[10:36] <sspiff> I think
[10:37] <sspiff> because the spec says that the timeout is specified in seconds and only has to be respected with a +0/+5s accuracy
[10:37] <sspiff> so we shouldn't rely on it to clear subtitles - we should always emit a clear BEFORE the timeout expires
[10:37] <plepere> ubitux : wow, your pastie does magic !
[10:45] <HelloWorld1> Hello
[10:57] <HelloWorld1> I`m trying to decode h264 video with VAAPI. When checking this ` pCodec = avcodec_find_decoder_by_name ("h264_vaapi");
[10:57] <HelloWorld1> i`m getting unsupported codec error. 
[10:59] <JEEB> as far as I can see that hwaccel is a hwaccel now, as it should be
[11:01] <HelloWorld1> JEEB: Thanks for your reply. I`m new to FFMPEG. I want  to enable hwaccel when decoding h264 video. 
[12:37] <cone-48> ffmpeg.git 03Georg Martius 07master:2727dbb05345: lavfi/vidstabdetect,vidstabtransform: update to vid.stab 0.98
[12:37] <cone-48> ffmpeg.git 03Georg Martius 07master:7012a9dc10b9: configure: add version check for pkg libraries, and use it for vidstab
[12:37] <cone-48> ffmpeg.git 03Stefano Sabatini 07master:a64333db5fef: lavfi/vidstabtransform: apply various cosmetical nits
[12:59] <BBB> ubitux: apply mine also ;)
[13:12] <ubitux> BBB: can you first fix the comments here?
[13:12] <ubitux> +    VP9_UNPACK_MULSUB_2D_4X 10, 11,  6,  7, 15893,  3981    ; m3/6=t3[d], m2/7=t2[d]
[13:12] <ubitux> +    VP9_UNPACK_MULSUB_2D_4X  3,  2,  4,  5,  8423, 14053    ; m10/4=t11[d], m11/5=t10[d]
[13:13] <BBB> oh I broke it again?
[13:13] <BBB> yes send a comment so I remember
[13:14] <BBB> so, for lf, you asked about the mmx versions
[13:14] <BBB> I would first focus on the 4_16 and 8_16, these are relatively simple given that you did 16_16 already
[13:14] <BBB> (so 44_16 and 88_16)
[13:14] <BBB> then I'd do the mix2 functions (i.e. 48_16 and 84_16)
[13:15] <BBB> these are essentially identical to 88_16 except with a fixed zero mask for the filter6 for the 4 portion
[13:15] <BBB> I'd leave the mmx versions (4_8, 8_8 and 16_8) for last then
[13:15] <ubitux> ok
[13:15] <ubitux> thanks :)
[13:28] <BBB> (also when I say mmx, I mean '8 pixel per call', I don't mean 'do them in mmx' - you _could_, but you'll have less regs, so if that's hard, obviously just do it in xmm and ignore the upper half)
[13:29] <BBB> oh I remember that comment part
[13:29] <BBB> oops
[13:29] <BBB> lemmefix that
[13:30] <BBB> done
[13:30] <ubitux> BBB: is it in a branch?
[13:31] <ubitux> (simpler for me :P)
[13:31] <BBB> vp9-simd
[13:31] <BBB> https://github.com/rbultje/ffmpeg/commit/d3b32220ca59c92a7ecd80b1caa6cf7a09402278
[13:32] <ubitux> i'll apply soon
[13:34] <ubitux> for some reason yasm is in infinite loop on vp9itxfm.o
[13:34] <ubitux> ah, done
[13:34] <ubitux> wth
[13:34] <ubitux> :))
[13:36] <ubitux> yasm -f elf -m amd64 -g dwarf2 -I./ -I.// -Pconfig.asm -I libavcodec/x86/ -M   32.92s user 0.03s system 100% cpu 32.943 total
[13:36] <ubitux> BBB: can't you do something about it? :D
[13:36] <ubitux> 30+ seconds of assembling sound a bit huge
[13:37] <ubitux> that's weird, it's not caused by your patch
[13:37] <ubitux> oh well
[13:37] <ubitux> yasm -f elf -m amd64 -g dwarf2 -I./ -I.// -Pconfig.asm -I libavcodec/x86/ -M   13.06s user 0.01s system 100% cpu 13.076 total
[13:39] <ubitux> (without your patch)
[13:40] <nevcairiel> that patch increases assembly time nearly 3 fold?
[13:40] <ubitux> yes
[13:40] <ubitux> 13s is quite huge actually
[13:40] <nevcairiel> might be time for a file split then
[13:40] <ubitux> probably too much macro madness
[13:40] <ubitux> nevcairiel: that's hard to split
[13:41] <JEEB> I've noticed that x265's asm files with a lot of macros tend to be slow to compile as well
[13:41] <ubitux> i think that's because there are too much macro within macro
[13:41] <ubitux> probably the %rep ones being evaluated a lot of time
[13:41] <ubitux> can we have some stats profiling with yasm?
[13:43] <JEEB> also it seems like the yasm creator just doesn't care about that one commit https://github.com/yasm/yasm/commits/master
[13:43] <JEEB> also it's kind of sad that yasm "nextgen" seems to be actually more dead than yasm itself https://github.com/yasm/yasm-nextgen
[13:44] <ubitux> what commit?
[13:44] <JEEB> the top one
[13:44] <JEEB> "test commit"
[13:44] <ubitux> adding a random space
[13:44] <ubitux> :D
[13:47] <BBB> yes I've noticed it
[13:47] <BBB> I ... couldn't really care less, tbh
[13:47] <BBB> but I can work on it if you think it's important
[13:47] <BBB> I don't really know what causes it
[13:47] <BBB> yes we use a lot of macros, but so did vp8, and that's ok'ishly fast
[13:48] <ubitux> i'm thinking about the zero %rep thing
[13:48] <BBB> we can ask our local yasm overlord pengvado
[13:48] <BBB> ah right
[13:48] <BBB> we can unroll that I guess i that's it
[13:49] <ubitux> there are 2 nested loops
[13:49] <BBB> have you seen how many nested loops a vp9/hevc decoder is in when it gets to actual reconstruction of a 4x4 idct?
[13:49] <BBB> that didn't take 30 seconds
[13:49] <BBB> :-p
[13:50] <ubitux> yasm -f elf -m amd64 -g dwarf2 -I./ -I.// -Pconfig.asm -I libavcodec/x86/ -M   29.46s user 0.03s system 100% cpu 29.486 total
[13:50] <ubitux> ok not that.
[13:50] <BBB> for (frame : frames) { for (h and v tile : tiles) { for (y : rows) { for (x : cols) { for (h/v 4x4 block : blocks) { idct4x4(); } } } } }
[13:50] <BBB> so 7 nested loops
[13:51] <ubitux> BBB: i mean the nested loop in ZERO_BLOCK
[13:51] <ubitux> :p
[13:51] <BBB> I know
[13:51] <BBB> but that was not it ;)
[13:51] <ubitux> yes :(
[13:51] <ubitux> 30 seconds is really slow
[13:52] <BBB> agreed
[13:52] <BBB> so my patch makes it slow, or it was always slow?
[13:52] <ubitux> 13 -> 33 sec
[13:53] <BBB> hum
[13:53] <BBB> pengvado: Skyler_: any thoughts on that? (yasm being slow, see ^^)
[13:53] <ubitux> [context] after this: https://github.com/rbultje/ffmpeg/commit/d3b32220ca59c92a7ecd80b1caa6cf7a09402278
[14:21] <BBB> I really don't want to split that file, it's very interleaved :-p
[14:22] <BBB> I suppose you could possibly split it by size, but if 16x16 only made it 13->30 already, that only helps so much...
[14:25] <BBB> we should still apply the patch :-p
[14:25] <BBB> how's lf going btw? :)
[14:26] <BBB> I'll probably do iadst8_1d and iadst4_1d, and then work on the context setting in decode_mode
[15:15] <cone-48> ffmpeg.git 03Ronald S. Bultje 07master:8173d1ffc0b7: vp9/x86: 16x16 iadst_idct, idct_iadst and iadst_iadst (ssse3+avx).
[15:15] <ubitux> BBB: applied, but we really need to do something about assembling time
[15:15] <ubitux> that is really way too slow
[16:01] <cone-48> ffmpeg.git 03Ronald S. Bultje 07fatal: ambiguous argument 'refs/tags/n1.2.5': unknown revision or path not in the working tree.
[16:01] <cone-48> Use '--' to separate paths from revisions
[16:01] <cone-48> refs/tags/n1.2.5:HEAD: vp9/x86: 16x16 iadst_idct, idct_iadst and iadst_iadst (ssse3+avx).
[17:13] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:534a89478d42: avcodec/huffyuv: add a field that represents the number of symbols
[17:13] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:6c004e8aada0: avcodec/huffyuvenc: init bps from pixel descriptor
[17:13] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:b53aab1a5819: libavcodec/huffyuv: >8 bit support
[17:36] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:ef0282125cf4: avcodec/huffyuv: add YUV{420,422,444}P{9,10,12,14} support
[17:49] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:3d87927959ba: avcodec/huffyuv: add YUVA{420,422,444}P{9,10} support
[17:54] <ubitux> no btc on bountysource, too bad
[18:11] <ramiro> BBB: I get yasm 1.2.0 hanging on vp9itxfm.asm. Does it just take exceptionally long or is there something wrong?
[18:18] <saste> ubitux, what's the fuzz about bitcoins?
[18:18] <saste> good for speculators? or the amount of energy -> pollution needed to fabric them?
[18:20] <JEEB> ramiro, it just takes time as far as I know.
[18:20] <JEEB> Similar occurence as with x265 it seems
[18:27] <ramiro> JEEB: it's been a few minutes already
[18:28] <ramiro> oh, wow. that too exceptionally long... but it's done.
[18:28] <ramiro> I'm on an atom, so it's not a proper CPU anyways...
[18:35] Action: michaelni understands why optimizing compilers can be slow but an assembler !? :/
[18:39] <cone-48> ffmpeg.git 03Michael Niedermayer 07master:2fd39642c9aa: avcodec/huffyuv: fix median prediction for >8bps
[18:44] <michaelni> hmm, building vp9itxfm.o takes 50sec on a otherwise idle 4ghz i7
[18:46] <michaelni> is this is joke ? the box could build the whole ffmpeg wthout ccache in less than 20sec with the right compiler and flags
[18:46] <JEEB> some parts seem to be rather unoptimized @ yasm
[18:50] <Compn> saste : its just easy to transfer once you have them. unlike banks, paypal, checks, credit cards, etc
[18:50] <ubitux> saste: not subject to taxes and papers
[18:51] <ubitux> basically, that's a huge advantage if you want to do freelance jobs without having to mess with the legal aspect
[18:51] <ubitux> that's also extremely simple to deal with
[18:51] <Compn> its subject to taxes, just you are hiding it from government ubitux :P
[18:51] <Compn> ehe
[18:51] <ubitux> yeah, well, it's extremely hard to track
[18:51] <ubitux> and no one will bother for now
[18:51] <Compn> its tied to your ip addy ! :P
[18:51] <Compn> >use tor
[18:52] <ubitux> i'm not in the usa, they won't call the fbi if i get a few btc
[18:52] <Compn> true
[18:52] <ubitux> michaelni: 30 sec on my laptop, yeah that's huge
[18:52] <ubitux> BBB, michaelni, ramiro, should we revert for now?
[18:53] <Compn> i forgot i was going to setup btc donations on ffmpeg
[18:53] <Compn> if everyone wanted me to handle btc donations anyhow
[19:11] <michaelni> ubitux, no idea, but if possible we should workaround that yasm problem if possible
[19:11] <ubitux> i have no idea where it comes from
[19:11] <michaelni> nasm here takes 0m3.660s
[19:11] <michaelni> for time make V=2 libavcodec/x86/vp9itxfm.o after a rm libavcodec/x86/vp9itxfm.o
[19:21] <kierank> 5:50 PM <"ubitux> saste: not subject to taxes and papers
[19:21] <kierank> 5:51 PM <"ubitux> basically, that's a huge advantage if you want to do freelance jobs without having to mess with the legal aspect
[19:21] <kierank> 5:51 PM <"ubitux> that's also extremely simple to deal with
[19:21] <kierank> 5:51 PM <"Compn> its subject to taxes, just you are hiding it from government ubitux :P
[19:21] <kierank> this ^
[19:21] <kierank> baaaad idea
[19:21] <kierank> even mobsters get screwed for not paying taxes
[19:22] <michaelni> ubitux, correction its 5.1 sec here with nasm, was mistakely testing the previous commit
[19:22] <ubitux> michaelni: still decent
[19:22] <Compn> kierank is the white knight :D
[20:57] <pengvado> BBB: you're running into this bug: http://akuvian.org/paste/ee7e76e8.diff
[21:01] <JEEB> ooh
[21:03] <JEEB> I'll need to apply that too for my yasm
[21:03] <pengvado> (which I never submitted upstream because it's a kludge, but it works)
[21:04] <ubitux> pengvado: any workaround?
[21:04] <ubitux> reduce the number of macro?
[21:05] <ubitux> there are only 27 macro in that file
[21:05] <ubitux> i guess it's including the ones from x264 as well
[21:06] <pengvado> yes, the ones in x86inc.asm that define new temporary variables for every instruction instantiated
[21:08] <pengvado> using %% to get local variables doesn't actually limit their scope, they still go in the global hash of macro names
[21:09] <ubitux> N times with a different name?
[21:09] <pengvado> yes
[21:10] <ubitux> let's see if taking it out helps...
[21:11] <ubitux> not enough it seems :p
[21:12] <pengvado> http://akuvian.org/paste/434723e7.diff provides almost as much speedup as fixing yasm
[21:14] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.6:0cbef38cb818: arm: Don't clobber callee saved registers in scalarproduct
[21:14] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.6:792ee2f7750c: Merge remote-tracking branch 'qatar/release/0.6' into release/0.6
[21:14] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.6:9a073b1f0f27: avcodec/parser: reset indexes on realloc failure
[21:14] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.6:101e1b36cccd: avcodec/jpeglsdec: check err value for ls_get_code_runterm()
[21:16] <ubitux> ah that's nice
[21:29] <Daemon404> a wild pengvado 
[21:43] <cone-48> ffmpeg.git 03Reinhard Tartler 07release/0.10:d4a24e43edd0: alsdec: check block length
[21:43] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:b5736759eefa: qdm2: check array index before use, fix out of array accesses
[21:43] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:6f6cd7dbe539: roqvideodec: check dimensions validity
[21:43] <cone-48> ffmpeg.git 03Reinhard Tartler 07release/0.10:b997a6a86d7a: Prepare for 0.8.10 Release
[21:43] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:b6fc0127ce84: arm: Don't clobber callee saved registers in scalarproduct
[21:43] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:e01d623e0193: dsicin: Add some basic sanity checks for fields read from the file
[21:43] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:61032c577db0: smacker: Don't return packets in unallocated streams
[21:43] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:68a1df13c460: smacker: Avoid integer overflow when allocating packets
[21:44] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:85c02da30768: cmdutils: update year
[21:44] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:9e96051d5d03: Merge commit '68a1df13c460adb6241cfdf96aad953b5d637623' into release/0.10
[21:51] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:d5c104c1ae56: smacker: Make sure we don't fill in huffman codes out of range
[21:51] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:712945d21e63: shorten: Use a checked bytestream reader for the wave header
[21:51] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:7d8a4bb8d2ca: shorten: Break out of loop looking for fmt chunk if none is found
[21:51] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:f241d5aa1fa4: ape: Don't allow the seektable to be omitted
[21:51] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:b81d804f2ac1: zmbvdec: Check the buffer size for uncompressed data
[21:51] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:486c45767587: mpc8: Check the seek table size parsed from the bitstream
[21:51] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:bd953f94044e: Merge commit '486c45767587151b517bb6fde602d43d178da203' into release/0.10
[21:57] <JEEB> ugh, did they break yasm's compilation, too?
[21:57] <JEEB> I get this after configuration:
[21:57] <JEEB> $ make all
[21:57] <JEEB> make: *** No rule to make target `x86insns.c', needed by `all'.  Stop.
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:9291012d52a7: mpc8: Make sure the first stream exists before parsing the seek table
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:7981b5c20e61: omadec: Properly check lengths before incrementing the position
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:b9e90b36cd3b: sierravmd: Do sanity checking of frame sizes
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:7f80928c0e7a: cavsdec: Make sure a sequence header has been decoded before decoding pictures
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:290783b84866: vp3: Check the framerate for validity
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:802deb2d136f: svq3: Check for any negative return value from ff_h264_check_intra_pred_mode
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:48f27c854f23: mace: Make sure that the channel count is set to a valid value
[21:57] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:75f811babc08: matroskadec: Verify realaudio codec parameters
[21:58] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:f0db793bee82: rv34: Check the return value from ff_rv34_decode_init
[21:58] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:33b88170d76a: oggparseogm: Convert to use bytestream2
[21:58] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:843b330c3c9a: ivi_common: Make sure color planes have been initialized
[21:58] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:e3ba6ff9357d: mpeg4videodec: Check the width/height in mpeg4_decode_sprite_trajectory
[21:58] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:b19eafa2b930: eacmv: Make sure a reference frame exists before referencing it
[21:58] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:5339a9f00051: Merge commit 'b19eafa2b930ee40abfde6d1f026b7fa5591c4dc' into release/0.10
[22:07] <JEEB> ok, got it to start building after manually calling python
[22:08] <ubitux> python? huh?
[22:08] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:2ef84218b292: truemotion2: Use av_freep properly in an error path
[22:08] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:2b71a7884196: ffv1: Make sure at least one slice context is initialized
[22:08] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:0a23055b8ab2: xmv: Add more sanity checks for parameters read from the bitstream
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:3b169044ca04: rv10: Validate the dimensions set from the container
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:75dabbff8b15: idroqdec: Make sure a video stream has been allocated before returning packets
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:036136fa89ac: asv1: Verify the amount of extradata
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:ff8837e9c60a: mpegaudiodec: Validate that the number of channels fits at the given offset
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:2c3114158510: qpeg: Add checks for running out of rows in qpeg_decode_inter
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:e80071892b14: segafilm: Validate the number of audio channels
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:f23b1cc7d979: wtv: Add more sanity checks for a length read from the file
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:c211ba9b59e8: rl2: Avoid a division by zero
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:456a9392103f: dca: Validate the lfe parameter
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:b29c31c21e6d: wnv1: Make sure the input packet is large enough
[22:09] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:159993acc7f4: vc1dec: Fix leaks in ff_vc1_decode_init_alloc_tables on errors
[22:09] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:d1c7a7776f67: Merge commit '159993acc7f4e3155510d42c543e09fe972b933c' into release/0.10
[22:09] <Daemon404> JEEB, ... are you using yasm git?
[22:09] <Daemon404> (dont)
[22:09] <JEEB> ubitux, yeah -- it needs to create a c file with python 2.x
[22:10] <JEEB> Daemon404, it has some fixes from D_S too
[22:10] <Daemon404> like?
[22:10] <JEEB> might as well
[22:10] <JEEB> https://github.com/yasm/yasm/commit/a8495ad151b6db29f633d8e005e9049c7949cd05
[22:10] <Daemon404> id rather use a known working ver ;)
[22:10] <JEEB> and not that there's many commits since https://github.com/yasm/yasm/commits/master
[22:11] <JEEB> (although yasm is surprisingly less dead than yasm-nextgen
[22:11] <Daemon404> well
[22:11] <Daemon404> that commit is entirely irrelevant
[22:11] <Daemon404> since no such code exists
[22:11] <Daemon404> afaik
[22:16] <JEEB> also there are some issues resolved looking at some of the commit messages, so now that yasm's last release is like ages behind, I might as well try using HEAD
[22:17] <JEEB> if it breaks then I'll fall back to 1.2.0
[22:29] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:cd254e8540b3: vc1dec: Undo mpegvideo initialization if unable to allocate tables
[22:29] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:61d56054a9d7: vqf: Make sure sample_rate is set to a valid value
[22:29] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:efe59ad90bce: vqf: Make sure the bitrate is in the valid range
[22:29] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:889bdc47f6bb: avidec: Make sure a packet is large enough before reading its data
[22:29] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:ea1c9424d1c4: xwma: Avoid division by zero
[22:29] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:35c39d2ee2a2: mvi: Add sanity checking for the audio frame size
[22:30] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:7fdb915fc712: riffdec: Add sanity checks for the sample rate
[22:30] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:416ad3ecf242: electronicarts: Add more sanity checking for the number of channels
[22:30] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:5e8243e843be: bfi: Avoid divisions by zero
[22:30] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:6c0c799bd586: Merge commit '5e8243e843be8f71da7dde199a71d095726533eb' into release/0.10
[22:31] <JEEB> Daemon404, another D_S's fix btw https://github.com/yasm/yasm/commit/31c99f1fb24a893e0bd7778ba165a779474246b5
[22:36] <Daemon404> very descriptive message
[22:36] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:213b8aa0a905: bfi: Add some very basic sanity checks for input packet sizes
[22:36] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:1438181a2943: mov: Make sure the read sample count is nonnegative
[22:36] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:ac1e61d5593d: pngdec: Stop trying to decode once inflate returns Z_STREAM_END
[22:36] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:5a40e4c64d90: pcx: Consume the whole packet if giving up due to missing palette
[22:37] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:993977032a0a: xan: Use bytestream2 to limit reading to within the buffer
[22:37] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:76c48a78d1c0: Merge commit '993977032a0adb47eb70e7fef6ce0d5370027e83' into release/0.10
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:380e37326768: xan: Only read within the data that actually was initialized
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:d3986f4f1baf: xxan: Disallow odd width
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:cb4a101fbe27: rpza: Fix a buffer size check
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:d92c908e235a: pcx: Check the packet size before assuming it fits a palette
[22:49] <cone-48> ffmpeg.git 03Anton Khirnov 07release/0.10:90294e31a199: mxfdec: set audio timebase to 1/samplerate
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:e972338e3596: asfdec: Check the return value of asf_read_stream_properties
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:591d5281f5bc: twinvqdec: Check the ibps parameter separately
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:871baf312791: rmdec: Validate the fps value
[22:49] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:49c1defee522: svq3: Avoid a division by zero
[22:50] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:fbc52044f3d0: fraps: Make the input buffer size checks more strict
[22:50] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:29fa517d40ed: r3d: Add more input value validation
[22:50] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:9925f7df0a50: vc1dec: Make sure last_picture is initialized in vc1_decode_skip_blocks
[22:50] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:93e3ec451caf: Merge commit '9925f7df0a50387ade8d83cb85b40c53e41e7041' into release/0.10
[23:26] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:c00e491aebec: vc1dec: Don't decode slices when the latest slice header failed to decode
[23:26] <cone-48> ffmpeg.git 03Martin Storsjö 07release/0.10:371659d1ad71: mpegvideo: Initialize chroma_*_shift and codec_tag even if the size is 0
[23:27] <cone-48> ffmpeg.git 03Derek Buitenhuis 07release/0.10:48d57650f121: pthread: Fix deadlock during thread initialization
[23:27] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:a26bfc444d25: Merge commit '48d57650f121d3d9e977832e9006bb334337d921' into release/0.10
[23:34] <Zeranoe> oes anyone know what kind of material I'm dealing with here: http://paste.debian.net/76683/ I added the line breaks at logical spots, but the patterns might be longer. I can resort it if it helps. I want to know how to properly deinterlace it
[23:35] <Zeranoe> does*
[23:36] <cone-48> ffmpeg.git 03Ben Jackson 07release/0.10:112387087983: pthread: Avoid spurious wakeups
[23:36] <cone-48> ffmpeg.git 03Sean McGovern 07release/0.10:3736b1375392: Changelog for 0.8.10
[23:36] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:86a423d2b8c6: Merge remote-tracking branch 'qatar/release/0.8' into release/0.10
[23:38] <cone-48> ffmpeg.git 03Michael Niedermayer 07release/0.10:41216ebb9e53: update for 0.10.11
[00:00] --- Fri Jan 17 2014