[Ffmpeg-devel-irc] ffmpeg-devel.log.20190831

[burek]

[00:18:39 CEST] <kierank> apparently deblock is 3 pixels
[00:18:44 CEST] <kierank> so who knows where that magic number comes from
[00:39:26 CEST] <BBB> kierank: 16 is probably "bottom of mb" (instead of top, which is mb_y)?
[00:41:45 CEST] <kierank> yes but why does it do "top -= deblock_border;"
[00:41:51 CEST] <kierank> where deblock_border = (16+4)
[00:42:03 CEST] <kierank> that's 100% illegal in slice threads, mode
[00:42:22 CEST] <kierank> I need to add logic for deblock_mode=2
[00:42:26 CEST] <kierank> detect the top of the slice
[00:42:55 CEST] <kierank> I need to read up on how the edges work in ffmpeg
[01:03:22 CEST] <kierank> I don't understand why ff_thread_report_progress is called before deblock
[01:03:33 CEST] <kierank> how can you tell downstream threads the frame is ready for use if it's not deblocked
[01:04:26 CEST] <kierank> probably doesn't matter in frame thread mode
[01:04:30 CEST] <kierank> single thread decoding a frame
[01:04:41 CEST] <kierank> probably only matters in sliced threads and deblock_mode=1 mode
[01:04:49 CEST] <kierank> which is rare/impossible
[01:48:08 CEST] <kierank> ok starts to make sense
[03:34:06 CEST] <kierank> so how do I know when I'm at the end of the slice
[03:34:14 CEST] <kierank> sl->next_slice_idx is wrong often
[13:37:36 CEST] <Lynne> anyone with an amd card on linux to test vulkan?
[14:55:09 CEST] <cone-081> ffmpeg 03Paul B Mahol 07master:c79d6728a755: avfilter/vf_v360: add cubemap 1x6 layout
[16:13:48 CEST] <jkqxz> Lynne:  <http://www.jkqxz.net/~mrt/eac3ba932a16aab5465425e41a3a4567/vulkaninfo> (Mesa 19.2.0-rc1).
[16:18:27 CEST] <ubitux> jamrial: i looked into the coverage thing
[16:18:29 CEST] <ubitux> but i get "lcov: ERROR: no valid records found in tracefile coverage.info.in
[16:18:43 CEST] <ubitux> are you actually able to make it work locally?
[16:18:53 CEST] <ubitux> or it's a problem on my side?
[16:19:37 CEST] <jamrial> i just looked at http://coverage.ffmpeg.org/ and saw the last run was may 2018, so assumed your client died or got stuck
[16:19:48 CEST] <jamrial> didn't try to run lcov locally
[16:19:59 CEST] <ubitux> the compilation works fine, the fate run passes as well
[16:20:07 CEST] <ubitux> but make lcov raises that error here
[16:20:17 CEST] <ubitux> with a lot of warnings such as:
[16:20:19 CEST] <ubitux> geninfo: WARNING: cannot find an entry for vp8.c##ff77268aa0f07be80ca01b01d1b9e93e.gcov in .gcno file, skipping file!
[16:20:30 CEST] <ubitux> (probably all files, leading to the finale error i pasted)
[16:20:56 CEST] <jamrial> maybe a package update broke it
[16:21:15 CEST] <ubitux> i tried upgrading lcov but got the same issue
[16:21:39 CEST] <jamrial> :(
[16:29:42 CEST] <jkqxz> tmm1:  If you need an empty device it's probably better to make it with alloc()/init() rather than adding hacks to create().
[16:30:29 CEST] <Lynne> jkqxz: that's very basic, there's basically nothing to test there, I thought at least you'd be able to import/export fds/dmabufs like on intel
[17:29:01 CEST] <durandal_1707> ubitux: clean all environment of stalled files?
[17:29:20 CEST] <ubitux> isn't fate doing that at every run?
[17:29:37 CEST] <durandal_1707> dunno
[17:35:24 CEST] <cone-081> ffmpeg 03Michael Niedermayer 07master:a370582ba9c4: tools/target_dec_fuzzer: Init parsepkt
[18:15:04 CEST] <cone-081> ffmpeg 03Marton Balint 07master:765c56bfa903: avformat/mpegts: fix teletext PTS when selecting teletext streams only
[18:15:05 CEST] <cone-081> ffmpeg 03Marton Balint 07master:2e31774b409d: avformat/avidec: add support for recognizing HEVC fourcc when demuxing
[18:15:06 CEST] <cone-081> ffmpeg 03Marton Balint 07master:ce1fcc8cede2: avformat/utils: return pending IO error on EOF in av_read_frame()
[18:17:15 CEST] <JEEB> did we have any PGS samples in fate?
[18:17:22 CEST] <JEEB> checked under subs and it didn't seem to have
[18:17:29 CEST] <JEEB> *sub/
[18:17:58 CEST] <JEEB> I'm thinking of finally getting to making that FATE test for that one sub2video use case which I think is still borked in master
[18:18:15 CEST] <JEEB> which I didn't merge my fix for because I never had the time to get the FATE test going :P
[18:19:13 CEST] <JEEB> nothing under tests for "pgs" in git grep
[18:19:16 CEST] <JEEB> so I guess we don't
[18:19:41 CEST] <JEEB> I'll test with the DVD subtitle file first I guess
[18:20:02 CEST] <JEEB> if it has something specific with the mux of the PGS-in-MPEG-TS file, then I'll have to poke a bit more
[18:33:41 CEST] <kierank> durandal_1707: I fixed bug kinfa
[18:33:44 CEST] <kierank> Kinda
[18:41:40 CEST] <cone-081> ffmpeg 03Michael Niedermayer 07master:679f34089060: avutil/mathematics: Fix 2 overflows in av_add_stable()
[18:41:41 CEST] <cone-081> ffmpeg 03Michael Niedermayer 07master:6962fd586e1a: avcodec/vc1_block: Check for double escapes
[18:41:42 CEST] <cone-081> ffmpeg 03Michael Niedermayer 07master:9c6b4004928e: avcodec/vc1dec: Require res_sprite for wmv3images
[18:41:43 CEST] <cone-081> ffmpeg 03Michael Niedermayer 07master:fe536b6d9984: avcodec/vc1_block: Check the return code from vc1_decode_p_block()
[18:41:44 CEST] <cone-081> ffmpeg 03Michael Niedermayer 07master:5a3dee65d4fa: tools/target_dec_fuzzer: adjust pixel threshold for TRUEMOTION2, as it allows coding gigantic images on tiny input
[18:41:45 CEST] <cone-081> ffmpeg 03Michael Niedermayer 07master:cc78783ce5e8: avcodec/truemotion2: Fix multiple integer overflows in tm2_null_res_block()
[18:43:00 CEST] <cpplearner> Guys, I want to save a complete list of I-frame timestamps to seek in demand. In this case, which timestamp do I need to save, pts or dts?
[18:51:11 CEST] <cpplearner> Sorry, I'll ask in #ffmpeg. Have a nice day!
[18:54:58 CEST] <thardin> both, probably
[19:35:17 CEST] <thardin> hum, cinepak.c accepts empty frames if they come with a palette change
[19:35:22 CEST] <thardin> something the vintage decoders do not
[19:52:29 CEST] <jamrial> thardin: 9033920bec9 should be reverted once my ff_reget_buffer patch is commited
[19:55:14 CEST] <thardin> agreed
[19:55:27 CEST] <thardin> so this nonsense has been going on for at least a year
[19:56:17 CEST] <thardin> silently accepting and outputing nothing is like the worst thing one can do
[19:56:44 CEST] <thardin> I'm going to have it simply reject any file where the strips don't cover the entire frame
[20:02:38 CEST] <durandal_1707> vel0city: why dngs have darker colors compared to dcraw?
[20:30:16 CEST] <kierank> Very quiet today
[20:52:21 CEST] <thardin> I also see it counts a frame as a keyframe if the first strip is marked as intra, even if the rest of them are not
[20:52:45 CEST] <thardin> which could explain why it outputs crap on seeking in certain files
[20:53:33 CEST] <jamrial> thardin: it should unref s->frame on seeking, so it needs a flush() callback
[20:54:57 CEST] <thardin> adding a TODO about that
[22:07:25 CEST] <durandal_1707> kierank: fixed issue?
[22:08:03 CEST] <JEEB> pushing out frames not yet done
[22:08:43 CEST] <kierank> durandal_1707: kind of
[22:08:46 CEST] <kierank> bframes don't work
[22:08:53 CEST] <kierank> but you should be using bframes in lowlatency mode
[22:10:12 CEST] <durandal_1707> never worked with bframes
[22:10:21 CEST] <kierank> code was not touched for 11 years
[23:48:07 CEST] <Chagall> can I get a review on patches 14408, 14409, 14410?
[00:00:00 CEST] --- Sun Sep  1 2019