[Ffmpeg-devel-irc] ffmpeg-devel.log.20170312

Mon Mar 13 03:05:04 EET 2017

[03:11:45 CET] <chatter29> hey guys
[03:11:48 CET] <chatter29> allah is doing
[03:11:55 CET] <chatter29> sun is not doing allah is doing
[03:11:57 CET] <chatter29> to accept Islam say that i bear witness that there is no deity worthy of worship except Allah and Muhammad peace be upon him is his slave and messenger
[03:22:13 CET] <Compn> this allah guy sounds needy
[04:38:49 CET] <cone-347> ffmpeg 03Michael Niedermayer 07master:47cc9c1d77f5: avcodec/wavpack: Fix runtime error: signed integer overflow: -2147483648 + -83886075 cannot be represented in type 'int'
[04:38:49 CET] <cone-347> ffmpeg 03Michael Niedermayer 07master:f4c2302ee24d: avcodec/dca_xll: Fix runtime error: signed integer overflow: 1762028192 + 698372290 cannot be represented in type 'int'
[04:38:49 CET] <cone-347> ffmpeg 03Michael Niedermayer 07master:44e2105189ac: avcodec/amrwbdec: Fix  runtime error: left shift of negative value -1
[04:38:49 CET] <cone-347> ffmpeg 03Michael Niedermayer 07master:ce010655a6b8: avcodec/dca_xll: Fix runtime error: signed integer overflow: 2147286116 + 6298923 cannot be represented in type 'int'
[11:33:58 CET] <paras2052> durandal_1707: Did you see the latest patch i sent ?
[11:48:35 CET] <durandal_1707> paras2052: yes i did
[11:48:51 CET] <paras2052> durandal_1707: is it fine ?
[11:49:28 CET] <durandal_1707> mostly
[11:50:32 CET] <paras2052> okay, what do i have to do next. any other task ?
[11:52:02 CET] <durandal_1707> you need to complete gsoc stuff
[11:52:50 CET] <paras2052> durandal_1707: yes i know that. Student applications date is 20th march. Should i start making proposal now ?
[11:52:58 CET] <durandal_1707> yes
[11:53:05 CET] <paras2052> okay
[12:37:03 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:807d5dcde9d8: avcodec/scpr: use correct linesize for prev frame
[13:11:10 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:b0f36a0043d7: avconv: stop using setpts for input framerate forced with -r
[13:11:11 CET] <cone-378> ffmpeg 03Clément BSsch 07master:993a9a3d72af: Merge commit 'b0f36a0043d76436cc7ab8ff92ab99c94595d3c0'
[13:13:26 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:6f40181cad8a: avconv_qsv: align the surface size to 32
[13:13:27 CET] <cone-378> ffmpeg 03Clément BSsch 07master:e514309a916e: Merge commit '6f40181cad8ac04adff7bd10e1e1ab65f22bc1f0'
[13:14:11 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:8b7a9729aa16: avconv_qsv: use the actual pixel format provided by lavc
[13:14:12 CET] <cone-378> ffmpeg 03Clément BSsch 07master:15f6e5f2a956: Merge commit '8b7a9729aa162e2bbd571933f1aa40767f1ff47b'
[13:16:38 CET] <ubitux> hwaccel folks: any comment on f65285aba ?
[13:18:38 CET] <ubitux> seems to apply cleanly
[13:19:44 CET] <ubitux> after f65285aba0df7d46298abe0c945dfee05cbc6028, the doxy of sw_pix_fmt becomes invalid
[13:19:54 CET] <ubitux> i'll leave this to someone else to fix it
[13:23:20 CET] <BtbN> in what way does it become invalid?
[13:23:27 CET] <BtbN> The check it adds seems reasonable to me.
[13:23:43 CET] <nevcairiel> the doxy probably says its unused for encoding or something
[13:23:49 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:d59641abfd25: lavc: initialize AVCodecContext.sw_pix_fmt properly
[13:23:50 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:f65285aba0df: lavc: set sw_pix_fmt for hwaccel encoding
[13:23:51 CET] <cone-378> ffmpeg 03Clément BSsch 07master:8d2d81709882: Merge commit 'd59641abfd25a1007bdf4723d952887b1e3619c6'
[13:23:52 CET] <cone-378> ffmpeg 03Clément BSsch 07master:5e193daaa28d: Merge commit 'f65285aba0df7d46298abe0c945dfee05cbc6028'
[13:24:33 CET] <ubitux> BtbN: what nevcairiel said
[13:24:47 CET] <ubitux> btw, should we merge ac7bfd69678f3966e38debdb27f4bde94dc0345c now?
[13:24:57 CET] <ubitux> i'm seeing a bunch of commit related to qsv
[13:25:05 CET] <ubitux> i think we can merge it now, don't we?
[13:26:07 CET] <nevcairiel> is that the scale filter?
[13:26:13 CET] <nevcairiel> probably can
[13:27:25 CET] <ubitux> i guess i'll have to test it a little
[13:27:35 CET] <ubitux> or someone wants to take over the qsv-mostly-noop merge?
[13:42:48 CET] <ubitux> ./ffmpeg -hwaccel qsv -c:v h264_qsv ...
[13:42:51 CET] <ubitux> is this supposed to crash?
[13:42:56 CET] <ubitux> Assertion avctx->codec_id != AV_CODEC_ID_NONE failed at src/libavcodec/parser.c:144
[13:43:08 CET] Action: ubitux still trying to figure out how to make qsv work
[13:43:21 CET] <nevcairiel> thats what you are supposed to do =p
[13:43:34 CET] <nevcairiel> but no, it should not trigger that assertion usually
[13:43:46 CET] <ubitux> well, i'm trying to have it working before merging the stuff
[13:43:54 CET] <ubitux> but it looks pretty sportive already
[13:44:14 CET] <ubitux> -hwaccel qsv has no effect and -hwaccel qsv -c:v h264_qsv crashe
[13:44:16 CET] <ubitux> +s
[13:44:23 CET] <nevcairiel> havent tried to use it in ages, but jkqxz would hopefully know
[14:18:08 CET] <jkqxz> ubitux:  What is the problem?
[14:26:59 CET] <jkqxz> -hwaccel qsv indeed does nothing on its own.
[14:33:21 CET] <ubitux> jkqxz: shouldn't it warn or something?
[14:34:07 CET] <ubitux> i had 3 problems: -qsv has no effect, hwaccel qsv -c:v h264_qsv asserts, and dropping the assert leads to "HWCTX Error initializing an MFX session"
[14:34:17 CET] <ubitux> i'm not yet familiar with that stuff though
[14:35:23 CET] <jkqxz> What platform?  Has it ever worked on that machine?
[14:40:18 CET] <jkqxz> <http://sprunge.us/GQSK>
[14:40:51 CET] <jkqxz> The ffmpeg default of having lookahead enabled on the encoder messes with it somewhat, because it needs a huge output pool for that to work.  Otherwise it's fine.
[15:07:15 CET] <ubitux> jkqxz: i guess not compat, i7-4770R
[15:07:44 CET] <ubitux> anyway, if i can't test, someone else should probably do the merges
[15:12:02 CET] <jkqxz> The next seven (2196226..ad71d32), which are all qsv?
[15:13:54 CET] <ubitux> + ac7bfd696 which was skipped
[15:13:59 CET] <jkqxz> How do I do the two-part merge step?  I've only ever cherry-picked from another branch before.
[15:15:14 CET] <ubitux> you can check http://git.videolan.org/?p=ffmpeg.git;a=blob;f=doc/libav-merge.txt;hb=HEAD#l64
[15:15:40 CET] <ubitux> it's basically running tools/libav-merge-next-commit merge or tools/libav-merge-next-commit noop [optionnal-hash-ref]
[15:16:13 CET] <ubitux> (with a remote to libav repository)
[15:23:28 CET] <jkqxz> How does the source commit get into the repo on push?  Does git do something magical to make that happen?
[15:25:46 CET] <BtbN> it just pushes it there
[15:33:01 CET] <jkqxz> So there is some internal magic?  There is only one commit in the ffmpeg tree, and I don't see how it is referencing the change.
[15:34:33 CET] <jkqxz> push --dry-run does not look like it's going to do the right thing, since it only mentions the main tree.
[15:40:32 CET] <BtbN> A merge commit has two parrents.
[15:40:35 CET] <BtbN> That's the entire point of them
[15:52:15 CET] <ubitux> jkqxz: push it in a private repo and we can check
[15:56:38 CET] <jkqxz> <https://github.com/fhvwy/FFmpeg/commits/master>  Looks like yes?
[15:59:18 CET] <ubitux> jkqxz: yes it looks fine
[16:00:49 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:21962261c74a: qsv: handle the semi-packed formats in map_fourcc as well
[16:00:50 CET] <cone-378> ffmpeg 03Mark Thompson 07master:210dd7bbb2c0: Merge commit '21962261c74aed4df00ae8348a5e2d1ecb67c52d'
[16:03:51 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:807a3b30d291: lavfi: add a QSV scaling filter
[16:37:44 CET] <jkqxz> The deinterlacer doesn't work because of some difference in lavfi.  It keeps calling ff_request_frame(), which succeeds but never gives it a frame.
[16:57:24 CET] <wm4> jkqxz: is that the thing that requests 2 output frames at once?
[16:57:28 CET] <jkqxz> Making the obvious change to return the lack of error kindof works, but it doesn't actually deinterlace anything.
[16:58:09 CET] <jkqxz> Maybe?  It sits in a loop asking for input frames until it gets some.
[16:59:11 CET] <jkqxz> Also it doesn't seem to actually do anything when run with avconv.  (Looks like the interleaved frames on input are just passed through unchanged?)
[16:59:19 CET] <ubitux> btw, be careful: if you merge a bunch of commit but someone pushes something upstream, you're going to be in trouble to rebase
[16:59:31 CET] <ubitux> are you already up to the deinterlace commit?
[16:59:53 CET] <jkqxz> Yes.
[17:00:06 CET] <jkqxz> Rebasing doesn't work somehow?
[17:00:22 CET] <ubitux> yeah, with merges it's a pita
[17:00:38 CET] <ubitux> you can trick with one, but multiples you're asking for troubles
[17:02:58 CET] <ubitux> the request frame thing changed in FFmpeg
[17:03:20 CET] <ubitux> see 02aa0701a i guess
[17:03:41 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:37a9015ee84c: qsvenc: add support for p010
[17:03:42 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:d9ec3c60143b: qsvenc: take only the allocated dimensions from the frames context
[17:03:43 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:95414eb2dc63: qsv: print more complete error messages
[17:03:44 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:0956fd460681: qsvenc: do not re-execute encoding on all positive status codes
[17:03:45 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:8e07c22e508b: qsvenc: print warnings from encode/init
[17:03:46 CET] <cone-378> ffmpeg 03Mark Thompson 07master:562f386c7715: Merge commit '37a9015ee84c15fec5247ba8f6577351a25fa8d2'
[17:03:47 CET] <cone-378> ffmpeg 03Mark Thompson 07master:723a542d6c67: Merge commit 'd9ec3c60143babe1bb77c268e1d5547d15acd69b'
[17:03:48 CET] <cone-378> ffmpeg 03Mark Thompson 07master:15887a410c5a: Merge commit '95414eb2dc63a6f934275b4ed33dedd4369f2c49'
[17:03:49 CET] <cone-378> ffmpeg 03Mark Thompson 07master:80fa5a0bcc13: Merge commit '0956fd460681e8ccbdae19f135f0d3970bf95c2f'
[17:03:50 CET] <cone-378> ffmpeg 03Mark Thompson 07master:a7434ef19585: Merge commit '8e07c22e508b349d145b9f142aa3ee8b3ce1d3a4'
[17:04:15 CET] <wm4> wow merges
[17:04:29 CET] <JEEB> I think those are still cherry-pick merges?
[17:04:32 CET] <JEEB> or is that merging proper?
[17:04:38 CET] <ubitux> no it's real merges
[17:04:45 CET] <JEEB> ooh, coal
[17:04:48 CET] <jkqxz> Those were made with the magic script.
[17:05:05 CET] <ubitux> only 994 left 
[17:05:07 CET] <ubitux> @_@
[17:06:04 CET] <ubitux> jkqxz: does it work if you simply drop the request_frame?
[17:07:01 CET] <ubitux> mmh i suppose it won't
[17:07:13 CET] <jkqxz> Yeah, it gets past that problem if you just return the result.  It doesn't do any deinterlacing, though (nor does avconv).
[17:09:10 CET] <durandal_170> bunch of remaining are bitstream commits
[17:10:09 CET] <ubitux> there are hundreds to deal with before bitstream
[17:21:13 CET] <wm4> I was starting to work on some threading commits (those which keep threading enabled for hwaccels), but stopped having time
[17:21:33 CET] <wm4> partially because I was fighting with michaelni for this side data merging shit
[17:41:22 CET] <atomnuker> bofh_: ping
[17:53:00 CET] <durandal_170> can i push my stuff?
[17:58:42 CET] <jkqxz> durandal_170:  Wait a few minutes, please.
[18:04:07 CET] <bofh_> atomnuker: pong, sec let me first grab the latest code from master since it seems you added a separate mdct15 from the imdct15?
[18:04:45 CET] <cone-378> ffmpeg 03Anton Khirnov 07master:ad71d3276fef: lavfi: add a QSV deinterlacing filter
[18:04:46 CET] <cone-378> ffmpeg 03Mark Thompson 07master:b9acc7fbd9fe: Merge commit 'ad71d3276fef0ee7e791e62bbfe9c4e540047417'
[18:04:53 CET] <jkqxz> durandal_170:  Go ahead.
[18:05:27 CET] <nevcairiel> there is some "tricks" to manually rebase, i  still wonder why git doesnt just rebase the merge commit like any other, but i'm sure there is some weird reason for that
[18:05:36 CET] <jkqxz> ubitux:  That's all of this set of qsv stuff.
[18:28:27 CET] <cone-378> ffmpeg 03Paras Chadha 07master:5dab7b91adf6: avcodec: add XPM decoder and demuxer
[18:28:28 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:eda78c8b7f8a: avcodec/xpmdec: make convert function more picky about its input
[18:28:29 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:404d2b977fde: avcodec/codec_desc: add mime type for X-PixMap
[18:28:30 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:01b069c1b81f: avcodec/codec_desc: add mime type for X-BitMap
[18:28:31 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:e73325b971b3: avcodec/xpmdec: skip everything before signature
[18:29:38 CET] <cone-378> ffmpeg 03James Almer 07master:88deeb3eba00: tests/api-seek: make the crc array uint32_t
[18:29:39 CET] <cone-378> ffmpeg 03James Almer 07master:bdd07d7796e9: test/api-seek: clean up properly on failure
[18:29:41 CET] <cone-378> ffmpeg 03James Almer 07master:ff17c76e92cd: tests/api-seek: fix memory leak on realloc() failure
[18:38:57 CET] <durandal_170> i had a dream with swscale having modular pix fmts support where each pix fmt can be disabled/enabled
[18:39:09 CET] <nevcairiel> was it a nightmare
[18:39:37 CET] <DHE> dream? that's the thing where you experience hallucinations at night and then suffer amnesia about them in the morning?
[19:04:39 CET] <durandal_170> no seriously, why not?
[19:05:35 CET] <durandal_170> ubitux: i fixed silly cube bug in lut3d \Ë/
[19:14:39 CET] <cone-378> ffmpeg 03Jun Zhao 07master:bf238a6a3ca9: vf_hwupload: Add missing return value check
[19:25:07 CET] <BBB> durandal_170: I think itd be cool, but I dont think well ever quite get there :-p
[19:43:42 CET] <adeel1> durandal_170: I have made the skip_chars_comment more compact and removed a parse_video method, not sure if there's any other useless code in there: https://gist.github.com/adl1995/ff687e9748d5b3b5aad04536f766e89f
[19:47:44 CET] <durandal_170> adeel1: still calling coded_frame, not checking allocation failure, also another student already completed qualification task
[20:00:26 CET] <adeel1> durandal_170: removed coded frame usage, added checks: https://gist.github.com/adl1995/ff687e9748d5b3b5aad04536f766e89f
[20:00:46 CET] <adeel1> what does it mean if  2 students complete same qualification task?
[20:02:25 CET] <durandal_170> adeel1: you havent passed, also i helped you too much,  even for pointer usage
[20:06:13 CET] <adeel1> durandal_170: In the beginning, yes. But, I did follow through along the way. Could you please assign be a different task? This time I will solve it without any external guidance.
[20:07:01 CET] <durandal_170> adeel1: i dont get it, FITS is already taken
[20:07:31 CET] <ubitux> durandal_170: it's curious i haven't spot that one, pretty sure i tested the DOMAIN thing
[20:07:40 CET] <ubitux> tell the guy to drop those lines as a workaround
[20:07:50 CET] <ubitux> (i suppose that's fixing the ffmpeg-user issue)
[20:17:45 CET] <adeel1> durandal_170: It was mentioned on the page "each student should work on a different qualification task", isn't it un-fair that you gave us both the same task? Also, don't you think I have the motivation and the will to improve? I have been contiously coming here for almost past 2 weeks.
[20:27:05 CET] <durandal_170> adeel1: dwa/dwb in exr is even harder task
[20:27:35 CET] <durandal_170> so i didnt want to give it to unexperienced students
[20:29:28 CET] <durandal_170> there is minterpolate and dicom task
[20:35:09 CET] <adeel1> durandal_170: okay, I can choose either?
[20:35:57 CET] <durandal_170> adeel1: they have different mentors, so ask them for more info
[20:36:28 CET] <adeel1> alright
[20:45:38 CET] <adeel1> michaelni: I want to work on Minterpolate improvement's qualification task. I have briefly gone through vf_minterpolate.c Is there any specific method to which I should add improvements?
[20:46:32 CET] <durandal_170> adeel1: first you need to know what you are doing
[20:47:04 CET] <durandal_170> do you know how to use filters via ffmpeg?
[20:47:58 CET] <adeel1> durandal_170: Through the CLI commads? yes, I tried some examples.
[20:53:07 CET] <durandal_170> adeel1: basically you need to find way to reduce number of artifacts in minterpolate filter when it produces extra frames
[21:07:38 CET] <cone-378> ffmpeg 03Muhammad Faiz 07master:6c7a0876fefd: swresample/swresample: do not reset tsf on swr_alloc_set_opts
[21:19:39 CET] <michaelni> adeel1, its up to you what you improve and how
[21:27:08 CET] <bofh_> atomnuker: posting code to the mailing list in a sec.
[21:48:39 CET] <jamrial> michaelni: give others the chance to comment on the spherical patch before pushing since afaik there were some in favor of instead dropping sizeof(sidedata) from ffprobe's output
[21:49:06 CET] <nevcairiel> there is no reason we cant do both
[21:51:46 CET] <jamrial> true
[23:09:05 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:7133ab435abb: avcodec/xpmdec: improve comment for one function and also fix 2 identation issues
[23:09:06 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:9d7e71a233e3: avcodec/xpmdec: rename convert to hex_char_to_number
[23:09:07 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:fbc1f323dbda: avcodec/xpmdec: rename yet another function
[23:09:08 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:2b790b1c9e3b: avcodec/xpmdec: do not allow number of colors to be higher than allocated
[23:09:09 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:dd0090eb2112: avcodec/xpmdec: fix colors values which are different for X11
[23:09:10 CET] <cone-378> ffmpeg 03Paul B Mahol 07master:58f0bbc1eafb: avcodec/xpmdec: avoid "magic" numbers in function hex_char_to_number()
[00:00:00 CET] --- Mon Mar 13 2017