[Ffmpeg-devel-irc] ffmpeg-devel.log.20150107

burek burek021 at gmail.com
Thu Jan 8 02:05:02 CET 2015


[00:03] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:15ea222778ca: vf_interlace: merge FIELD_LOWER check
[00:03] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:82b1bcd7f9b6: Merge commit '15ea222778caaec0877b3f9938140b707c931d96'
[00:13] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:4438c256387a: vf_interlace: remove the warning tag in the message
[00:13] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:ed97963bdbf3: ulti: invert the order of parameters of ulti_decode_frame()
[00:13] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:f2f88b952d66: Merge commit '4438c256387a5845e5cdc5c14c43f4047b68301b'
[00:13] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:40e123bc67cc: Merge commit 'ed97963bdbf3bb17fca4f9ea0aa1a97722dec907'
[00:33] <cone-150> ffmpeg.git 03Vittorio Giovara 07master:1c6183233d56: msmpeg4: check memory allocations and propagate errors
[00:33] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:89388cf60bd5: Merge commit '1c6183233d56fb27a4a154e7e64ecab98bd877f1'
[00:39] <cone-150> ffmpeg.git 03Justin Ruggles 07master:8e03ca2fff40: isom: add 'mp1v' fourcc
[00:39] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:4009ee3eccac: Merge commit '8e03ca2fff40ce065ea5a516200e777c96cd785e'
[00:46] <cone-150> ffmpeg.git 03Martin Storsjö 07master:e581e88cbdb3: movenc: Readd an accidentally removed condition
[00:47] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:0d0a905152f6: Merge commit 'e581e88cbdb3323ca0026a54b39a9716a3686e9f'
[01:24] <cone-150> ffmpeg.git 03Martin Storsjö 07master:2a1500fb5523: movenc: Fix a typo in a comment
[01:24] <cone-150> ffmpeg.git 03Martin Storsjö 07master:8a3d9ca603f4: libavcodec: Add an OpenH264 encoder wrapper
[01:24] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:cb03d29dec64: Merge commit '2a1500fb552364488cfbee3e29e296788a3a6bdd'
[01:24] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:c09eeca5468c: Merge commit '8a3d9ca603f4d15ecaa9ca379cbaab4ecaec8ce4'
[01:24] <cone-150> ffmpeg.git 03Michael Niedermayer 07master:ddae03f69bc1: avcodec/allcodecs: prefer x264 over openh264 if both are available
[04:01] <jamrial> german spambot?
[04:05] <Timothy_Gu> lol
[04:07] <Timothy_Gu> lol x265 is classified by GitHub as being an assembly project
[04:21] <jamrial> Isn't x265 a mercurial project hosted on bitbucket? what are you looking at?
[04:36] <Timothy_Gu> jamrial: https://github.com/mirror/x265
[04:43] <jamrial> ah, interesting
[04:45] <jamrial> but yeah, x265 doesn't like to use yasm macros so their .asm files are huge, apparently making up 56% of total lines of code
[06:26] <ac_slater_> hey guys. I'm making a series of patches to fix the stagefright android example. Issue is, when I do a clean build against bionic (libc) via `./configure` ... apparently log2() isnt available. If I disable most codecs then it works.
[06:26] <ac_slater_> any hints?
[06:26] <ac_slater_> (even though log2 says it's available in config.log)
[06:40] <ac_slater_> also, ffmpeg 2.5 
[06:54] <Timothy_Gu> ac_slater_: you should find out 1. if log2 is supposed to work, and 2.why log2 doesn't work, and then fix it in configure
[06:55] <Timothy_Gu> ac_slater_: seems like it is supposed to work by reading bionic source
[06:56] <ac_slater_> Timothy_Gu: exactly ... 
[06:56] <Timothy_Gu> ac_slater_: what's the error?
[06:56] <ac_slater_> Timothy_Gu: http://paste.debian.net/plain/139519
[06:58] <Timothy_Gu> ac_slater_: so apparently log2 is not defined in a header
[06:58] <ac_slater_> Timothy_Gu: and I dont really believe that
[07:00] <Timothy_Gu> ac_slater_: seems like it is
[07:00] <Timothy_Gu> ac_slater_: can you run `make V=1` and paste config.h and config.log?
[07:00] <Timothy_Gu> the decl is guarded with __BSD_VISIBLE
[07:01] <ac_slater_> oooh intersting
[07:01] <ac_slater_> working on those pastes
[07:03] <Timothy_Gu> the exact line is "#if __BSD_VISIBLE || __ISO_C_VISIBLE >= 1999 || __XSI_VISIBLE"
[07:04] <ac_slater_> Timothy_Gu: stupid question, should I just force one?
[07:05] <Timothy_Gu> ac_slater_: you can for this build, but it would: a. be stupid and b. not benefit other users
[07:06] <ac_slater_> agreed
[07:10] <Timothy_Gu> ac_slater_: any progress?
[07:11] <ac_slater_> Timothy_Gu: yea interesting. config.log is showing the same thing... log2 and log2f implicitly defined
[07:11] <ac_slater_> (I have to scrub my paths since my work policy says I have to :( )
[07:12] <Timothy_Gu> ac_slater_: sed -i 's,/my/path,/censored,g' config.log
[07:12] <Timothy_Gu> ac_slater_: you work at Raytheon?
[07:16] <ac_slater_> Timothy_Gu: a subsidiary, but yes ;)
[07:19] <ac_slater_> alright, almost there... apparently some sites dont like pasting 400+kb
[07:21] <Timothy_Gu> curl -F 'sprunge=<file' http://sprunge.us | should work
[07:21] <ac_slater_> Timothy_Gu: http://sprunge.us/cRbQ
[07:21] <Timothy_Gu> haha
[07:21] <ac_slater_> haha yea, I couldnt remember the name of it!
[07:22] <ac_slater_> so yes, this isnt as simple as `./configure -os=android` ... but I hope it's close enough to debug
[07:24] <Timothy_Gu> ac_slater_: yeah I think it's fine but not everybody here thinks so, so don't be surprised when you meet a person not as nice as me ;)
[07:25] <ac_slater_> I've probably met them all haha. I'm usually here asking ridiculous things
[07:26] <Timothy_Gu> ac_slater_: stupid question: what android sdk are you using
[07:26] <Timothy_Gu> oh 4.2.2 saw that
[07:26] <ac_slater_> well, 4.2.2 is directory I made ... I'm using NDK r10
[07:27] <Timothy_Gu> nvrmd
[07:30] <ac_slater_> If it'd help, I can the minimal config
[07:31] <Timothy_Gu> that's dumb, bionic doesn't map _ISOC99_SOURCE to it's very own __ISO_C_VISIBLE or _C99_SOURCE
[07:31] <ac_slater_> interesting
[07:32] <Timothy_Gu> *its
[07:33] <Timothy_Gu> ok try this
[07:33] <Timothy_Gu> Add
[07:33] <Timothy_Gu> add_${pfx}cppflags -D_POSIX_C_SOURCE=200112 -D_XOPEN_SOURCE=600
[07:33] <Timothy_Gu> to after
[07:33] <Timothy_Gu> eval ${pfx}libc_type=bionic
[07:35] <Timothy_Gu> _POSIX_C_SOURCE implies __ISO_C_VISIBLE == 1999 so it *should* work
[07:36] <ac_slater_> cool, testing now
[07:37] <anshul_mahe> ubitux: for add_insert_timeline_graph is it possible to add it in parser instead of cmdutil
[07:42] <ubitux> what do you mean?
[07:42] <ubitux> what parser?
[07:43] <ac_slater_> Timothy_Gu: :(
[07:44] <ac_slater_> no go
[07:46] <ac_slater_> and I could see the cppflags added
[07:46] <ac_slater_> can * 
[07:46] <Timothy_Gu> ac_slater_: can you paste the preprocessor output?
[07:46] <Timothy_Gu> of a file that contains #include <math.h>
[07:47] <aetasx> you guys know if theres a way to pull gcc's default optimization param values? (the ones passed in with --param)
[07:50] <Timothy_Gu> aetasx: man gcc | grep default ?
[07:50] <ac_slater_> Timothy_Gu: (899 lines)
[07:51] <Timothy_Gu> aetasx: do you mean like `gcc -fthis-is-an-op`?
[07:51] <Timothy_Gu> aetasx: 768 here ;)
[07:51] <aetasx> nah theres some other parameters get passed in and set that change how those options work.  lemme grab an example
[07:52] <Timothy_Gu> grepping for "Options That Control Optimization" in man page might help
[07:53] <aetasx> so like "gcc -fgcse --param max-gcse-memory=1024..."
[07:53] <ac_slater_> Timothy_Gu: http://sprunge.us/GgDi
[07:54] <aetasx> yeah the manual lists some but the manual also lists other optimization options I don't have so its obviously different versions and most of the actual params don't state their default value in the manual anyway
[07:54] <ac_slater_> Timothy_Gu: shit... doesnt have the #define statements
[07:54] <anshul__> ubitux: in libavcodec in file allcodec.h there is provision to register any parser, to make any file and then register there
[07:54] <anshul__> exact file path is libavcodec/allcodecs.c
[07:54] <aetasx> https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#Optimize-Options  if you just hit END and skip to the bottom of the page for instance
[07:55] <Timothy_Gu> ac_slater_: try -D_POSIX_C_SOURCE=200112
[07:55] <Timothy_Gu> ac_slater_: that will be more realisitic like FFmpeg
[07:55] <ubitux> anshul__: no it doesn't belong here at all
[07:56] <ac_slater_> Timothy_Gu: yea no difference 
[07:56] <Timothy_Gu> aetasx: I don't see any undocumented params
[07:56] <Timothy_Gu> aetasx: or you can use man page which should be corresponding to your local gcc
[07:56] <Timothy_Gu> ac_slater_: hm
[07:57] <ac_slater_> Timothy_Gu: this is really odd
[07:57] <aetasx> no, those are documented but I don't know which ones are available in my gcc version since there's no listing for them and I was trying to find out what their default "sane" value is set at
[07:57] <aetasx> before I go and set it to something ludicrous
[07:58] <anshul_m> is that libavcodec parser only for some specific purpose?
[07:58] <Timothy_Gu> ac_slater_: do you have permission to modify ndk headers?
[07:59] <Timothy_Gu> aetasx: yes that's why I told you to read your *local* man page
[08:00] <aetasx> Timothy_Gu: thanks but it looks like I found the flag.  Looks like you can list them through "gcc --help=params"
[08:00] <ac_slater_> Timothy_Gu: yea. Im just curious why it's not there
[08:03] <anshul_m> ubitux: is that libavcodec parser only for some specific purpose parser?
[08:04] <ac_slater_> Timothy_Gu: I found some things on stackoverflow that point to this being intentional
[08:04] <Timothy_Gu> ac_slater_: can you compile with gcc -E -dM /tmp/c.c
[08:04] <Timothy_Gu> ac_slater_: that's odd
[08:05] <Timothy_Gu> ac_slater_: this cmd will show the predefined variables
[08:05] <ubitux> anshul_m: yes
[08:05] <anshul_m> they could only used by libavcodec or something like that, just for information I am asking
[08:06] <ubitux> this code is parsing format data, and constructing a lavfi filtergraph, it doesn't belong in a codec parser
[08:06] <ac_slater_> Timothy_Gu: http://sprunge.us/PNjR
[08:07] <anshul_m> ubitux: is it possible to have same parser register feature in libavformat
[08:07] <ubitux> what for?
[08:10] <Timothy_Gu> ac_slater_: apparently -D_POSIX_C_SOURCE didn't work because a) it's not shown in the output and b) __POSIX_VISIBLE is 200809 which is the default
[08:10] <ac_slater_> Timothy_Gu: would it be weird to disable log2 detection and rely on libavutil/libm.h to fix it?
[08:10] <ac_slater_> Timothy_Gu: I see, that makes sense
[08:10] <Timothy_Gu> ac_slater_: why would you not use a (potentially optimized) log2() and depend on a fallback?
[08:11] <anshul_m> ubitux: we have parser which works in avformat, even in my scte case my parser need to be share by different muxer's
[08:11] <ac_slater_> Timothy_Gu: cause it's not there ;) at least not in a header
[08:11] <Timothy_Gu> ac_slater_: sorry to bother you again but this time try: gcc -E -D_POSIX_C_SOURCE=200112 -dD /tmp/c.c
[08:11] <Timothy_Gu> ac_slater_: it is there just somehow not enabled
[08:12] <ac_slater_> no bother
[08:12] <ubitux> anshul_m: well then just write shared code
[08:13] <anshul_m> ubitux: where should I register that file with shared code
[08:13] <ac_slater_> Timothy_Gu: ooooo
[08:13] <ac_slater_> Timothy_Gu: http://sprunge.us/EHPC
[08:13] <Timothy_Gu> aha
[08:13] <ubitux> anshul_m: "register"? you just add it wherever you want in lavf and reference it in the makefile?
[08:13] <ac_slater_> Timothy_Gu: so what does that actually mean?
[08:14] <Timothy_Gu> ac_slater_: nothing
[08:14] <Timothy_Gu> ac_slater_: now try: gcc -E -D_POSIX_C_SOURCE=200112 /tmp/c.c
[08:14] <Timothy_Gu> it just means that a log2() exists
[08:14] <Timothy_Gu> wait
[08:15] <ac_slater_> that works too... ie - log2 is exposed
[08:15] <Timothy_Gu> oh I think I gave the wrong instructions
[08:15] <Timothy_Gu> <@     Timothy_Gu> ac_slater_: try -D_POSIX_C_SOURCE=200112 <--- did you *really* follow this?
[08:15] <anshul_m> ubitux: In libavformat each reference in Makefile is related to some muxer_opt, is same file in different muxer_opt is possible
[08:16] <Timothy_Gu> ac_slater_: btw your c.c is *really* broken btw ;) casting pointer to int
[08:16] <ac_slater_> Timothy_Gu: it's taken from the ffmpeg configure script :p
[08:17] <ubitux> anshul_m: see OBJS? these are not (de)muxers
[08:17] <Timothy_Gu> ac_slater_: hehe
[08:18] <Timothy_Gu> ac_slater_: now try to compile an FFmpeg file
[08:18] <Timothy_Gu> with -D_POSIX_C_SOURCE=200112
[08:18] <ac_slater_> Timothy_Gu: hmm well. I upgraded to ndk r10 with android-19. Let me put it back
[08:18] <anshul_m> ubitux: thanks
[08:19] <Timothy_Gu> ac_slater_: the ndk version shouldn't matter
[08:19] <anshul_m> ubitux: is it also possible to share that code in libavfilter
[08:19] <Timothy_Gu> I'm looking at the source of android-4.0.1_r1
[08:19] <ac_slater_> Timothy_Gu: interesting
[08:19] <ac_slater_> Timothy_Gu: one sec
[08:19] <ubitux> anshul_m: that code could be in libavfilter, but there is no point, and it means exporting the symbol
[08:20] <ubitux> i don't want that, because it's app specific, and thus belongs in the tools sources
[08:21] <anshul_m> ubitux: I understand your patch now, from my side no prob where it is, for your patch
[08:21] <ac_slater_> Timothy_Gu: hmm well, ndk with the android-19 sysroot requires no posix std hacks... but my code breaks. I brought this issue up with android-17... so that's what I need to fix 
[08:23] <anshul_m> ubitux: I was aking it for my scte case, i was writing a filter that take 2 program and give one prog in output, which also use the same code shared by muxers
[08:23] <ac_slater_> Timothy_Gu: I think the last two pastes I gave you were from android-19... let me verify
[08:23] <anshul_m> ubitux: should I write same code again in that filter or export it 
[08:23] <anshul_m> export it from libavformat
[08:24] <ac_slater_> Timothy_Gu: yup... my mistake.... I miss led you. This `gcc -E` calls we from android-19 .... I redid them with android-17 and no log2 functions were exposed.
[08:25] <ac_slater_> misled * 
[08:25] <Timothy_Gu> darnit
[08:26] <Timothy_Gu> android 17 is 4.2* right?
[08:26] <ac_slater_> yea
[08:26] <ac_slater_> Timothy_Gu: yea it's ok. I was merely trying to fix libavcodec/libstagefright.cpp ... the buildscript in the ffmpeg tree is soooo broken
[08:26] <ac_slater_> I managed to get it to build and work with 4.3+
[08:26] <ac_slater_> but not 4.2
[08:27] <ac_slater_> but sadly to use libstagefright (and link it) ... you need some of the android source tree. So only about ~300mb of shit to get to build and link 1 file
[08:28] <Timothy_Gu> I made a mistake too
[08:28] <Timothy_Gu> I was looking at sys/cdefs.h from 4.0.1 but not math.h
[08:29] <ac_slater_> I wonder what issues will come up building my libs against the 4.3 sysroot and using them in 4.2 .... 
[08:29] <Timothy_Gu> yeah the 4.2 math.h is completely broken
[08:29] <Timothy_Gu> see https://android.googlesource.com/platform/bionic/+/a0ee07829a9ba7e99ef68e8c12551301cc797f0f%5E%21/libm/include/math.h
[08:31] <ac_slater_> lol +doublelog2(double);
[08:31] <ac_slater_> as in, I see the definition added
[08:31] <Timothy_Gu> and the commit is included in 4.3 or api level 18
[08:32] <Timothy_Gu> so yes, the best work around is using libavutil/libm.h
[08:32] <Timothy_Gu> after everything we've been through
[08:32] <Timothy_Gu> what a standard-compliant libc
[08:33] <ubitux> anshul_m: you can "private export" with avpriv_*
[08:34] <ac_slater_> Timothy_Gu: so would you force libavutil/libm.h or upgrade to level 18 if you could?
[08:34] <Timothy_Gu> ac_slater_: for you guys I don't know cuz I'm no android dev
[08:34] <ac_slater_> Timothy_Gu: me either ;)
[08:34] <Timothy_Gu> ac_slater_: but for FFmpeg we'll of course try to support both
[08:35] <ac_slater_> well now I know what the deal is. I'll make a note and submit a patch in a few days
[08:35] <ac_slater_> hmm but I guess it might be hard to detect bionic version
[08:35] <Timothy_Gu> ac_slater_: you don't really need to I guess
[08:35] <Timothy_Gu> remember the implicit decl error?
[08:35] <Timothy_Gu> it is originally a warning
[08:36] <ac_slater_> yea it still wouldnt continue though.. -Werror must have been on
[08:36] <Timothy_Gu> but then turned into an error with -Weror=implicit-function-declaration
[08:36] <Timothy_Gu> yes 
[08:36] <Timothy_Gu> but the check for log2() did not have the -Werror on
[08:37] <ac_slater_> I see. Interesting
[08:37] <Timothy_Gu> so we can either move the entire l. 5267 (master) block up
[08:38] <Timothy_Gu> or we can manually add -Werror=implicit-function-declaration -Werror=missing-prototypes to check_mathfunc (l. 4851)
[08:38] <Timothy_Gu> need to sleep now
[08:38] <Timothy_Gu> school tomorrow
[08:38] <ac_slater_> thanks for the help mate!
[08:38] <Timothy_Gu> ac_slater_: good luck
[08:38] <wbs> ac_slater_: fwiw, you shouldn't need to use the bionic libc headers at all, the normal ndk headers for that should work just fine
[08:38] <ac_slater_> take it easy
[08:38] <ac_slater_> Timothy_Gu: thanks!
[08:38] <Timothy_Gu> yeah wbs should know more about this stuff than I do
[08:38] <wbs> ac_slater_: just take the frameworks/av, system/core and hardware/libhardware headers from whatever android version you need
[08:38] <ac_slater_> wbs: I think that's what I point it to anyways
[08:39] <ac_slater_> oh *those*
[08:39] <anshul_m> ubitux: thanks I will look at that aproch
[08:39] <ubitux> anshul_m: that is, only if you need to share it between libs
[08:39] <ac_slater_> wbs: I was assuming the sysroot the ndk produced would be enough for stdlib stuff
[08:39] <ubitux> otherwise just use ff_ 
[08:40] <wbs> ac_slater_: yes, it should, and the math.h from the NDK should be identical for all android-3 to android-19
[08:41] <ac_slater_> yea that's what I was going. Timothy_Gu and I figured out that android-17 has some funky libm definitions
[08:42] <wbs> hmm, no, indeed, apparently android-18 in the NDK headers has got a change in math.h that may or may not be broken
[08:42] <Timothy_Gu> wbs: -18 changed it: https://android.googlesource.com/platform/development/+/master/ndk/platforms/android-18/include/math.h
[08:43] <Timothy_Gu> and -18 is the non-broken one
[08:43] <Timothy_Gu> now really gotta go
[08:44] <Timothy_Gu> it is identical for 13-17
[08:44] <Timothy_Gu> then 18 changed it
[08:44] <Timothy_Gu> then 21 changed it
[08:48] <ac_slater_> wbs: it's working with 18 and 19
[08:49] <ac_slater_> but all of this came from trying to find out what was causing the symbol "ff_isom_write_hvcc" for not being available for me with my reduced configure call. I'm sure that's a separate topic though (and probably not for this channel)
[08:49] <ac_slater_> s/reduced/reduced feature/g
[09:02] <wbs> ac_slater_: so to recap your issue (I didn't read all of it properly), configure does detect the log2 function in android-17 even though it isn't available in math.h in that version? what exact NDK version is this?
[11:58] <Loriker> does anyone know the reason for closing this ticket (#4219) ?
[12:37] <rcombs> Loriker: because the report is extremely vague, and outright states that it can't be reproduced on any other machine?
[12:39] <rcombs> &and it was closed by the reporter
[13:35] <akira4> ubitux, would you please have a look at this diff? http://pastebin.com/tfvpDGwv. Still no subtitle stream :( 
[13:45] <cone-93> ffmpeg.git 03Michael Niedermayer 07master:69ee915e1c62: avcodec/parser: add fuzzy mode to ff_fetch_timestamp()
[13:45] <cone-93> ffmpeg.git 03Michael Niedermayer 07master:ad465e789721: avcodec/mpeg12: Try to fetch pts/dts from both the packet associated with the picture start code as well as earlier start codes
[13:56] <wm4> ubitux: I'd rather call it "more lazy" than "less risky" (because you seem to plan to break it anyway)
[14:58] <aetasx> what was it that brought about adding the --enable-random/--disable-random configure flags?  automated testing?
[14:59] <wm4> holy fuck it exists
[14:59] <wm4> pretty fun
[14:59] <aetasx> lol, those flags?
[15:00] <Daemon404> wm4, what does
[15:00] <thardin> do they enable random features?
[15:00] <wm4>   --enable-random          randomly enable/disable components
[15:00] <wm4>   --disable-random
[15:00] <thardin> or just make the binary deterministic?
[15:00] <thardin> lol
[15:00] <Daemon404> o
[15:01] <Daemon404> good for fuzzing maybe
[15:01] <thardin> I wonder how many combinations are possible. probably billions upon billions
[15:01] <Daemon404> indeed
[15:02] <aetasx> I dunno, I saw that and was like.....what possible need could've drove this feature into here
[15:02] <Daemon404> its found quite a number of build failures
[15:03] <aetasx> wouldn't it be more thorough to just bruteforce the combinations though?  you wouldn't have repeats at least
[15:03] <Daemon404> uh no
[15:04] <saste> and now we have this silly reference.ppm with food instead of the beautiful lena...
[15:04] <Daemon404> saste, because lena wast "free" enough for debian
[15:04] <Daemon404> and for some reason we care
[15:04] <Daemon404> (i was against removing it)
[15:04] <wm4> it was also sexist
[15:04] <Daemon404> oh that too. supposedly.
[15:04] <wm4> you should be ashamed
[15:04] <aetasx> lol
[15:05] <saste> wm4, I'm vegetarian and I'm a bit offended by the meat-food in reference.ppm ;-)
[15:05] <aetasx> Daemon404: were you a part of the project when those flags got added?
[15:05] <Daemon404> probably
[15:05] <wm4> saste: surely you agree that women have more rights than cattle
[15:05] <aetasx> interesting
[15:05] <thardin> that or start using burly men in image compression tests, see how long it takes people to feel uncomfortable
[15:06] <saste> wm4, I don't eat women, at least not literally
[15:06] <kierank> use a picture of a certain religious figure
[15:06] <Daemon404> aetasx, if you were to try and build every combination of: config options, enabled muxers, demuxers, parsers, filters, etc
[15:06] <Daemon404> i would probably outlive the universe
[15:06] <Daemon404> too many to try
[15:06] <wm4> use a picture of allah eating pork, get ffmpeg bombed
[15:06] <thardin> aetasx: as I said, there's probably billions of combinations. there's no hope of even ./configuring them all, let alone compiling and linking
[15:07] <aetasx> true, but handling the various tests outside of the script would at least give you more control over it.  was just curious what the history was behind it
[15:07] <thardin> it would perhaps be better to have it outside configure, but whatevs
[15:07] <Daemon404> i dont really follow what 'more control' means
[15:08] <thardin> Daemon404: how do you repeat a full configure + build + run thing when it just does it randomly?
[15:08] <Daemon404> thardin, log it bro
[15:08] <Daemon404> its in config.log
[15:08] <aetasx> meaning you can generate the random lists without the internal randomizer...anyway I wasn't debating its purpose, was just curious of its history
[15:08] <Daemon404> and e.g. FATE keeps all config logs
[15:08] <Daemon404> for all runs
[15:08] <Daemon404> so there is no repeatability issue.
[15:08] <thardin> ah, so it "translates" the config line or somesuch
[15:08] <wm4> y4m for chrome? wtf
[15:08] Action: kierank wonders whether to send a chrome patch to decode vp9 with ffmpeg
[15:09] <Daemon404> aetasx, it was probably so it could be hooked up to FATE
[15:09] <aetasx> gotcha
[15:09] <Daemon404> kierank, will be rejected
[15:09] <Daemon404> libvpx has better ER
[15:09] <wm4> kierank: ffmpeg still can't decode vp9 in webm correctly
[15:10] <Daemon404> all of 0 people care too
[15:10] <Daemon404> because nobody outside of youtube can use it
[15:10] <kierank> i respect the strategy though
[15:10] <kierank> not playing codec wars
[15:11] <kierank> since atm hevc is just noise
[15:11] <compn> whas incorrect about it ?
[15:11] <Daemon404> only annoying since they periodically contact us
[15:11] <Daemon404> with bs info
[15:11] <Daemon404> on how much It Is Great
[15:11] <Daemon404> and try to get us to switch
[15:11] <compn> standard vp9 trolling
[15:11] <wm4> compn: ffmpeg drops timestamps for a certain type of frame
[15:11] <compn> er standard on2 trolling*
[16:10] <aetasx> hey Compn, looks like those K-S-V patches fixed all that PTS nonsense I was getting
[16:12] <compn> aetasx : of course thats why i told you to use ksv build :p
[16:15] <aetasx> I know.  It was totally my fault for assuming the official repo was the best place to get it :p
[16:17] <compn> only took you a month to test? :P
[16:17] <compn> ehe
[16:17] Action: compn kidding
[16:17] Action: compn goes afk
[16:19] <aetasx> yeah I know, theres like 300gigs of video on my disk that I didn't go through
[16:58] <iive> A user seeks help with raw image format. the image seems to be variety of nv12, but with bigger chroma channel (aka nv12->420, nvXX->422). Does somebody know if such format is supported by ffmpeg/swscale ? Or at least how it is named?
[17:00] <iive> hum. nv16 seems like it.
[17:01] <nevcairiel> nv16 is 422, but if swscale likes that i couldnt tell you
[17:04] <kierank> I think it does
[18:23] <cone-93> ffmpeg.git 03Michael Niedermayer 07master:dfffc4aadc7c: avformat/dvbsub: remove unused variable
[18:23] <cone-93> ffmpeg.git 03Michael Niedermayer 07master:e00499eb4c80: avcodec/libxvid: fix potential integer overflow
[18:23] <cone-93> ffmpeg.git 03Michael Niedermayer 07master:2d7fd4cc33d5: ffmpeg_opt: Use av_realloc_array()
[18:23] <cone-93> ffmpeg.git 03Michael Niedermayer 07master:7efabffc2899: ffmpeg_opt: Check for av_realloc_array() failure
[20:13] <akira4> uhm ubitux , are you free right now?
[20:14] <ubitux> akira4: hi
[20:14] <ubitux> sorry, in a moment
[20:14] <ubitux> akira4: btw, i'm going to provide you a dvd sample
[20:14] <ubitux> which i need to upload (might take a while)
[20:14] <ubitux> so we can test on the same data
[20:14] <akira4> that'd be great.
[20:18] <ubitux> akira4: ETA 25 minutes
[20:18] <ubitux> i'll be back then
[20:19] <akira4> alright. I'll stick around till then 
[21:00] <ubitux> wm4: deprecating a side data is way complex and annoying that a whole C structure api with headers
[21:00] <ubitux> you honestly want a full MOV/MP4 specific field in AVFormatContext and AVStreams?
[21:00] <ubitux> with all the structures that it means?
[21:08] <wm4> I just don't want an idiotic API that tells me "fuck you"
[21:09] <wm4> though it might still be slightly better than nothing
[21:10] <wm4> but that's the only thing that speaks for it
[21:56] <cone-947> ffmpeg.git 03Michael Niedermayer 07master:86e7d20ac2e7: avcodec/aic: Use av_malloc_array
[22:47] <cone-947> ffmpeg.git 03Luca Barbato 07master:b3f04657368a: segment: Fix the failure paths
[22:47] <cone-947> ffmpeg.git 03Michael Niedermayer 07master:2a741baeff35: Merge commit 'b3f04657368a32a9903406395f865e230b1de348'
[22:47] <cone-947> ffmpeg.git 03Michael Niedermayer 07master:e2404096b02e: avformat/segment: Clear contexts after deallocation
[23:02] <cone-947> ffmpeg.git 03Luca Barbato 07master:01168bf1405d: mpeg: Remove unused field
[23:03] <cone-947> ffmpeg.git 03Michael Niedermayer 07master:0e1bb03cb034: Merge commit '01168bf1405ddbf671f1074a927b951b8777adbb'
[00:00] --- Thu Jan  8 2015


More information about the Ffmpeg-devel-irc mailing list