[Ffmpeg-devel-irc] ffmpeg-devel.log.20140307

burek burek021 at gmail.com
Sat Mar 8 02:05:02 CET 2014


[00:03] <cone-884> ffmpeg.git 03Luca Barbato 07master:5b977c1d4fd9: mxf: Return meaningful errors
[00:04] <cone-884> ffmpeg.git 03Michael Niedermayer 07master:39f5003a07db: Merge commit '5b977c1d4fd99f325049d48fb8f0a3240d50eef8'
[00:15] <cone-884> ffmpeg.git 03Rainer Hochecker 07master:b66382101cff: dxva2: Increase maximum number of slices for mpeg2
[00:15] <cone-884> ffmpeg.git 03Michael Niedermayer 07master:b7c1f6f9cc75: Merge commit 'b66382101cff33e2ce66500327a90d0a105eedeb'
[00:33] <cone-884> ffmpeg.git 03Derek Buitenhuis 07master:d67dbfa7c195: libx265: Support API version 9
[00:33] <cone-884> ffmpeg.git 03Michael Niedermayer 07master:0707a34dd817: Merge remote-tracking branch 'qatar/master'
[00:46] <cone-884> ffmpeg.git 03Hendrik Leppkes 07master:276839b8de7f: dxva2_h264: set the correct ref frame index in the long slice struct
[00:46] <cone-884> ffmpeg.git 03Hendrik Leppkes 07master:8b2a130d3fe6: dxva2_h264: add a workaround for old intel GPUs
[00:49] <michaelni> nevcairiel, ok, thanks, applied
[00:54] <cone-884> ffmpeg.git 03Michael Niedermayer 07master:64e4bd7414fc: doc/APIchanges: update versions & hashes
[11:02] <nevcairiel> this locked trac database thing is stupid
[11:05] <ubitux> and annoying
[11:06] <nevcairiel> if its really a backup, someone should move it into EU sleepy times
[11:07] <nevcairiel> and make it faster!
[11:12] <wm4> obviously the server should just run a script that deletes the lock every other minute
[11:13] <wm4> or just restart the server altogether to be sure
[11:13] <nevcairiel> when a backup process locks the database, such a lock is somewhat meaningful
[11:13] <nevcairiel> as long as it unlocks again once its done
[11:13] <nevcairiel> and doesnt take half an hour to backup o.o
[11:13] <wm4> oh I see, so the backup process is locking it
[11:14] <nevcairiel> thats what someone claimed is happening
[11:37] <nevcairiel> i'm starting to wonder if the theory isnt wrong
[11:37] <nevcairiel> its still happening
[12:08] <michaelni> nevcairiel, ubitux is that db locked thing still happening ? which url shows it ?
[12:15] <anshul> michaelni,no lock is not now, it gives lock when i try to login
[12:18] <ubitux> michaelni: you just need to try to login to trigger the lock when that happens
[12:19] <nevcairiel> michaelni: seems its gone again
[12:19] <nevcairiel> i just tried to open a issue link as spammed by the bot
[12:20] <nevcairiel> (but i am logged in, so that may trigger it)
[12:53] <cone-388> ffmpeg.git 03Patrice Clement 07master:93d216d37a3f: dict: K&R formatting cosmetics
[12:53] <cone-388> ffmpeg.git 03Michael Niedermayer 07master:774d0c9dbaf5: Merge commit '93d216d37a3f95190ecb9d51cf72f54ea4e04ec7'
[12:58] <cone-388> ffmpeg.git 03Martin Panter 07master:5b2ad78f97d4: rtmppkt: Handle extended timestamp field even for one-byte header
[12:58] <cone-388> ffmpeg.git 03Michael Niedermayer 07master:d540e7577a4c: Merge commit '5b2ad78f97d43299adcb038c04346999fe9b196c'
[13:10] <cone-388> ffmpeg.git 03Martin Panter 07master:48cfb5f41140: rtmppkt: Add a descriptive comment
[13:10] <cone-388> ffmpeg.git 03Michael Niedermayer 07master:94ec27131c7a: Merge commit '48cfb5f4114048383a4ee85b2f500eefce24cad2'
[13:49] <cone-388> ffmpeg.git 03Martin Panter 07master:8bbb02ae4f44: rtmppkt: Properly handle extended timestamps when writing
[13:49] <cone-388> ffmpeg.git 03Michael Niedermayer 07master:9baef60ae400: Merge commit '8bbb02ae4f44f1cddc1ce1e74a71fa2022e93da3'
[14:14] <cone-388> ffmpeg.git 03Martin Storsjö 07master:77eed91fab9b: rtmppkt: Rename the ts_delta field to ts_field
[14:14] <cone-388> ffmpeg.git 03Michael Niedermayer 07master:4fc339faea11: Merge remote-tracking branch 'qatar/master'
[14:30] <plepere> BBB : if punpcklbw/punpckhbw is slower than movhlps+pmovzxbw, should I do 2 different versions of the functions ?
[14:30] <plepere> one for SSSE3, one for SSE4
[15:01] <plepere> actually, the punpck is faster. :P
[15:16] <Skyler_> plepere: in my experience pmov* is basically useless except for sign-extension
[15:16] <plepere> ok
[15:17] <Skyler_> you'd think it would be faster in some case, but often it's reliably slower
[15:18] <Skyler_> (I really only have wild guesses as to why)
[15:24] <ubitux> i've written 3 lines of arm asm
[15:24] <ubitux> and already triggered an objdump bug
[15:24] <ubitux> awesome.
[15:24] <ubitux> (and same in gdb ofc)
[15:24] <ubitux> "all versions of objdump that I have seem to disassemble 'vshll.u16' instruction incorrectly"
[15:24] <ubitux> great..
[15:25] <Skyler_> what do they disassemble it as?
[15:26] <ubitux> stuff like "vshll.u16 q6, d26, #6" becomes "vshll.u16 d12, q13, #6"
[15:26] <Skyler_> oh, geez, so they just get the registers wrong completely
[15:26] <ubitux> yeah they swapped wrongly
[15:27] <Skyler_> though isn't d26 == q13 and q6 == d12?
[15:27] <ubitux> d12/d13 are q6 and d26/d27 are q13
[15:27] <Skyler_> oh, I see, you mean it displays wrong because it should show {Q, D} not {D, Q}
[15:27] <ubitux> well it displays wrongly but the "wrong" version makes no sense and isn't assemblable
[15:28] <ubitux> q6 is d12+d13, and d26+d27 are q13
[15:32] <ubitux> -            val = get_xbits(gb, nbits);
[15:32] <ubitux> +            val                                 = get_xbits(gb, nbits);
[15:32] <ubitux> lol
[15:32] <ubitux> so much better
[15:39] <plepere> it's all in the indentation
[15:50] <plepere> isn't it possible to  assign a string in ASM preprocessor ?
[15:50] <plepere> %assign %%lod movd
[15:50] <plepere> for instance
[15:55] <Skyler_> %define is what you use for that, not %assign, I think
[15:55] <Skyler_> %assign is for variables
[15:57] <plepere> ok
[15:57] <plepere> thanks
[15:58] <plepere> works !
[15:58] <plepere> :D
[16:08] <cone-388> ffmpeg.git 03Hendrik Leppkes 07master:930f67b71294: dxva2_vc1: fix intensity compensation condition with interlaced fields
[16:08] <cone-388> ffmpeg.git 03Hendrik Leppkes 07master:61ff0431ba8e: dxva2_vc1: include the start code in wMBbitOffset
[17:47] <ubitux> BBB: i'll probably start doing the vp9 mc for arm, since i don't know how they work
[17:47] <ubitux> any suggestion or whatever information about those before i start?
[19:58] <troulouliou_dev> http://bitbin.it/Ky30I5LP
[19:58] <troulouliou_dev> a small bug correction patch for the demux example
[20:00] <llogan> troulouliou_dev: patches should be sent to ffmpeg-devel at ffmpeg.org...otherwise we tend to forget
[20:00] <troulouliou_dev> llogan, ok
[20:01] <llogan> troulouliou_dev: and please use git format-patch, or git send-email if you can
[20:47] <wm4> wow requiring ML interaction for such trivial patches
[20:55] <Compn> i could have sent it to the list or committed it
[20:55] <Compn> i just wasnt here at the moment wm4 :P
[20:57] <Compn> i guess i can commit it now
[20:57] <Compn> good time to install git i think
[20:58] <Compn> see how long this takes
[21:11] <Compn> ok cloning the websites
[21:11] <Compn> now how to developer access the actual repo
[21:14] <Compn> anyone remember pls pm me :)
[21:14] <llogan> websites?
[21:14] <Compn> ffmpeg-web and rtmpdump-web
[21:14] <llogan> for that trivial patch?
[21:14] <Compn> no
[21:14] <Compn> just because i had to reinstall git
[22:40] <Compn> michaelni : you there to help me with dumb git question ?
[22:43] <michaelni> Compn, maybe ;)
[22:44] <Compn> michaelni : do you remember how to set git up so ican commit to ffmpeg ?
[22:44] <Compn> i have a fresh install and forgot the steps
[22:46] <michaelni> maybe http://www.ffmpeg.org/git-howto.html
[22:46] <michaelni> and if not then it should be fixed
[22:46] <Compn> is the developer repo git url the same as the read git repo ?
[22:46] <Compn> :)
[22:47] <michaelni> push urls differ from anon read urls
[22:47] <michaelni> normally
[22:47] <Compn> ok , wheres the push url ?
[22:49] <michaelni> it should be in the git howto and maybe download page but its possibly not
[22:49] <michaelni> for web its gil at ffmpeg.org:ffmpeg-web
[22:49] <michaelni> feel free to add it
[22:50] <michaelni> if its missig
[22:50] <Compn> michaelni : i have forgotten the url, whats in your .git/config ?
[22:51] <michaelni> for web:  url = gil at ffmpeg.org:ffmpeg-web
[22:53] <Compn> not for web
[22:53] <Compn> i know web
[22:53] <Compn> i mean ffmpeg itself
[22:56] <michaelni> githowto says: git clone git at source.ffmpeg.org:ffmpeg
[23:03] <Compn> ehe
[23:06] <Compn> heh
[23:06] <Compn> git clone git://source.ffmpeg.org/ffmpeg is what the git-howto says
[23:10] <Compn> oh theres two
[23:10] <Compn> humm
[23:12] <wm4> michaelni: are you sure you want him to push _anything_ to the git repo
[23:14] <michaelni> Compn, what do you want to push to the repo ?
[23:14] <Compn> michaelni : just that patch fixing typo in the doc
[23:16] <michaelni> you can also send me a pull req or give me a url of that git patch.
[23:16] <michaelni> if you push directly make sure you dont make a mistake
[23:17] <michaelni> always do a "git push origin master --dry-run"  before actual push
[23:17] <Compn> its on ffmpeg-devel already :P
[23:17] <Compn> the patch
[23:18] <beastd> doesn't mean you can't fuck up the push
[23:18] <nevcairiel> in some crazy oddball patch format
[23:18] <Compn> i know
[23:18] <beastd> look at the dryrun output closely
[23:18] <Compn> ive committed to ffmpeg before, and followed dry run instructions beastd :)
[23:18] <Compn> just cant remember how i setup last time
[23:20] <beastd> just put the url there michaelni mentioned, i cannot think of problems besides ssh key auth
[23:20] <Compn> i'm guessing its ssh key auth problem
[23:22] <cone-388> ffmpeg.git 03compn 07master:84bccae8e9ac: doc: fix typo
[23:22] <Compn> pfft
[23:22] <Compn> i guess i had to restart my git program for it to work
[23:22] <Compn> thanks guys :)
[23:24] <wm4> ubitux: I should have read that post more closely, I thought he wanted to _use_ libavfilter
[23:25] <Compn> wm4 : you doubted my skills? :P
[23:25] <beastd> Compn: Do you have the priv key for the corresponding pub key you gave for ffmpeg.git repo on that machine at a location where ssh will find it?
[23:25] <Compn> beastd : ^^ i just committed, thanks for helping. :)
[23:25] <Compn> and no , i dont know what was wrong.
[23:26] <beastd> Compn: Oh ok.  That commit came so fast that I doubted it was you who pushed it ^^;
[23:26] <Compn> or how i fixed it
[23:26] <beastd> That always leaves an uneasy feeling IMHO
[23:26] <Compn> 	url = git at git.videolan.org:ffmpeg.git
[23:27] <Compn> is whats in my .git/config
[23:27] <beastd> url = git at source.ffmpeg.org:ffmpeg # that is what i have, and i never had any problems with push
[23:37] <Compn> i dont even want to know what kind of voodoo is going on with the two git urls :)
[23:53] <Compn> michaelni : i guess i dont have any wisdom to add to the git-howto , only 'if its not working, try turning it off and on again' :)
[00:00] --- Sat Mar  8 2014


More information about the Ffmpeg-devel-irc mailing list