[Ffmpeg-devel-irc] ffmpeg-devel.log.20150624

[burek]

[00:34:08 CEST] <cone-309> ffmpeg 03Mariusz SzczepaDczyk 07master:71034163acbb: lavf/libsmbclient: implement move and delete callbacks
[00:51:41 CEST] <cone-309> ffmpeg 03Lukasz Marek 07master:e96f0a692b3c: lavf/libssh: implement directory listing callbacks
[00:51:42 CEST] <cone-309> ffmpeg 03Mariusz SzczepaDczyk 07master:0cce94fb1037: lavf/libssh: read empty path from url as /
[00:51:43 CEST] <cone-309> ffmpeg 03Mariusz SzczepaDczyk 07master:b23d2bac0d70: lavf/libssh: implement move and delete callbacks
[03:25:27 CEST] <cone-309> ffmpeg 03Michael Niedermayer 07master:89a420b71b53: avcodec/mpegaudio_parser: Discard ID3v1 tag at the end
[03:25:28 CEST] <cone-309> ffmpeg 03Michael Niedermayer 07master:041aa800f220: avformat/mp3dec: Remove the ID3v1 tag removial code
[05:05:57 CEST] <jamrial> nevcairiel: since you were wondering, gcc 4.9.3 is being released later this week
[05:19:16 CEST] <metalwood> good
[05:20:49 CEST] <pengvado> durandal_1707: yes
[07:37:47 CEST] <nevcairiel> philipl: i dont know anything about vdpau, but i can read it and see if i spot anything obvious
[07:37:57 CEST] <nevcairiel> jamrial: i saw, ya
[08:10:59 CEST] <philipl> nevcairiel: thats all i need. thanks.
[09:59:22 CEST] <j-b> 'matin
[10:01:53 CEST] <durandal_170> what's up
[10:33:24 CEST] <rcombs> https://gist.github.com/c4a7a17b7c05a07a216c <-- is this the right place for this check?
[10:35:00 CEST] <rcombs> fixes a crash when remuxing this with an unnecessary `-bsf:v h264_mp4toannexb`: https://dl.dropboxusercontent.com/u/105828/sample%20crash%20file.mkv
[10:36:47 CEST] <rcombs> but I could see an argument that it should go either in ff_avc_parse_nal_units_buf or higher up the stack in matroskaenc
[10:37:11 CEST] <nevcairiel> how does it end up with a empty packet
[10:40:07 CEST] <rcombs> nevcairiel: good question
[10:40:32 CEST] <rcombs> I get some whining on stderr: "Failed to open bitstream filter h264_mp4toannexb for stream 0 with codec copy: Invalid argument"
[10:40:47 CEST] <rcombs> usually that's pretty harmless
[10:41:09 CEST] <rcombs> but in this case there's actually an invalid NAL in there somewhere, at least according to the h264 decoder
[10:46:05 CEST] <wm4> empty packet = flush packet?
[10:46:52 CEST] <rcombs> hmm
[10:47:23 CEST] <rcombs> isn't that usually a NULL packet, rather than a packet with NULL data?
[10:47:41 CEST] <wm4> at leats not on the API level
[11:01:33 CEST] <__gb__> BtbN, FYI, my download of the MSDK community edition through wget completed. The enclosed site_license_materials.txt file mentions: No components have site license restrictions -- so it looks to me that only the process needs to be improved, by e.g. just publishing the download link as is?
[11:01:38 CEST] <__gb__> didn't have the time to try yet
[11:02:07 CEST] <__gb__> the tarball contains a tarball that contains a tarball that contains packages for CentOS, SLES and a tarball for "generic" versions :)
[11:02:24 CEST] Action: __gb__ hides
[11:03:05 CEST] <BtbN> interesting
[11:03:15 CEST] <BtbN> sounds like putting the direct link in an ebuild is fine then
[12:10:59 CEST] <cone-218> ffmpeg 03Janne Grunau 07master:c2de2cf0d292: arm64: constify src in h264qpel dsp function definitions
[12:11:00 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:8f454c315454: Merge commit 'c2de2cf0d2927f3f584dab6d54276fbda92a0a71'
[12:39:24 CEST] <cone-218> ffmpeg 03Mark Webster 07master:a88e21f5b78e: Fix missing symbols when libvpx has only vp8 enabled
[12:39:25 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:7e63a622d14a: Merge commit 'a88e21f5b78ef5e75b073ecbfd4e92921002a703'
[12:58:35 CEST] <cone-218> ffmpeg 03Janne Grunau 07master:41740ef8be6e: libvpx: fix test for VPX_IMAGE_ABI_VERSION
[12:58:36 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:49038fc17dd0: Merge commit '41740ef8be6ec409f7eff3932ddba9a9eeec27b1'
[13:57:02 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:8f0f678f090d: ffmpeg: Do not use the data/size of a bitstream filter after failure
[14:58:34 CEST] <BBB> durandal_170: btw, re: your evrc patch, some codecs use avctx->skip_loop_filter for that
[14:58:44 CEST] <BBB> (I know its not exactly a loopfilter, but the idea is the same)
[15:06:22 CEST] <durandal_170> it is used for video only
[15:21:57 CEST] <BBB> we could use it for audio; many of the speech codecs have postfilters that we could turn off with this flag if we wanted to
[15:22:20 CEST] <BBB> (not that I have any strong opinion, its rally only useful for testing afaik)
[15:22:25 CEST] <BBB> (so feel free to ignore me)
[15:40:47 CEST] <cone-218> ffmpeg 03Mariusz SzczepaDczyk 07master:346624be4aff: lavf: set is_connected flag so url can be properly closed
[16:08:57 CEST] <cone-218> ffmpeg 03Paul B Mahol 07master:5a1f78524959: avfilter: add ssim filter
[16:31:16 CEST] <ubitux> the way to entangle the init/retrieve/uninit hwaccel thing is weird
[16:32:27 CEST] <ubitux> like, the uninit cb being set in the init
[16:51:20 CEST] <ubitux> and mmh it seems you need to decode at least once to know what kind of pixel format you'll get in output?
[17:00:08 CEST] Action: ubitux still hasn't figured out the purpose of get_format
[17:27:50 CEST] <durandal_170> I was thinking abou writing drawgraph filter which would display various graphical representations of metadata numbers user pick up
[17:28:52 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:8ff09e909869: avcodec/jpeg2000dec: Use 32x32->64bit for 5/3 dequantization
[17:28:53 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:05ef6324d110: avcodec/jpeg2000dec: Add 5/3 de-quantization special case for the lossless case
[17:28:54 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:ffa1de8a3b93: avcodec/jpeg2000dec: Use 32x32->64bit for 9/7i dequantization
[17:32:46 CEST] <rcombs> OK, I'm screwing around in bsf_h264_mp4toannexb.c, trying to see what needs doing to do the equivalent transform for HEVC
[17:34:06 CEST] <rcombs> I've got h264_extradata_to_annexb tweaked to parse HEVC extradata (it's not _hugely_ different), but once I actually hit the NAL parsing in h264_mp4toannexb_filter things get very strange indeed
[17:35:04 CEST] <rcombs> the data passed in doesn't match what I see in mkvinfo, and it looks as if something else got halfway through the annex B conversion already
[17:35:14 CEST] <rcombs> so I'm certainly missing several things
[17:37:54 CEST] <nevcairiel> dunno, the hevc mp4 extradata is quite a bit more complex than the h264 version
[17:39:51 CEST] <rcombs> if you're just looking to parse out SPS and PPSs (which is what I was going for, just to see if I could roughly shoehorn this into the existing AVC code), it's not hard
[17:40:17 CEST] <rcombs> https://gist.github.com/f0411529f620d81508b4 <-- the results of my tomfoolery
[17:40:51 CEST] <rcombs> (diff -w'd for sanity)
[17:41:10 CEST] <rcombs> https://gist.github.com/318dbc4a324a13473a79 <-- sans--w
[17:54:12 CEST] <rcombs> https://gist.github.com/rcombs/6c82e2fbe6a98441aaa7 :/
[18:14:28 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:d58025557155: vcodec/jpeg2000: Increase cblk size limit to 128
[18:14:29 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:d7e224ec24f7: avcodec/jpeg2000: change flags array type to 16bit
[19:33:31 CEST] <cone-218> ffmpeg 03James Almer 07master:58d7dde081d1: brstm: fix use of uninitialized variables
[20:02:15 CEST] <llogan> durandal_170: what does drawgraph do? (too lazy to apply patch & compile)
[20:03:31 CEST] <durandal_170> draw graph from metadata numbers
[20:04:45 CEST] <durandal_170> to visualize ssim calculations, for example
[20:05:51 CEST] <llogan> i see. thanks
[20:48:12 CEST] <cone-218> ffmpeg 03Paul B Mahol 07master:5a79bf0284a3: PCM signed 16-bit big-endian planar decoder
[20:48:13 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:8e2fcd1cb4be: Merge commit '5a79bf0284a37855636390e620d189a964214c9e'
[20:48:14 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:ffd9d60201a7: avcodec: Add codec_desc and remapping for AV_CODEC_ID_PCM_S16BE_PLANAR_DEPRECATED
[21:07:00 CEST] <cone-218> ffmpeg 03Paul B Mahol 07master:ded5957d75de: segafilm: Remove deplanarization hack
[21:07:01 CEST] <cone-218> ffmpeg 03Luca Barbato 07master:4d5548439111: dds: Fix palette mode on big endian
[21:07:02 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:41070923490a: Merge commit 'ded5957d75def70d2f1fc1c1eae079230004974b'
[21:07:03 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:87544ce8dcf9: Merge commit '4d55484391112a30d16812597b9ad81605af0c24'
[21:21:31 CEST] <cone-218> ffmpeg 03Tom Butterworth 07master:5dbd491eb38e: TextureDSP: fix erroneous condition which produced blocky output for DXT
[21:21:32 CEST] <cone-218> ffmpeg 03Michael Niedermayer 07master:eb823df81766: Merge commit '5dbd491eb38efab1d1313d4129ed76ab2e98176d'
[00:00:00 CEST] --- Thu Jun 25 2015