[Ffmpeg-devel-irc] ffmpeg-devel.log.20180911

[burek]

[00:03:18 CEST] <atomnuker> that's some spam
[09:48:27 CEST] <LigH> Hi. Trying to build FFmpeg with libaom (both trunk); is there already a known issue with AOM_ERROR_RESILIENT_PARTITIONS ? I did not yet find a trac issue reported.
[10:15:15 CEST] <LigH> If I report this, shall I prefer "avcodec" or "build system" as category?
[10:18:08 CEST] <durandal_1707> LigH: what you want to report?
[10:18:29 CEST] <LigH> A compilation error related to libaomenc.c
[10:19:25 CEST] <LigH> The sttandalone aomenc.exe compiles without error, it happens only in an ffmpeg building sequence.
[10:19:49 CEST] <cone-344> ffmpeg 03Paul B Mahol 07master:7115ad53b424: avcodec/loco: rotate other planes too for non-power of 2 widths
[10:25:22 CEST] <LigH> durandal_1707, "libavcodec/libaomenc.c:744:109: error: 'AOM_ERROR_RESILIENT_PARTITIONS' undeclared here (not in a function); did you mean 'AOM_ERROR_RESILIENT_DEFAULT'?"
[10:34:13 CEST] <LigH> Well, can be fixed later.
[10:36:23 CEST] <JEEB> libaom breaking its API? Why I would have never thought 8)
[10:36:38 CEST] <LigH> No surprise, I know
[10:36:58 CEST] <LigH> But hey, FDK-AAC does it too
[10:37:08 CEST] <JEEB> yea, but at least fdk-aac has more time behind it
[10:37:12 CEST] <JEEB> and x264 also does it
[10:37:23 CEST] <JEEB> but both do it relatively rarely
[10:37:44 CEST] <JEEB> also not sure if there's an easy way to ifdef that
[10:37:53 CEST] <JEEB> because I think libaom made a release?
[10:38:03 CEST] <JEEB> I mean when AV1 froze
[10:38:08 CEST] <LigH> Could it be that they took more time to think ahead, what may happen in the future, to design an API with reserve for new ideas?
[10:38:36 CEST] Action: JEEB moves back to $dayjob
[10:39:21 CEST] <LigH> My dayjob is similar. Solve half the issue now. Later, the rest of the issue implies a database redesign...
[10:39:31 CEST] <LigH> Bye
[11:25:53 CEST] <cone-344> ffmpeg 03Paul B Mahol 07master:06e990ce8911: avfilter/vf_limiter: add slice threading
[12:01:46 CEST] <cone-344> ffmpeg 03Paul B Mahol 07master:492b312ddd85: avfilter/vf_datascope: reduce block height
[12:50:40 CEST] <durandal_1707> i can't force myself to work on bink 2 anymore
[13:07:31 CEST] <BBB> durandal_1707: so don't
[13:07:42 CEST] <BBB> you should do things that you think are fun
[13:16:28 CEST] <durandal_1707> BBB: working on Dav1d is fun
[13:46:50 CEST] <BBB> durandal_1707: I agree!
[13:47:03 CEST] <BBB> durandal_1707: come to VDD and we can talk about it
[14:31:38 CEST] <cone-344> ffmpeg 03Karthick Jeyapal 07master:606879ec4298: avformat/dashenc: Add CODECS tag to HLS master playlist
[14:31:40 CEST] <cone-344> ffmpeg 03Karthick Jeyapal 07master:44ee858b6df0: avformat/dashenc: Remove muxer overhead from Bandwidth field in DASH manifest
[15:19:23 CEST] <maxim_d33> those qsv related trac reports - will be more meaningful way of dealing 
[18:09:51 CEST] <cone-344> ffmpeg 03Luca Barbato 07master:f89ec87afaf0: frame: Simplify the video allocation
[18:09:52 CEST] <cone-344> ffmpeg 03James Almer 07master:8822e2b9543b: Merge commit 'f89ec87afaf0d1abb6d450253b0b348fd554533b'
[18:20:43 CEST] <kurosu> /ignore fflogger
[18:23:05 CEST] <durandal_1707> how one interpolate linearly given only integers?
[18:26:09 CEST] <jamrial> durandal_1707: you mean in x86 asm?
[18:27:42 CEST] <durandal_1707> jamrial: nope, in C
[18:27:59 CEST] <jamrial> oh, nevermind. i misread tha as interleave
[18:32:50 CEST] <kurosu> (not a request) Now that the participants list is finalized, neither of you 2 are coming to vdd ?
[18:33:12 CEST] <atomnuker> durandal_1707: same as floats but with fixed point math
[18:34:03 CEST] <cone-344> ffmpeg 03Maxym Dmytrychenko 07master:325aa63dd1a3: qsv: enforcing continuous memory layout
[18:34:04 CEST] <cone-344> ffmpeg 03Andrey Utkin 07master:5d01bd181bb7: http: pass return code from http_open_cnx_internal() on its failure
[18:34:05 CEST] <cone-344> ffmpeg 03Simon Thelen 07master:c194b9ad6dbe: network: Use ff_neterrno instead of AVERROR(errno) for poll errors
[18:34:06 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:22f98ac19cf2: network: Check for EINTR in ff_poll_interrupt
[18:34:07 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:6a9c00c09d2b: tls_openssl: Fix checks for SSL_ERROR_WANT_WRITE in nonblocking operation
[18:34:08 CEST] <cone-344> ffmpeg 03James Almer 07master:e1e1e8dbb204: bsf: add a flushing mechanism to AVBSFContext
[18:34:09 CEST] <cone-344> ffmpeg 03James Almer 07master:d6321851ba35: h264_mp4toannexb_bsf: implement a AVBSFContext.flush() callback
[18:34:10 CEST] <cone-344> ffmpeg 03James Almer 07master:eb1d1c764c7f: vp9_superframe_split_bsf: implement a AVBSFContext.flush() callback
[18:34:11 CEST] <cone-344> ffmpeg 03James Almer 07master:7f01c209f219: vp9_superframe_bsf: implement a AVBSFContext.flush() callback
[18:34:12 CEST] <cone-344> ffmpeg 03James Almer 07master:0e27e2767001: h264_redundant_pps_bsf: implement a AVBSFContext.flush() callback
[18:34:13 CEST] <cone-344> ffmpeg 03James Almer 07master:ad99cbc9b3f4: decode: flush the internal bsfs instead of constantly reinitalizing them
[18:34:14 CEST] <cone-344> ffmpeg 03James Almer 07master:662558f985f5: decode: copy the output parameters from the last bsf in the chain back to the AVCodecContext
[18:34:15 CEST] <cone-344> ffmpeg 03James Almer 07master:14fe81b3a88d: Merge commit '325aa63dd1a3abc2453914d0bc111d297833d725'
[18:34:16 CEST] <cone-344> ffmpeg 03James Almer 07master:1a8be90adbaf: Merge commit '5d01bd181bb77e6740462095d7be4e0733a59420'
[18:34:17 CEST] <cone-344> ffmpeg 03James Almer 07master:5c1aafff2d84: Merge commit 'c194b9ad6dbe65f5abd68158c4811ed84e2a2b95'
[18:34:18 CEST] <cone-344> ffmpeg 03James Almer 07master:03210fe138f3: Merge commit '22f98ac19cf29f22b3e1d10314df9503f06fe683'
[18:34:19 CEST] <cone-344> ffmpeg 03James Almer 07master:f3881c04e93e: Merge commit '6a9c00c09d2bc50c0ea64ba092b2f4afc46aa978'
[18:34:20 CEST] <cone-344> ffmpeg 03James Almer 07master:0ff76ca86e0e: Merge commit '662558f985f50834eebe82d6b6854c66f33ab320'
[18:34:51 CEST] <durandal_1707> atomnuker: using floats is not issue
[18:38:10 CEST] <durandal_1707> https://github.com/richardpl/FFmpeg/commit/65fdab9e9c1e6cda4eee8088af50ca7d9f021890
[18:38:28 CEST] <durandal_1707> what to use for tx/ty?
[18:40:36 CEST] <atomnuker> are you converting lerpf to use ints only?
[18:40:50 CEST] <durandal_1707> atomnuker: no
[18:42:30 CEST] <durandal_1707> i need factors/average that do actual interpolation
[18:43:29 CEST] <cone-344> ffmpeg 03Zhong Li 07master:e05e5920a4e1: qsv: Error out if getting session handle failed in avfilter
[18:43:30 CEST] <cone-344> ffmpeg 03Zhong Li 07master:76eef04f30a7: qsvenc: Fix a misleading log message
[18:43:31 CEST] <cone-344> ffmpeg 03Zhong Li 07master:69caad895998: qsvdec: Release packet on decoding failure for  mpeg2/vp8/vc1
[18:43:32 CEST] <cone-344> ffmpeg 03James Almer 07master:185aa5e896e1: Merge commit 'e05e5920a4e1f1f15cc8a7c843159d519f6ec18e'
[18:43:33 CEST] <cone-344> ffmpeg 03James Almer 07master:50ae1f7e0ff1: Merge commit '76eef04f30a768fa80366d679f3de7e9447b67d5'
[18:43:34 CEST] <cone-344> ffmpeg 03James Almer 07master:23ce57af3ad6: Merge commit '69caad8959982580504643d36aef22528e4aa6ce'
[18:56:59 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:9b4c3f5aadf5: network: Add RFC 8305 style "Happy Eyeballs"/"Fast Fallback" helper function
[18:57:00 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:8c76bfacf663: tcp: Use ff_connect_parallel for RFC 8305 style connecting
[18:57:01 CEST] <cone-344> ffmpeg 03James Almer 07master:762c2b5dcd99: Merge commit '9b4c3f5aadf54ffd2a6e15746b1fd736379883c4'
[18:57:02 CEST] <cone-344> ffmpeg 03James Almer 07master:ef71ef5f30dd: Merge commit '8c76bfacf663ff71cee5264a74d0f9c86addd325'
[19:02:29 CEST] <atomnuker> does 8c76bfacf663 make dash work better?
[19:02:41 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:83678dbbae64: libopenh264dec: Export the decoded profile and level in AVCodecContext
[19:02:42 CEST] <cone-344> ffmpeg 03Zhong Li 07master:e16b20782a59: lavc/qsvenc: allow to set qp range for h264 BRC
[19:02:43 CEST] <cone-344> ffmpeg 03Zhong Li 07master:c8bca9fe466f: lavc/qsvenc: dump BufferSizeInKB message
[19:02:44 CEST] <cone-344> ffmpeg 03James Almer 07master:ff44c2d4f483: Merge commit '83678dbbae64ad8c501e0c732c1117e642c25dae'
[19:02:45 CEST] <cone-344> ffmpeg 03James Almer 07master:27e30c73d722: Merge commit 'e16b20782a597e36a9c7488487c3179375a25b97'
[19:02:46 CEST] <cone-344> ffmpeg 03James Almer 07master:a12063b118ad: Merge commit 'c8bca9fe466f810fd484e2c6db7ef7bc83b5a943'
[19:03:31 CEST] <nevcairiel> atomnuker: only if you had ipv4/v6 troubles
[19:21:14 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:141c960e21d2: libfdk-aacenc: Fix building with libfdk-aac v2
[19:21:15 CEST] <cone-344> ffmpeg 03Joe Olivas 07master:642fd4769bec: qsvvpp: Perform full init only when needed
[19:21:16 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:ffb9b7a6bab6: libfdk-aac: Consistently use a proper version check macro for detecting features
[19:21:17 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:2edaafe5b938: libfdk-aacdec: Allow setting the new dynamic range control effect setting
[19:21:18 CEST] <cone-344> ffmpeg 03Martin Storsjö 07master:7e929dac1009: libfdk-aacenc: Allow enabling the ELDv2 profile
[19:21:19 CEST] <cone-344> ffmpeg 03James Almer 07master:ee96ab2db507: Merge commit '141c960e21d2860e354f9b90df136184dd00a9a8'
[19:21:20 CEST] <cone-344> ffmpeg 03James Almer 07master:f8377ffce352: Merge commit '642fd4769becc2f4827f8375a3d9e8edd2f5df77'
[19:21:21 CEST] <cone-344> ffmpeg 03James Almer 07master:203bbaccfaae: Merge commit 'ffb9b7a6bab6c6bfd3dd9a7c32e3724209824999'
[19:21:22 CEST] <cone-344> ffmpeg 03James Almer 07master:94d98330ed6c: Merge commit '2edaafe5b93832715781851dfe2663da228a05ad'
[19:21:23 CEST] <cone-344> ffmpeg 03James Almer 07master:c7488f746154: Merge commit '7e929dac100916fc45cb95e231025f3439c20156'
[19:21:25 CEST] <JEEB> wee
[19:21:27 CEST] <JEEB> fdk-aac stuff
[19:22:30 CEST] <durandal_1707> so nobody understands linear interpolation?
[19:27:48 CEST] <jamrial> durandal_1707: michaelni probably does, judging by his swr code
[20:22:35 CEST] <kurosu> durandal_1707, I don't understand, because I assume my answer is incorrect, otherwise you'd have thought of it
[20:22:50 CEST] <kurosu> you want to work in 'fixed point' precision
[20:23:07 CEST] <kurosu> the danger is properly evaluating precision to avoid overflows
[20:23:45 CEST] <kurosu> say you are pretty sure the values are in [-2^24, 2^24]
[20:23:56 CEST] <kurosu> that leaves you around 7 bits of precision left for 32b
[20:24:22 CEST] <kurosu> so you use tx = 0.5*(1<<N) with N your precision, ie 1<<(N-1)
[20:24:58 CEST] <kurosu> and all through computations you take into account that
[20:25:50 CEST] <kurosu> s + (e - s) * t; <- as float, so as in it is (s<<N) + (((e-s)<<N) * (t<<N))
[20:26:08 CEST] <kurosu> to get the rounded integer, you right shift by N
[20:26:22 CEST] Action: kurosu feels like captain obvious
[20:27:32 CEST] <kurosu> (s*(1<<N) expressed as integer but you get the point)
[20:31:54 CEST] <durandal_1707> kurosu: i already said, that floats can be used
[20:32:36 CEST] <durandal_1707> even if i wanted fixed point, from where I would get tx and ty?
[20:32:43 CEST] <kurosu> oh, right, I missed the line in between your url and the cone-344 spam
[20:34:07 CEST] <kurosu> what is the interpolation filter you want to use ? bilinear, lanczos, some other?
[20:34:40 CEST] <durandal_1707> kurosu: bilinear
[20:34:55 CEST] <kurosu> it's just a weighted average
[20:35:21 CEST] <kurosu> if point is at n+dx with p[n] and p[n+1] known
[20:35:41 CEST] <kurosu> then it's (1-dx)*p[n] + dx*p[n+1]
[20:35:53 CEST] <kurosu> rince and repeat for any number of dimensions
[20:36:37 CEST] <durandal_1707> i have 4 points for bilinear interpolation input already
[20:36:51 CEST] <durandal_1707> how to get dx/dy
[20:37:10 CEST] <kurosu> (though you do interpolation of all points for the next dimensions, so (1-dy)*((1-dx)*p[n+0*stride] + dx*p[n+1+0*stride]) + dy*((1-dx)*p[n+1*stride] + dx*p[n+1+1*stride])
[20:37:38 CEST] <durandal_1707> https://github.com/richardpl/FFmpeg/commit/65fdab9e9c1e6cda4eee8088af50ca7d9f021890
[20:38:23 CEST] <kurosu> this computes the point half-way between c00 c10 c01 and c11
[20:39:19 CEST] <kurosu> if you think the position is slightly shifted see the equation above
[20:40:00 CEST] <durandal_1707> i do not want to have tx/ty hardcoded to .5
[20:40:03 CEST] <kurosu> atomnuker could have said as much, but he's likely away
[20:40:32 CEST] <kurosu> you have to know how it is biased towards one point or another
[20:40:52 CEST] <kurosu> if the interpolated point is right in the middle, then 0.5 is the value to use for bilinear
[20:41:22 CEST] <kurosu> if it is, say, 1/4 away, then you get it, weights will be 0.75 and 0.25
[20:42:57 CEST] <kurosu> say another way: compute the parameters of the line passing through 2 points, then, having it parameterized, you can then compute any point on that line
[20:43:03 CEST] <kurosu> *said another
[20:54:22 CEST] <cone-344> ffmpeg 03James Almer 07master:092cbbfa74d8: avformat/tcp: re-add checks for setsockopt return values
[20:54:46 CEST] <durandal_1707> kurosu: correct answer is finding centroid
[22:19:32 CEST] <jkqxz> Anyone want to say anything about AVS2?
[22:21:17 CEST] <JEEB> it seems like actual content in it doesn't exist or something
[22:21:30 CEST] <JEEB> and almost everything goes back to those two or so libraries
[22:24:28 CEST] <jkqxz> I feel like there must be more in Chinese-only places.  Hardware decoders are clearly available for it.
[22:25:04 CEST] <JEEB> yea, I tried to baidu my way somewhere but ended up in authentication-only
[22:25:45 CEST] <durandal_1707> ask for samples?
[22:26:04 CEST] <JEEB> I did
[22:26:14 CEST] <durandal_1707> and?
[22:26:15 CEST] <jkqxz> JEEB did, and was given some obviously created with that encoder.
[22:26:19 CEST] <JEEB> ^
[22:36:43 CEST] <jkqxz> Oh well.  I think it's sensible to have so I'm going to apply it.
[22:37:44 CEST] <JEEB> I'd just double-check that the license is acceptable in the libraries and if the patches look acceptable then... yea
[22:39:07 CEST] <jkqxz> GPL2 everywhere.  (It's based on libx264, so it has to be.)
[22:40:00 CEST] <JEEB> ok
[22:40:19 CEST] <JEEB> I just remembered some chinese projects having some GPL plus no corporate usage etc licenses
[22:44:23 CEST] <durandal_1707> EU gonna kill internet anyway, tomorrow 
[23:02:38 CEST] <J_Darnley> Well, the Internet had a good run.
[23:03:24 CEST] <JEEB> the EU kills everything yet we're still eating cuecumbers
[23:03:35 CEST] <J_Darnley> I'd better check that I have all I need from it.
[23:05:26 CEST] <jamrial> jkqxz: do you have libdrm? does it show up for you in avutil extralibs when you enable it during configure?
[23:07:22 CEST] <jkqxz> I'm unsure how to check that?
[23:07:38 CEST] <jkqxz> It's in the output pkgconfig file.
[23:07:48 CEST] <jamrial> run configure, check what config.mak lists in EXTRALIBS-avutil
[23:07:52 CEST] <jkqxz> ls
[23:07:53 CEST] <jamrial> or that, in libavutil.pc
[23:07:59 CEST] <jamrial> mmh
[23:08:00 CEST] <jkqxz> Oops.
[23:08:29 CEST] <jkqxz> "EXTRALIBS-avutil=-pthread -lva -lva-drm -lva -lva-x11 -lX11 -lvdpau -lX11 -lm -L/usr/local/lib -ldrm -L/usr/local/lib -lmfx -lstdc++ -ldl -lva-drm -lva -lva-x11 -lva -lOpenCL -lva"
[23:08:38 CEST] <jkqxz> The "-L/usr/local/lib -ldrm" is from libdrm.
[23:09:01 CEST] <jamrial> well, that's odd
[23:09:16 CEST] <jamrial> i'm replying to this libdrm patch for the build system
[23:09:17 CEST] <jkqxz> Let me try with fewer things enabled.
[23:09:30 CEST] <jamrial> the guy reports it's not being listed for libavuil
[23:09:32 CEST] <jamrial> http://autobuild.buildroot.net/results/515/5152dcca58944cf732d09fba6e6c9af8a9243c75/build-end.log
[23:09:39 CEST] <jamrial> but that log shows it's listed for libavdevice
[23:10:25 CEST] <jkqxz> Was that right in version 3.4.4?
[23:10:29 CEST] <jamrial> i don't know
[23:10:41 CEST] <jamrial> 3.4 was before the configure dep overhaul
[23:11:03 CEST] <jkqxz> "--disable-autodetect --enable-libdrm" gives me "EXTRALIBS-avutil=-pthread -lm -L/usr/local/lib -ldrm".
[23:11:09 CEST] <jamrial> so i think this can't be 3.4
[23:12:32 CEST] <durandal_1707> iiuc op have issue with dynamic build only
[23:13:00 CEST] <jamrial> "This situation is valid only in the case you link *statically*. For dynamic linking there's no problem at all."
[23:13:06 CEST] <jamrial> no, the other way around
[23:13:54 CEST] <durandal_1707> ah, yes
[23:21:34 CEST] <cone-344> ffmpeg 03hwren 07master:85a921f46184: lavc, doc, configure: add libxavs2 video encoder wrapper
[23:21:35 CEST] <cone-344> ffmpeg 03hwren 07master:ed0bd0a71b5e: lavf: add raw avs2 muxer
[23:24:22 CEST] <jamrial> jkqxz: ah, after looking again i guess this could be 3.4 before the overhaul. but then it wouldn't make sense making a patch on top of git master
[23:25:05 CEST] <jkqxz> I was going by the ffmpeg-3.4.4 in the log above.
[23:25:16 CEST] <jamrial> oh, missed that
[23:25:23 CEST] <jamrial> well, he just confirmed it
[23:25:33 CEST] <jamrial> now, how to fix this for a point release...
[23:26:00 CEST] <durandal_1707> raw avs2 muxer missing makefile line?
[23:34:10 CEST] <cone-344> ffmpeg 03Giulio Benetti 07release/3.4:c50dc77ac708: configure: add LIBDRM to extralibs_avutil
[23:35:26 CEST] <jkqxz> durandal_1707:  Yes, oops, sorry.  Will fix in a moment...
[23:39:14 CEST] <cone-344> ffmpeg 03Mark Thompson 07master:ff21d4735ed0: lavf/Makefile: Fix standalone build of AVS2 muxer
[00:00:00 CEST] --- Wed Sep 12 2018