[Ffmpeg-devel-irc] ffmpeg-devel.log.20150224

burek burek021 at gmail.com
Wed Feb 25 02:05:03 CET 2015

[01:09] <cone-692> ffmpeg.git 03Michael Niedermayer 07master:1895d414aaac: swscale/utils: More carefully merge and clear coefficients outside the input
[01:09] <cone-692> ffmpeg.git 03Michael Niedermayer 07master:fb8e5044b420: swscale/utils: add asserts to check filterpos
[03:31] <cone-692> ffmpeg.git 03Michael Niedermayer 07master:8f4cbf940212: avcodec/snowdec: Fix ref value check
[03:31] <cone-692> ffmpeg.git 03Michael Niedermayer 07master:721c916dd55b: avcodec/snow: Free all planes and bands independent of the last count
[06:55] <kurosu> jamrial, michaelni: valgrind needed? sounds like not enough padding for the unconditional AVZERO (I thought there was enough padding)
[07:01] <jamrial> no, the segfaults were on scalaproduct_int16, and valgrind complained as such (invalid reads of 16 for sse2 version, 2 for c version)
[07:03] <kurosu> DECLARE_ALIGNED(16, int16_t, filter)[MAX_PREDICTORS];
[07:03] <kurosu> with padding, I meant enough room for the overreads
[07:04] <kurosu> so MAX_PREDICTORS+8
[07:05] <kurosu> or FF_PADDING_whatever/sizeof(uint16_t) if you want to be careful
[07:06] <kurosu> I don't understand why, when I sometimes do a git am, it doesn't leave any conflict markers, and just aborts
[07:06] <kurosu> anyway, gotta go
[07:07] <jamrial> valgrind still complains
[07:07] <jamrial> the segfault was reading s->residues btw, not s->filter
[07:13] <kurosu> ok
[07:14] <kurosu> well, I don't really now, as it seemed residues had enough padding
[07:24] <jamrial> kurosu: i think lines 462 and 631 aren't taking into account the padding from s->residues
[11:23] <cone-631> ffmpeg.git 03Clément BSsch 07master:cc69ccb6d440: avformat/nutdec: fix isom and mpa build deps
[11:33] <ubitux> with a rawvideo input, isn't avformat_find_stream_info/av_find_best_stream supposed to "find" something?
[11:34] <wm4> isn't it rather the opposite?
[11:35] <ubitux> ah my bad, i think i just forgot the rawvideo decoder
[11:35] <ubitux> yeah that was it, sorry for the noise
[12:20] <cone-631> ffmpeg.git 03Christophe Gisquet 07master:e87e20613b01: lavc/pthread_slice: release entries
[14:18] <cone-631> ffmpeg.git 03Paul B Mahol 07master:418be7ceb471: avcodec/mss4: check return value of init_get_bits8()
[15:22] <ubitux> wow
[15:22] <wm4> wow?
[15:22] <ubitux> so i'm rescaling a video without much color from 640x640 to 256x256
[15:22] <ubitux> the number of colors goes from 100k to...
[15:22] <ubitux> 200k
[15:22] <ubitux> wth o_o
[15:22] <wm4> and suddenly RAINBOWS?
[15:22] <wm4> aha
[15:22] <ubitux> yeah...
[15:22] <ubitux> i don't get it :p
[15:23] <wm4> also why are people editing the gsoc article, if it's clear as day that google will reject
[15:23] <compn> google hasnt told us specifically not to apply
[15:23] <compn> so we will continue to apply
[15:27] <Daemon404> compn, and third party candidates can keep running for president with about as much success
[15:28] <compn> you guys spend as much time complaining about what other devs work on as they spend working on gsoc applications.
[15:28] <compn> does the gsoc application process hurt you or the project in some way ?
[15:30] <Daemon404> im not comaplaining, im trolling
[15:30] <ubitux> it probably hurts the project a bit in the sense that it's probably a waste of resources
[15:31] <ubitux> but that's up to everyone to assign his time as he wishes to :p
[15:31] <Daemon404> indeed
[15:35] <rcombs> and hey, keep applying and maybe one year they'll loosen up
[15:36] <Daemon404> considering we're internally blacklisted...
[15:37] <compn> cant hurt to apply then.
[15:37] <compn> unless we get put on double secret probation!
[15:37] <compn> animal house is a fun film
[15:38] <Daemon404> i prefer futurama's versin
[15:38] <compn> i'm not a fan of comedy via references
[15:38] <Daemon404> thats not the comedy
[15:39] <Daemon404> i fully agree current fox toons are awful reference-piles
[15:58] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:847daac5c013: avformat/dss: Set packet position correctly
[16:00] <ubitux> ok i think i found the explanation at the exploding number of colors
[16:01] <ubitux> scale=256:-1:flags=bicubic,format=yuv444p ’ 205724 colors
[16:01] <ubitux> scale=256:-1:flags=bicubic,format=rgb32 ’ 64686 colors
[16:01] <ubitux> erm, swapped
[16:01] <ubitux> scale=256:-1:flags=bicubic,format=rgb32 ’ 205724 colors
[16:01] <ubitux> scale=256:-1:flags=bicubic,format=yuv444p ’ 64686 colors
[16:03] <ubitux> now the question is, how the hell 205724 can appears out of 108503 colors
[16:03] <rcombs> what's the source resolution/format?
[16:03] <ubitux> source is yuv444p, 640x640
[16:04] <rcombs> a better question might be how you ended up with more than 65536 colors in a 256x256 image
[16:05] <Daemon404> ubitux, 
[16:05] <Daemon404> movflags sure is confusing...
[16:05] <ubitux> rcombs: it's a video
[16:05] <rcombs> ah, well then
[16:05] <Daemon404> { "write_colr", "Write colr atom", 0, AV_OPT_TYPE_CONST, {.i64 = FF_MOV_FLAG_WRITE_COLR}, INT_MIN, INT_MAX, AV_OPT_FLAG_ENCODING_PARAM, "movflags" },
[16:05] <ubitux> rcombs: 6 seconds video
[16:05] <Daemon404> avoptions struct saus that the .i64 = <..> makes it default
[16:05] <Daemon404> but that is not he case
[16:05] <Daemon404> the*
[16:05] <Daemon404> the only way i can see is to edit the movflags entry default
[16:05] <Daemon404> which is butt ugly
[16:06] <ubitux> yeah that's what you need to do
[16:06] <Daemon404> meh
[16:06] <nevcairiel> i dont think you can set the default on flag-aliases, it needs to be set on the flag member itself
[16:06] <Daemon404> yeah i can see that now
[16:06] <Daemon404> it's non-ibviosu
[16:06] <ubitux> Daemon404: what's so confusing?
[16:06] <Daemon404> non-obvious*
[16:06] <Daemon404> ubitux, because the member  {.i64 = FF_MOV_FLAG_WRITE_COLR} is called "Default"
[16:07] <ubitux> yeah, it's actually the value itself for CONSTs
[16:07] <Daemon404> exactly
[16:07] <Daemon404> thats VERY non-obvious.
[16:07] <ubitux> but what you want to set default is the "movflags"
[16:07] <Daemon404> also not documented
[16:07] <ubitux> you'll get over it
[16:07] <ubitux> ;)
[16:08] <Daemon404> heh
[16:08] <Daemon404> also the movflags are indented to look like top level options
[16:08] <Daemon404> rather than flags
[16:08] <rcombs> I'd expect you're just getting the interpolated values, and with yuv444p you end up generating fewer distinct colors, maybe because the luma channel is somewhat more uniform?
[16:09] <ubitux> note that the number of color doesn't change if i do scale=640:-1,format=rgb32
[16:09] <ubitux> but it does if i scale=256:-1,format=rgb32
[16:09] <ubitux> (i have the "counting" filter just after the format, and it's taking only rgb32)
[16:10] <rcombs> yeah, it shouldn't unless there's dithering or something going on
[16:13] <ubitux> i'm really confused at what's going on but yeah, let's say it's dithering
[16:13] <ubitux> i don't see anything to disable it
[16:14] <ubitux> the 0 of sws_dither should be reserved for none
[16:15] <ubitux> it's actually auto
[17:37] <cone-631> ffmpeg.git 03Kevin Wheatley 07master:3664703780a3: avformat/movenc: Fix writing ACLR atoms for DNxHD
[18:26] <cone-631> ffmpeg.git 03Tobias Rapp 07master:f34aca8a1612: doc/filters.texi: add reference link to video size syntax description
[21:30] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:c23a0e77dd49: avcodec/h264: Only reinit quant tables if a new PPS is allowed
[21:44] <cone-631> ffmpeg.git 03Martin Storsjö 07master:26524e358147: rtsp: Interpret the text media type as AVMEDIA_TYPE_DATA
[21:44] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:c6792275f43c: Merge commit '26524e358147aade6e9dd18fff42d61b966bbc70'
[22:03] <cone-631> ffmpeg.git 03Gilles Chanteperdrix 07master:04a1be8e28e8: libavformat: add T.140 RTP depacketization (RFC 4103)
[22:03] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:8c1c06937156: Merge commit '04a1be8e28e81f3967eace7705343c450616cc95'
[22:10] <cone-631> ffmpeg.git 03Martin Storsjö 07master:3cd5828fc000: rtpdec: Add missing trailing commas
[22:10] <cone-631> ffmpeg.git 03Martin Storsjö 07master:74d318f138f2: rtsp: Fix the indentation of a linewrapped statement
[22:11] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:30c863fdfabe: Merge commit '3cd5828fc0007e6d056067444de053a34ddedd02'
[22:11] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:eb24186b457b: Merge commit '74d318f138f2a3f1b2fe81aea826d80d1e60f54c'
[22:19] <cone-631> ffmpeg.git 03Martin Storsjö 07master:fe414059ec1f: rtpdec_amr: Use the common indentation style
[22:19] <cone-631> ffmpeg.git 03Martin Storsjö 07master:247764a592f8: rtpdec_h264: Remove unnecessary struct padding
[22:19] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:5d06b6e98470: Merge commit 'fe414059ec1fe8237574bed34cdaa8ac3893917b'
[22:19] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:ae2a8d874236: Merge commit '247764a592f83521f9c1ccdc900b00a8b84f3c87'
[22:26] <cone-631> ffmpeg.git 03Martin Storsjö 07master:0752f700d63f: rtpdec_h264: Remove an unnecessary include
[22:26] <cone-631> ffmpeg.git 03Martin Storsjö 07master:353b492d0f2a: rtpdec: Change enc_name to a pointer instead of a fixed-size buffer
[22:26] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:a4f50ec9ddaa: Merge commit '0752f700d63f6b15ffd5369ac7615414cce71247'
[22:26] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:8c6651a14db2: Merge commit '353b492d0f2a21ae8eb829db1ac01b54b2a4d202'
[22:33] <cone-631> ffmpeg.git 03Martin Storsjö 07master:db158f0dd217: rtpdec: Remove unnecessary _if_needed suffixes on functions
[22:33] <cone-631> ffmpeg.git 03Martin Storsjö 07master:b651c9139e1a: rtpdec_mpa_robust: Move .enc_name to the start of the struct
[22:33] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:a51b192f3b8f: Merge commit 'db158f0dd217cf839be8af195d66cf49a76537a8'
[22:33] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:2bedcbf82274: Merge commit 'b651c9139e1ab222d5aab9151dcd7d6e40e49885'
[22:41] <cone-631> ffmpeg.git 03Martin Storsjö 07master:b7a4c319fda2: rtpdec: Allow setting the need_parsing field in RTPDynamicProtocolHandler
[22:41] <cone-631> ffmpeg.git 03Martin Storsjö 07master:2b982e92f42a: rtpdec: Set need_parsing via a handler field
[22:41] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:4c8a556a773c: Merge commit 'b7a4c319fda22aa91ce29692d728ec6103b514f6'
[22:41] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:11fb6258582d: Merge commit '2b982e92f42a6e661d90b12b6592cd13cae496d4'
[22:55] <cone-631> ffmpeg.git 03Martin Storsjö 07master:e72605f80bf5: rtpdec: Allow allocating and freeing the private data without explicit functions
[22:55] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:48e3cd4fcd49: Merge commit 'e72605f80bf5cbe32053a554ccc137e0a99cf3dd'
[23:02] <cone-631> ffmpeg.git 03Martin Storsjö 07master:5d8cae45737b: rtpdec: Get rid of all trivial .alloc/.free functions
[23:02] <cone-631> ffmpeg.git 03Martin Storsjö 07master:5956f489d045: rtpdec_hevc: Add asterisks at the start of each long comment line
[23:02] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:3804d73e4651: Merge commit '5d8cae45737bed6239bd6b6e0698802dbe1463c8'
[23:02] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:df38883607d2: Merge commit '5956f489d0452ff6dea6b6b81b4fa8e596fc5684'
[23:10] <cone-631> ffmpeg.git 03Martin Storsjö 07master:8633fb47db2e: rtpdec_hevc: Share the implementation of parsing a=framesize with h264
[23:10] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:29f9d82d23ff: Merge commit '8633fb47db2ec39eb8bd1bd65302af75a94ff5d0'
[23:23] <cone-631> ffmpeg.git 03Martin Storsjö 07master:f3449062a8d1: rtpdec_hevc: Reduce indentation level by returning early on errors
[23:23] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:0f8de2b5f17b: Merge commit 'f3449062a8d100ac4f703647336c32b126aa99f1'
[23:30] <cone-631> ffmpeg.git 03Martin Storsjö 07master:3567b91e49c6: rtpdec_hevc: Share the implementation of fragmented packets with h264
[23:30] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:17773f49b7da: Merge commit '3567b91e49c6ae101c9a35c90f46b8ad9890ac15'
[23:38] <cone-631> ffmpeg.git 03Martin Storsjö 07master:c82bf15dca00: rtpenc: Merge the h264 and hevc packetizers
[23:38] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:cf95877474ad: Merge commit 'c82bf15dca00f67a701d126e47ea9075fc9459cb'
[23:44] <cone-631> ffmpeg.git 03Martin Storsjö 07master:12251f997bbc: rtpdec: Remove unnecessary checks
[23:44] <cone-631> ffmpeg.git 03Martin Storsjö 07master:ac0e54fda930: rtpdec: Add const to string parameters in internal fmtp parsing functions
[23:44] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:1c5d9dd2b739: Merge commit '12251f997bbc0abb93be39c51021e6d404ca385f'
[23:44] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:499da717b8f3: Merge commit 'ac0e54fda9305cc7d149007f5b512bb8619f7c78'
[23:52] <cone-631> ffmpeg.git 03Martin Storsjö 07master:ec96a89c3e50: rtpdec: Don't pass non-const pointers to fmtp attribute parsing functions
[23:52] <cone-631> ffmpeg.git 03Martin Storsjö 07master:88434f9725e7: rtpdec: Remove unnecessary inline attributes
[23:52] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:d2a5c6f2104d: Merge commit 'ec96a89c3e507cf0fb1f2b159b28a53f2bad9a74'
[23:53] <cone-631> ffmpeg.git 03Michael Niedermayer 07master:99c76902cdc3: Merge commit '88434f9725e7c9484dcbcf323566ae88a2904f32'
[00:00] --- Wed Feb 25 2015

More information about the Ffmpeg-devel-irc mailing list