[Ffmpeg-devel-irc] ffmpeg-devel.log.20131006

[burek]

[00:03] <JEEB> <JEEB> doom9 is compromised
[00:03] <JEEB> <JEEB> disabling JS seems "OK" for now, but there is no way of knowing if they got access to an admin as well, and not only a moderator
[00:03] <JEEB> <JEEB> if you had opened a (sub)forum and it took a while and had an announcement, and you had JS enabled, you can be relatively sure that your username/password combo got sent out to a 3rd party server
[00:44] <cone-459> ffmpeg.git 03Michael Niedermayer 07master:6853e40106ca: avformat/matroskadec: only set r_frame_rate if the value is within reasonable limits
[01:45] <Compn> put doom9 news on the site
[01:45] <Compn> luckily i use throwaway passwords on most things :)
[01:46] <Compn> now wheres lou so he can make a tweeet about it :P
[03:33] <rcombs> who can I pay, and how much, to fix the dithering-causes-banding-and-discoloration issue?
[03:52] Action: rcombs wonders if a Riemersma dither could be done fast enough
[03:59] <Compn> rcombs : open a trac ticket with a sample and command line
[03:59] <Compn> and slap a bounty on it
[04:04] <rcombs> Compn: already got a ticket (someone else filed, I added some sample images)
[04:04] <rcombs> not sure how best to stick a bounty on an issue
[04:05] <Compn> "$500 to person who fixes" ?
[04:15] <rcombs> ^ relevant
[04:16] <Compn> sounds good
[04:16] <Compn> i'm kind of not paying attention here :D
[04:16] <Compn> ehe
[04:19] <Compn> rcombs : btw can you do --disable-optimizations or possibly --disable-yasm and rerun your tests ?
[04:19] <rcombs> Compn: sure, gimme a minute
[04:19] <Compn> wonder if its asm related
[04:21] Action: rcombs facepalms; forgot that was a configure option
[04:22] <Compn> yeah one of those
[04:22] <Compn> recompile
[04:23] <rcombs> yup
[04:23] <Compn> so uh... can you make a small sample file
[04:23] <rcombs> oh joy, a compile error
[04:23] <Compn> i dont see one ...
[04:23] <Compn> oh there it is
[04:23] <Compn> vtst0.mp4
[04:24] <rcombs> Compn: also, if you take the rgb48 PNG I snapshotted, it makes a good sample source
[04:24] <Compn> ok the png source shows the same thing ? :)
[04:25] <rcombs> the RGB48 PNG looks good; using ffmpeg to convert the RGB48 to RGB24 shows the issue
[04:26] Action: rcombs ran out of registers during register allocation
[04:30] <rcombs> apparently clang doesn't like something or other that shows up when optimizations are disabled
[04:31] <rcombs> and& gcc apparently screws up its output
[04:32] Action: rcombs checks out n2.0.1; still has issue
[04:33] <rcombs> libavcodec/x86/dsputil_mmx.c:266:9: error: ran out of registers during register allocation <-- joy
[04:33] <Compn> add -fomit-frame-pointer ? maybe ?
[04:33] <Compn> i forgot clang 
[04:33] <Compn> its old error i mean, but i dont know if that works with clang
[04:34] Action: rcombs brings himself back to master
[04:35] <rcombs> nope
[04:37] Action: rcombs tries some more stuff
[04:37] <rcombs> there we go
[04:38] <rcombs> for the record, my successful configure line was ./configure --enable-gpl --enable-version3 --enable-nonfree --enable-postproc --enable-libass --enable-libfdk-aac --enable-libfreetype --enable-libmp3lame --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-openssl --enable-libtheora --enable-libvorbis --enable-libvpx --enable-libx264 --enable-libxvid --prefix=/usr/local --enable-static --enable-shared --cc="clang -m32 -fomit
[04:38] <rcombs> -frame-pointer" --disable-optimizations --disable-yasm --disable-asm
[04:39] <Compn> cool
[04:43] <rcombs> http://dl.dropboxusercontent.com/u/46348978/Madoka_Snapshot_02_24_noasm.png <-- no change in output
[05:11] <Compn> ok thanks for checking
[05:11] <Compn> was a dumb idea
[05:12] <Compn> theres dithering in there ?
[05:12] Action: Compn doesnt even know what dithering means
[05:12] <Compn> looks like luma is off
[05:12] <rcombs> it is, yup
[05:13] <rcombs> dithering (in this case) means improving the quality of color quantization by& it's hard to explain; the Wikipedia article has some great images demonstrating this
[10:53] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:ccf96f8c642d: swscale/options: switch default to bicubic
[11:12] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:931a4b433dce: avfilter/vf_scale: change the default scaler to bicubic
[11:24] <cone-291> ffmpeg.git 03Luca Barbato 07master:4272bb6ef153: doxy: Document avlog
[11:24] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:90cecd3c9b6b: Merge commit '4272bb6ef1533846a788c259cc498562d0704444'
[11:30] <cone-291> ffmpeg.git 03Diego Biurrun 07master:0b8b2ae5e93d: x86: xviddct: Employ more specific ifdefs
[11:30] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:d0b27036763f: Merge commit '0b8b2ae5e93d616c2ece59f7175f483154cff918'
[11:36] <cone-291> ffmpeg.git 03Diego Biurrun 07master:258414d07718: x86: fdct: Initialize optimized fdct implementations in the standard way
[11:36] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:7fb123429e6c: Merge commit '258414d0771845d20f646ffe4d4e60f22fba217c'
[11:49] <cone-291> ffmpeg.git 03Vittorio Giovara 07master:6b3ff6f91a53: swscale: provide a default scaler if none is set
[11:49] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:5710f55e490d: Merge commit '6b3ff6f91a535d6383f41ca7bdf760165dcb6015'
[11:55] <cone-291> ffmpeg.git 03Diego Biurrun 07master:2ddb35b91131: x86: dsputil: Separate ff_add_hfyu_median_prediction_cmov from dsputil_mmx
[11:55] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:c86955d24a43: Merge commit '2ddb35b91131115c094d90e04031451023441b4d'
[11:59] <ubitux> BBB: in the itxfm_wrapper(), why is the input block reset to 0?
[12:00] <ubitux> is it simpler for the caller then?
[12:08] <ubitux> well it seems useful anyway.
[12:15] <ubitux> there are not much additionnal int in ssse3, but some look very useful to me&
[12:33] <cone-291> ffmpeg.git 03Vittorio Giovara 07master:90e15e340d97: oggparsevorbis: return meaningful errors
[12:33] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:4634871b513c: Merge remote-tracking branch 'qatar/master'
[12:40] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:3ee796760a89: avfilter/vf_scale: dont duplicate default.
[12:50] <BBB> ubitux: someone has to reset the block contents to zero
[12:51] <BBB> ubitux: otherwise the block reader has to do it after reading the end-of-block
[12:51] <ubitux> is it always necessary?
[12:51] <BBB> only for the ones that are non-zero
[12:51] <ubitux> yeah well, ok
[12:51] <BBB> so you really only need to memset(block, 0, sizeof(*block)*eob);
[12:51] <BBB> but I didn't do that :)
[12:51] <BBB> feel free to change the C code that way
[12:51] <ubitux> ah the eob thing
[12:51] <ubitux> ok
[14:52] <ubitux> SUMSUB_BA should be used with a temporary register when available right?
[15:02] <durandal_1707> wow. doom9 hacked
[15:30] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:4916088a9302: fate: fix dpx on big endian
[15:33] <JEEB> durandal_1707, at the very least a moderator's account was compromised, in the worst case they also got access to the server
[15:34] <JEEB> the most visible part of it was that JS script embedded in an announcement that was shown on every subforum
[15:35] <durandal_1707> so it was because forum software was not updated?
[15:36] <JEEB> well, the forum software letting anyone put a script tag into a title (or anything for that matter) is fucked up, but I have no idea if the updates would have fixed that
[15:36] <JEEB> I think they were more or less up-to-date with the 3.x branch
[15:36] <JEEB> which is, granted, quite old
[16:48] <durandal_1707> why was lavc minor not bumped when vp9 was added?
[16:49] <ubitux> likely a mistake
[16:49] <durandal_1707> well not minor but micro
[16:49] <durandal_1707> minor is bumped when new codec id is added
[16:50] <durandal_1707> or whatever
[16:52] <durandal_1707> the commit check script (which check for tabs iirc) could check for this too
[16:56] <cone-291> ffmpeg.git 03Michael Niedermayer 07master:efff7564d4d6: configure: filters should select the parts they need not use deps
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:eec833b10dda: avcodec/parser: reset indexes on realloc failure
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:510da4fe2b45: apichanges: fix 2 wrong hashes
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:ba7cd748c179: doc/APIchanges: List merge commit hashes and version numbers
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:ff29290e265d: update all trac links to use the trac subdomain
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:0dff3171ce10: avcodec/rpza: Perform pointer advance and checks before using the pointers
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:38ca79b04d9a: jpeg2000: check log2_cblk dimensions
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:1b05b0005bec: matroska_read_seek: Fix used streams for subtitle index compensation
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:e3d643cf7587: avcodec/pngdsp: fix (un)signed type in end comparission
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:90ee388b2883: avcodec/dsputil: fix signedness in sizeof() comparissions
[17:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:bd66456866e9: avcodec/ffv1enc: update buffer check for 16bps
[18:17] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.8:a248117f2618: update for 0.8.15
[18:21] <durandal11707> Compn: why you put that on ffmpeg news?
[18:47] <cone-291> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n0.8.15': unknown revision or path not in the working tree.
[18:47] <cone-291> Use '--' to separate paths from revisions
[18:47] <cone-291> refs/tags/n0.8.15:HEAD: configure: filters should select the parts they need not use deps
[19:06] <cone-291> ffmpeg.git 03Vicente Jimenez Aguilar 07release/0.7:c10582e703c5: doc: filters: Correct BNF FILTER description
[19:06] <cone-291> ffmpeg.git 03Paul B Mahol 07release/0.7:537c17385350: smacker: fix off by one error
[19:06] <cone-291> ffmpeg.git 03Carl Eugen Hoyos 07release/0.7:e2d529424f3f: Fix type of shared flac table ff_flac_blocksize_table[].
[19:06] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:b6783b882682: indeo5: update AVCodecContext width/height on size change
[19:06] <cone-291> ffmpeg.git 03Luca Barbato 07release/0.7:ee6b868ac835: oggdec: make sure the private parse data is cleaned up
[19:06] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:8489c0599f0e: update year to 2013
[19:06] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:f13f6f82c63f: wmadec: require block_align to be set.
[19:06] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:7e6625a9afbe: xxan: fix invalid memory access in xan_decode_frame_type0()
[19:06] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:c313f3160ae1: Update RELEASE file for 0.7.8
[19:06] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:7d4c38d58d01: h264: check for luma and chroma bit depth being equal
[19:06] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:d86a5ce03f69: vorbisdec: Error on bark_map_size equal to 0.
[19:06] <cone-291> ffmpeg.git 03Luca Barbato 07release/0.7:5bfa208e6565: matroskadec: request a read buffer for the wav header
[19:06] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:5025dbc577c9: wmaprodec: return an error, not 0, when the input is too small.
[19:06] <cone-291> ffmpeg.git 03Justin Ruggles 07release/0.7:d785f6940144: shorten: validate that the channel count in the header is not <= 0
[19:06] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:a694b2b15865: shorten: set invalid channels count to 0
[19:06] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:5ebb5a32bdd9: shorten: report meaningful errors
[19:06] <cone-291> ffmpeg.git 03Luca Barbato 07release/0.7:6d4d186e9e9f: shorten: use the unsigned type where needed
[19:07] <cone-291> ffmpeg.git 03Mans Rullgard 07release/0.7:c65763a2c677: lavf: fix arithmetic overflows in avformat_seek_file()
[19:07] <cone-291> ffmpeg.git 03Xi Wang 07release/0.7:78889be3fb4b: atrac3: avoid oversized shifting in decode_bytes()
[19:07] <cone-291> ffmpeg.git 03Xi Wang 07release/0.7:f8d3bb8961d2: flacdec: simplify bounds checking in flac_probe()
[19:07] <cone-291> ffmpeg.git 03Mans Rullgard 07release/0.7:799000af702a: dsputil: fix invalid array indexing
[19:07] <cone-291> ffmpeg.git 03Xi Wang 07release/0.7:530d10792de2: lzo: fix overflow checking in copy_backptr()
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:9c713f30e491: parser: fix large overreads
[19:07] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:33c9e18b09f4: avfiltergraph: check for sws opts being non-NULL before using them.
[19:07] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:96cf80609af6: dfa: check for invalid access in decode_wdlt().
[19:07] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:280998b13cb8: wmaprodec: require block_align to be set.
[19:07] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:76c97f19634f: lavf: make sure stream probe data gets freed.
[19:07] <cone-291> ffmpeg.git 03Kostya Shishkov 07release/0.7:f844cb9bced3: iff: validate CMAP palette size
[19:07] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:9b052bfb8609: rv10: check that extradata is large enough
[19:07] <cone-291> ffmpeg.git 03Anton Khirnov 07release/0.7:96481c5e1843: qdm2: check that the FFT size is a power of 2
[19:07] <cone-291> ffmpeg.git 03Luca Barbato 07release/0.7:053c19cd88b6: oggdec: fix faulty cleanup prototype
[19:07] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:ade4f3e74635: aac: check the maximum number of channels
[19:07] <cone-291> ffmpeg.git 03Reinhard Tartler 07release/0.7:3197a9c4fa46: Update changelog for 0.7.8 release
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:d04dc7b5a78b: h264_cavlc: fix reading skip run
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:23d835f6113b: alacenc: Fix missing sign_extend()
[19:07] <cone-291> ffmpeg.git 03Carl Eugen Hoyos 07release/0.7:2a1bebfc83db: Autodetect idcin only if audio properties allow decoding.
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:c997dcd38be7: mpegts: only reopen pmt_cb filter if its different from the previous.
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:f8c4dbe45ee2: mpeg12dec: avoid reinitialization on PS changes when possible.
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:afe09e490a30: Merge commit '5025dbc577c9a9e0109cb363ac630a9eeda6dc1d' into release/0.8
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:0a41da3e9d20: Merge commit 'd785f6940144eb6ce4c24309ed034056b81395bc' into release/0.8
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:44ebb2556d09: Merge commit '5ebb5a32bdd910a8afb316c51ed0b322f5600ae5' into release/0.8
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:49d597f058a9: Merge commit '9c713f30e4913a28d93eb37ea5db7f62be4c0ef6' into release/0.8
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:70a1182a4844: Merge commit 'f844cb9bced3148fca2db5bbb092929526108005' into release/0.8
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:9b0736c08ac5: Merge remote-tracking branch 'qatar/release/0.7' into release/0.8
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:eec833b10dda: avcodec/parser: reset indexes on realloc failure
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:510da4fe2b45: apichanges: fix 2 wrong hashes
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:ba7cd748c179: doc/APIchanges: List merge commit hashes and version numbers
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:ff29290e265d: update all trac links to use the trac subdomain
[19:07] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:0dff3171ce10: avcodec/rpza: Perform pointer advance and checks before using the pointers
[19:08] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:38ca79b04d9a: jpeg2000: check log2_cblk dimensions
[19:08] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:1b05b0005bec: matroska_read_seek: Fix used streams for subtitle index compensation
[19:08] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:e3d643cf7587: avcodec/pngdsp: fix (un)signed type in end comparission
[19:08] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:90ee388b2883: avcodec/dsputil: fix signedness in sizeof() comparissions
[19:08] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:bd66456866e9: avcodec/ffv1enc: update buffer check for 16bps
[19:08] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:a248117f2618: update for 0.8.15
[19:08] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:0ca3dbbfcef0: Merge tag 'n0.8.15' into release/0.7
[19:08] Action: durandal11707 gonna ban cone-291 
[19:09] <cone-291> ffmpeg.git 03Michael Niedermayer 07release/0.7:63d5db37cba3: update for 0.7.16
[19:10] <rcombs> 0.o
[19:10] Action: rcombs didn't realize releases that old were still maintained at all
[19:13] <michaelni> rcombs, they arent really "maintained", iam just making new releases as some stuff has accumulated in the branches
[19:13] <rcombs> ah
[19:16] <cone-291> ffmpeg.git 03Michael Niedermayer 07fatal: ambiguous argument 'refs/tags/n0.7.16': unknown revision or path not in the working tree.
[19:16] <cone-291> Use '--' to separate paths from revisions
[19:16] <cone-291> refs/tags/n0.7.16:HEAD: configure: filters should select the parts they need not use deps
[20:24] <ubitux> BBB: are we supposed to play with special instruction for the rounding 14-bit rounding, or you manually add/shift just like in C?
[20:28] <BBB> ubitux: depends, most of the (a*x+b*y+round) >> shift should be done as such (pmaddwd for a*x+b*y, then paddd for round, and psrad+packuswb for shift -> word)
[20:28] <BBB> ubitux: some of them (like ((a+b)*x+rnd)>>shift) can use pmulhrsw if a or b is zero
[20:29] <BBB> ubitux: so e.g. for the dc only 4x4 case, you can use out=pmulhrsw(pmulhrsw(pmulhrsw(dc, 11585*2), 11585*2), 512)
[20:30] <BBB> ubitux: which is the same as the *11585, rounded-shift 14, *11585, rounded-shift 14, rounded-shift 6
[20:30] <ubitux> ok i can use the builtin-16 inst just for the t0/t1 cases in idct func
[20:31] <ubitux> thanks
[20:31] <BBB> for the 4x4/8x8, you can also use the pmulhrsw for the case where a and b are both nonzero (in ((a+b)*x+rnd)>>shift)
[20:31] <BBB> but that doesn't work for 16x16/32x32
[20:32] <ubitux> mmh
[20:33] <BBB> 4x4: see line 890-891
[20:33] <BBB> 8x8: see line 924-925 and 942-943
[20:33] <BBB> in vp9dsp.c
[20:33] <BBB> those can use pmulhrsw
[20:34] <ubitux> yes the t0/t1 i was mentioning
[20:34] <BBB> right
[20:34] <BBB> anything that looks like that in the 4x4/8x8
[20:34] <ubitux> BBB: what's the range of the values block values btw?
[20:34] <ubitux> they really are full 16-bits?
[20:35] <BBB> 4x4: 14bits, 8x8: 15bits, 16x16/32x32: 16bit
[20:35] <BBB> that's why it only works for the 4x4/8x8
[20:35] <BBB> for the 16x16/32x32, it can overflow in theory
[20:35] <ubitux> ah i see
[20:35] <BBB> (the add/sub)
[20:35] <BBB> that's why it can only be done if a or b is zero
[20:35] <BBB> not if both are nonzero
[20:35] <BBB> because the sub/add is 17bits
[20:35] <ubitux> ok
[20:35] <BBB> for 4x4/8x8, that's not a concern
[20:36] <BBB> (I think)
[20:36] <ubitux> no risk of signed 32-bit overflow with the mult in 16/32 btw?
[20:37] <BBB> shouldn't happen no
[20:40] <durandal_1707> imagemagick is full of shit
[20:50] <llogan> how do i show the changes in point releases in the online git browser?
[20:51] <durandal_1707> changes as?
[20:51] <llogan> as in the commits added for 0.8.15 from 0.8.14 for example
[20:52] <llogan> s/from/since/
[21:06] <cptspiff> llogan: http://git.videolan.org/?p=ffmpeg.git;a=shortlog;hp=<startrev>;h=<endrev>
[21:16] <llogan> cptspiff: thanks
[21:16] <BBB> ubitux: oh where I said packuswb, I of course meant packssdw
[21:16] <BBB> ubitux: silly me
[21:16] <BBB> (it's dword->word, not word->byte)
[22:22] <cone-291> ffmpeg.git 03Paul B Mahol 07master:8ec328668a7d: avcodec: add dpx parser
[22:22] <cone-291> ffmpeg.git 03Paul B Mahol 07master:ee3d03bf4ec6: avcodec/dpx: make .long_name more descriptive
[00:00] --- Mon Oct  7 2013