Friday, 2020-12-04

riniguspiggz: thanks for photo! I guess sending notification for each 100m before approaching a turn would be spamming06:41
piggzrinigus: yes, but lets see what happens if i override that on a more capable watch,....08:59
T42<adampigg> (Photo, 1920x2560)
T42<Lillian %lastname%> Fwd from Lillian Benson:
T42<adampigg> (Photo, 1920x2560)
rinigus@adampigg: started to wonder how many watches per hand do you have there?09:30
T42<adampigg> same watch, but reimplemented those methods in the subclass which is specific to this watch to send instructions to the music app09:45
T42<寿修禁 %lastname%> (Photo, 720x547) TG飞机群发,私聊,在线,增粉;三万优质中文群全网监听服务限时特价;需要请咨询09:57
T42Yan %lastname% was added by: Yan %lastname%11:49
deathmist_piggz: yes, I've indeed been working on MSM8998 mainlining for a while :) for some reason Matrix is completely broken for me and none of my messages ever got to the IRC side of this chat16:09
deathmist_mal: what is needed for proper gmp-droid enablement? YouTube playback from browser (and perhaps other things) completely broke after I added it to my droid-configs: I can send the logs again in a moment that I sent a few weeks back16:13
deathmist_s/sent/tried to send/16:14
maldid you build correct version?16:22
maldeathmist_: ^16:24
deathmist_I should've, it's an offline 3.4 build with gmp-droid pulled from git, I can check on device16:25
deathmist_gmp-droid is on v0.1.6 and I see there's been a few revisions in the last 2 weeks, I suppose I'll upgrade gmp-droid on device and try again16:29
abransondeathmist_: stay on that version for 3.4. those changes are for esr6017:01
maldeathmist_: I mean which version of gmp-droid17:14
maldeathmist_: I use 0.1.217:15
malabranson: wasn't 0.1.6 too new for 3.4? that is for esr6017:17
deathmist_mal: that explains it, let me downgrade it instead to 0.1.217:35
deathmist_btw I suppose hybris-17.1 is near since a few days back a lot of Android 10 support work was finally merged? :o17:36
malonce I add a new disk to my build machine so I make a test build17:37
deathmist_ exists and it suggests latest commit is fine unless it's not been set to the proper one yet17:51
malyou should not use that17:52
maljus t use 0.1.217:52
deathmist_mal: am I doing this wrong? I used "rpm/dhd/helpers/ --offline --mw=hybris/mw/gmp-droid" since without "--offline" it updated the sources to 0.1.818:03
T42<elros34> maybe use 0.1.4, that was recommended version some time ago by mal :). I used to use 0.1.6 and that also worked so changes are probably not critical18:07
T42<elros34> sorry or maybe it was abranson18:09
deathmist_ok gmp-droid 0.1.4 at least built with the same method so I'll test it; I'll grab the logs again in a bit18:10
deathmist_first thing I see in any shell I open on the device is "NOTICE: Env value ignored: MOZ_GMP_PATH=/usr/lib/gmp-droid/0.1/"; wonder if that has something to do with my issues18:14
T42<elros34> it's only relevant if you start browser from shell18:15
deathmist_ok this time the video at least starts playing in the browser but it's completely "corrupted" :D here's a preview:
T42<elros34> did you use dont-use-droid-convert?18:23
deathmist_@elros34 I did! removing them didn't seem to fix it though (I still left "use-codec-supplied-height;use-codec-supplied-width" for the codecs), I'll capture journal and logcat logs after rebooting and playing back the video18:28
T42<elros34> I am afraid those quirks are not used anymore18:29
deathmist_ah, well I removed my /etc/gst-droid/gstdroidcodec.conf with the previous quirks, no dice; journal: logcat:
maldeathmist_: try backporting
malgmp-droid does not use quirks18:46
deathmist_mal: alright, it applied cleanly but sadly nothing changed playback wise, should I send updated logcat?18:55
T42森屿海巷@ %lastname% was added by: 森屿海巷@ %lastname%18:59
maldeathmist_: then try
deathmist_mal: yep, that works19:55
deathmist_seems this issue at least affect MSM8998 SoCs19:55
deathmist_I guess SW decoding is better than broken decoding ¯\_(ツ)_/¯ I'll probably finish up the other issue I've yet to fix on 3.4 and finally update the public port20:00
maldeathmist_: the decoding still happen in hw, just to color conversion is not done android library20:09
malso many typos20:11
deathmist_hmm, it somehow still manages to mostly use >=60% of the first four cores during playback of a YouTube video through the browser20:18
T42<elros34> Beware that this advice is device specific but I had to disable media.mediasource.webm.enabled so youtube start to use avc (IIRC instead vp9)  which is hardware decoded on my device20:32
maldeathmist_: do you see anything in logcat20:51
deathmist_mal: "uid=1046(mediacodec) VideoDecMsgThre identical 682 lines" and "C2DColorConvert: unknown format passed for luma alignment number" definitely keep appearing during playback, here's a snippet from starting the browser and letting a YT video play:
deathmist_(that is with applied if it makes a difference)21:01
maldeathmist_: that is normal to get that "2DColorConvert: unknown format passed for luma alignment number"21:47
maldeathmist_: was there any difference how the output looked with and without
deathmist_I'll check21:55
maldeathmist_: of course when you do not use i.e. the broken state21:56
deathmist_mal: huh, so now I tested normal 0.1.4 release without any patches again and the video didn't even start to play and I got the following logcat:
deathmist_after rebooting and letting it idle for a bit before trying it works again lol, I'll gather logs with vs without c1644bb on a clean tree now22:05
maldeathmist_: hmm, I think I saw similar on my device randomly22:07
maldeathmist_: also check visually how those look, not just logs22:07
deathmist_mal: image artifacting looked identical between the screenshots; logcat with c1644bb: without it:
maldeathmist_: what happens if set m_stride = ALIGN_SIZE (width, 4); and m_slice_height = ALIGN_SIZE (m_slice_height, 4); both22:53
deathmist_I can try22:55
deathmist_ugh, I'm having difficulties even getting the video to start playing now23:08
deathmist_mal: artifacting still looks about the same, logcat:
maldeathmist_: can you tell which video you use for testing, or do all fail the same way?23:38
T42<Katherina %lastname%> Fwd from Katherina Wood:
deathmist_mal: yeah it doesn't matter which one, any video from YouTube on the default browser does the same; I just picked a random one from the trending page because I'm lazy :p23:44
malat least we know that when not using the android conversion library it works23:50
deathmist_yeah that'll do as a workaround for now so I can get a public 3.4 build out23:51
deathmist_besides I was already using it before as well (
deathmist_for gmp-droid testing I've obviously now had that deleted and will delete it from the repo tomorrow as well23:52
malmaybe we need to add support for the quirk to gmp-droid also23:52
maldon't you still need the quirk for gst-droid?23:52
deathmist_uh guess I do, just remembered gmp-droid was for gecko only xD I should probably sleep soon23:54

Generated by 2.17.1 by Marius Gedminas - find it at!