[Ffmpeg-devel-irc] ffmpeg-devel.log.20150518

burek burek021 at gmail.com
Tue May 19 02:05:02 CEST 2015


[00:48:31 CEST] <Antony> Hello 
[00:48:32 CEST] <Antony> I have an error when compiling ffmpeg 
[00:48:32 CEST] <Antony> this the error: gcc is unable to create an executable file. 
[00:48:32 CEST] <Antony> If gcc is a cross-compiler, use the --enable-cross-compile option. 
[00:48:32 CEST] <Antony> Only do this if you know what cross compiling means. 
[00:48:32 CEST] <Antony> C compiler test failed 
[00:48:32 CEST] <Antony> https://www.filepicker.io/api/file/4Z61EL0BTaqPhT2OBDsR
[00:49:53 CEST] <Daemon404> libnettle.so doesnt exist in any library dir
[00:50:02 CEST] <Daemon404> you, for some reason, put it in extra ldflags.
[00:50:05 CEST] <Daemon404> also wrong channel.
[00:52:44 CEST] <Antony> sorry thanks
[00:53:32 CEST] <cone-898> ffmpeg 03Michael Niedermayer 07master:57e38043def6: ffmpeg: Do not copy the display matrix if rotation meta-data is manually added
[00:58:59 CEST] <jamrial> michaelni: you wrote the wrong sha1 value in that ticket, in case you care about fixing it
[01:00:35 CEST] <michaelni> fixed, thx
[02:18:44 CEST] <cone-898> ffmpeg 03Michael Niedermayer 07master:66f4b1e9c4a4: cmdutils: Fix sign error in display matrix auto-rotation code
[02:32:06 CEST] <BBB> michaelni: for my patchset, do you want to wait longer for reviews, or shall I give you a merge tree?
[02:33:11 CEST] <michaelni> BBB, depends on what you prefer, i see no reason to wait from my side
[02:33:51 CEST] <BBB> ok, https://github.com/rbultje/ffmpeg/commits/vp9-bugfixes is current tree
[02:34:04 CEST] <BBB> all based on samples that I can unfortunatey not share :( so no new fate tests
[02:34:05 CEST] <BBB> but ohwell
[02:34:37 CEST] <BBB> Im still not sure about the changes to the x86 assembly, but Ill look at that closer sometime soon
[02:44:31 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:96d30c34951b: vp9: disable all pmulhrsw in 8/16 iadst x86 optimizations.
[02:44:32 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:d32d0593f172: vp9: disable more pmulhrsw optimizations in idct16/32.
[02:44:33 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:e12188e1438b: vp9: fix segmentation map referencing upon framesize change.
[02:44:34 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:dc96c0f9fc96: vp9: read all 4x4 blocks in sub8x8 blocks individually with scalability.
[02:44:35 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:1e4a77d0018e: vp9: apply mv scaling workaround only when subsampling is enabled.
[02:44:36 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:96a58a8daa4d: vp9: set skip flag if the block had no coded coefficients.
[02:44:37 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:29045fbfd4c7: vp9: clip motion vectors in the same way as libvpx does.
[02:44:38 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:3e634e3e9821: vp9: extend loopfilter workaround for vp9 h/v mix-up to work for 422.
[02:44:39 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:88126916c819: vp9: fix crash when playing back 440/440 content with width%64<56.
[02:44:40 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:c81677e9b73c: vp9: reset loopfilter mode/ref deltas on keyframe.
[02:44:41 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:5de142d316a6: vp9: clamp segmented lflvl before applying ref/mode deltas.
[02:44:42 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:32b6d31ef3ab: vp9: don't allow compound references if error_resilience is enabled.
[02:44:43 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:68c1e9131654: vp9: improve signbias check.
[02:44:44 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:ccfb03ecc4f6: vp9: fix scaled motion vector clipping for sub8x8 blocks.
[02:44:45 CEST] <cone-898> ffmpeg 03Ronald S. Bultje 07master:900e3af85787: vp9: match another find_ref_mvs() bug in libvpx.
[02:44:46 CEST] <cone-898> ffmpeg 03Michael Niedermayer 07master:451be676f3ff: Merge remote-tracking branch 'rbultje/vp9-bugfixes'
[03:01:23 CEST] <BBB> michaelni: ty
[06:06:32 CEST] <Compn> kierank : thats a nice pdf on ffv1 :)
[06:06:39 CEST] <Compn> i wonder if lou will post that to ffmpeg twitter, http://download.das-werkstatt.com/pb/fsfe/presentations/2015/the_ffv1_story.pdf
[06:06:50 CEST] <Compn> or whoever runs facebook too
[09:15:29 CEST] <j-b> morning
[09:24:38 CEST] <TimNich> morning j-b
[11:09:39 CEST] <cone-106> ffmpeg 03Michael Niedermayer 07master:58afb3128bc6: avformat/mxfenc: Set the component depth from the pixel format if available
[11:09:39 CEST] <cone-106> ffmpeg 03Michael Niedermayer 07master:e32d832a8211: avformat/mxfenc: Set horizontal chroma subsample value from pixel format if available
[13:42:55 CEST] <cone-106> ffmpeg 03Michael Niedermayer 07master:0c0d1a4ef202: avfilter/avfiltergraph: Rename ff_avfilter_graph_config_pointers to graph_config_pointers.
[13:42:56 CEST] <cone-106> ffmpeg 03Michael Niedermayer 07master:c6bf27d59be6: avcodec/hevc_cabac: Rename ff_hevc_transform_skip_flag_decode() to hevc_transform_skip_flag_decode()
[14:27:53 CEST] <cone-106> ffmpeg 03Michael Niedermayer 07master:4b0f78ad1f5e: avcodec/snow_dwt: Remove ff_ prefix from ff_spatial_idwt_init() and ff_spatial_idwt_slice()
[14:31:25 CEST] <BBB> michaelni: your lpc patch never sets improved=1
[14:31:28 CEST] <BBB> (its commented out)
[14:31:37 CEST] <BBB> so the loop seems to never run more than once
[14:31:39 CEST] <BBB> is that intentional?
[14:57:10 CEST] <aftab> Hello, I am trying to install ffmpeg with xampp on windows 7, i have downloaded ffmpeg_new and copied php_ffmpeg.dll in ext directory and also the other 2 dll files in windows\system32 directory but it is not loading. Whenever I run php it gives me this error: Unable to load dynamic link library 'D:\xampp\php\ext\php_ffmpeg.dll' - The specified procedure could not be found.
[14:57:20 CEST] <aftab> Please let me know how to fix it ?
[15:05:17 CEST] <aftab^> Hello, I am trying to install ffmpeg with xampp on windows 7, i have downloaded ffmpeg_new and copied php_ffmpeg.dll in ext directory and also the other 2 dll files in windows\system32 directory but it is not loading. Whenever I run php it gives me this error: Unable to load dynamic link library 'D:\xampp\php\ext\php_ffmpeg.dll' - The specified procedure could not be found.
[16:58:12 CEST] <michaelni> BBB, setting  improved=1 didnt seem to improve it, its either buggy or already close enough to the minimum
[19:59:37 CEST] <relaxed> When compiling ffmpeg with libx265's high bit depth enabled, "ffmpeg -h encoder=libx265" doesn't list any supported pixel formats.
[20:09:26 CEST] <Daemon404> because theyre set at runtime.
[20:09:33 CEST] <Daemon404> libx264 is the same wya
[20:10:16 CEST] <relaxed> no, ffmpeg -h encoder=libx264 will list the 10 bit pixel formats, and so does vp9
[20:10:43 CEST] <Daemon404> vp9 is unrelated here
[20:11:35 CEST] <relaxed> well, if you enable its high bit depths they're listed with ffmpeg -h encoder=libvpx-vp9
[20:12:26 CEST] <relaxed> it would be helpful if libx265 did as well :)
[20:12:37 CEST] <Daemon404> there is nothing different in libx265.c though.
[20:12:44 CEST] <Daemon404> it does it teh exact same way libx264 does
[20:14:23 CEST] <jamrial> yet it's not working. just tried and same happens with libx265 8bit. No pixel formats are listed
[20:14:47 CEST] <relaxed> ok, I just think it's helpful that libx264 lists "Supported pixel formats: yuv420p10le yuv422p10le yuv444p10le nv20le"
[20:15:47 CEST] <relaxed> jamrial: 8bit libx265 build shows "Supported pixel formats: yuv420p yuv422p yuv444p" here
[20:15:55 CEST] <Daemon404> jamrial, no idea why
[20:15:58 CEST] <Daemon404> it's literally the same
[20:16:09 CEST] <jamrial> relaxed: not for me, but let me recompile git head and check again
[20:17:45 CEST] <jamrial> i'm using libx265 shared on mingw64, though. will try static as well
[20:18:35 CEST] <jamrial> also, there's a patch by steve borho on the ml that changes the static init function that deals with this. maybe it's related?
[20:18:50 CEST] <Daemon404> jamrial, that will only add the 10bit formats if available
[20:18:56 CEST] <Daemon404> it wouldnt fix *no* formats showigm
[20:20:22 CEST] <Daemon404> https://github.com/videolan/x265/blob/069168543c08a994dbecae39073d73c38a1c41d3/source/common/version.cpp#L84
[20:20:25 CEST] <Daemon404> fuck me
[20:20:29 CEST] <Daemon404> the ysilently changed it from 12 to 10
[20:20:42 CEST] <Daemon404> although that would not explain why it does not work with 8bit
[20:20:58 CEST] <relaxed> it does work with 8bit
[20:21:02 CEST] <Daemon404> https://github.com/videolan/x265/commit/9f2269206bceab4963f5c86be40e43a4012d2830
[20:21:06 CEST] <Daemon404> this broke it
[20:21:08 CEST] <Daemon404> silently.
[20:21:11 CEST] <jamrial> ok, shared mingw doesnt' work
[20:21:13 CEST] <jamrial> static does
[20:21:16 CEST] <jamrial> 8bit, i mean
[20:21:23 CEST] <Daemon404> i am going to push steve's patch today
[20:21:27 CEST] <Daemon404> it will fix it regardless
[20:21:43 CEST] <Daemon404> (but only by coincidence)
[20:23:02 CEST] <Daemon404> jamrial, do you have push rights
[20:23:18 CEST] <jamrial> yes, why?
[20:23:30 CEST] <Daemon404> can you push steve's patch? I ok'd it yesterday.
[20:23:36 CEST] <Daemon404> my ffmpeg vm is down and i mlazy
[20:23:37 CEST] <Daemon404> <_<
[20:23:52 CEST] <jamrial> haha, ok :p
[20:29:21 CEST] <cone-106> ffmpeg 03Steve Borho 07master:b012bd50685e: libavcodec/x265: detect csps in libx265_encode_init_csp()
[20:30:21 CEST] <Daemon404> thanks
[20:30:51 CEST] <jamrial> np
[20:44:27 CEST] <jamrial> Daemon404: this may need to be fixed in maintained branches. not by using the new x265_api_get() stuff but by replacing the == 12 line with <= 12 or such
[20:56:26 CEST] <Daemon404> "maintained"
[20:56:35 CEST] <Daemon404> sporatically maybe
[20:59:30 CEST] <rcombs> hmm
[20:59:35 CEST] <jamrial> well, https://trac.ffmpeg.org/wiki/Downstreams 2.6 at the very least
[20:59:49 CEST] <jamrial> it got a point release yesterday, even
[21:00:28 CEST] <rcombs> think it'd be useful to provide an AVDictionary on AVCodecContext for additional codec properties that don't map directly into the members?
[21:01:24 CEST] <rcombs> e.g. whether or not an H.264 stream has a sequence scaling matrix, or uses CABAC
[21:04:17 CEST] <rcombs> I suppose for simple booleans, an additional int field and a public AV_CODEC_FLAG_H264_CABAC or the like would do
[21:04:38 CEST] <Daemon404> sidedata maybe?
[21:06:55 CEST] <rcombs> hmmm
[21:07:26 CEST] <rcombs> the keen-eyed reader will notice that I'm trying to make lavc a functional substitute for libmediainfo
[21:07:37 CEST] <rcombs> because mediainfo is god-awful
[21:08:05 CEST] <Daemon404> i have much experience with libmediainfo
[21:08:28 CEST] <Daemon404> i do not think that kind of stuff fits anywhere but metadata
[21:08:29 CEST] <Daemon404> er
[21:08:31 CEST] <Daemon404> but sidedata
[21:08:45 CEST] <Daemon404> adding new fields to that end is just ... ew
[21:10:21 CEST] <rcombs> AVCodecContext is already full of ridiculously specific crap like that (including some that aren't documented!)
[21:10:46 CEST] <rcombs> side data wouldn't suck too much, I suppose
[21:11:20 CEST] <Daemon404> "but some bad decisiosn were made in the past" is not a sufficient argument for "lets make more bad decisions"
[21:11:25 CEST] <rcombs> obviously
[21:11:25 CEST] <Daemon404> re: avctx fields
[21:11:43 CEST] <rcombs> which is why I'd like to come up with something better
[21:12:14 CEST] <Daemon404> consider that lavc and pals were never intended to fill the same niche as mediainfo
[21:12:21 CEST] <Daemon404> anything tacked on will be horrible
[21:12:29 CEST] <rcombs> a single new bitmask field for codec-specific wouldn't be particularly evil
[21:12:40 CEST] <Daemon404> no new bitmasks
[21:12:41 CEST] <rcombs> *codec-specific flags
[21:12:41 CEST] <Daemon404> EVER
[21:12:47 CEST] <Daemon404> because 64 is NEVER enough
[21:12:52 CEST] <Daemon404> you *will* run out of bits.
[21:13:28 CEST] <rcombs> are there really any codecs with more than 64 boolean options you might care about?
[21:13:48 CEST] <rcombs> (in my case, it's for deciding whether or not a player will be able to handle it)
[21:14:10 CEST] <Daemon404> youre going to need so many awful enums or defines for all the flags in the oublic header.
[21:14:33 CEST] <Daemon404> and i do not think it is lavc's responsibility to provide info for non-lavc decoders
[21:14:45 CEST] <Daemon404> especially not when it requeis an abominablre api
[21:14:48 CEST] <Daemon404> abominable*
[21:15:33 CEST] <Daemon404> codec-specific flags defined in avcodec.h is pretty much the worst thing i can think of
[21:15:36 CEST] <Daemon404> aside from a new avctx field
[21:15:59 CEST] <rcombs> you'd prefer, what, a codec-specific struct in side_data?
[21:16:14 CEST] <rcombs> or a bunch of codec-specific ints holding bytes in side_data?
[21:16:21 CEST] <rcombs> *ints holding bools
[21:16:26 CEST] <Daemon404> a dictionary is the least bad thing i can think of
[21:16:27 CEST] <Daemon404> key/val
[21:16:53 CEST] <rcombs> that was my first thought
[21:16:53 CEST] <Daemon404> but remember i dont necessarily agere lavc is even suited to this.
[21:21:14 CEST] <wm4> side data is the answer to everything
[21:21:44 CEST] <Daemon404> wm4, why can't admitting that maybe, just maybe, some things are not suited to be in libav* be teh answer
[21:21:47 CEST] <Daemon404> to anything
[21:21:47 CEST] <Daemon404> ever
[21:22:00 CEST] <rcombs> Daemon404: well, what _is_ the answer here?
[21:22:02 CEST] <wm4> because lavc is monolithic
[21:22:20 CEST] <Daemon404> rcombs, either or.
[21:23:37 CEST] <Daemon404> one thing mediainfo has that lavc will never do, is reporting it eitehr correctly, or not at all.
[21:23:43 CEST] <Daemon404> rather than making up "best effor" info
[21:23:48 CEST] <rcombs> my current one is libmediainfo, but that means fuckery with locale shit, matching streams between mediainfo and lav*, and an API made of parsing lots and lots of strings
[21:24:11 CEST] <Daemon404> been there done that
[21:24:19 CEST] <Daemon404> except i was sane enough to not do teh string parsing in C.
[21:24:26 CEST] <rcombs> (C++ here)
[21:45:23 CEST] <cone-106> ffmpeg 03Michael Niedermayer 07master:b0322e4a9e32: ffmpeg_opt: Compensate for DTS/PTS difference in seeking when its based on DTS
[00:00:00 CEST] --- Tue May 19 2015


More information about the Ffmpeg-devel-irc mailing list