Tuesday, 2018-09-18

T4<abhishek_0> meanwhile i have updated the docker image for gitlab-ci05:07
T4<abhishek_0> https://gitlab.com/abhishek9650/sailfishos-wt88047-ci05:07
T4<abhishek_0> check the latest tag05:07
T4<abhishek_0> if anyone else wants to use it: abhishek9650/sfossdk:latest05:08
*** ChanServ sets mode: +v T405:46
T4<abhishek_0> lol why was libwifi commented out https://paste.ubuntu.com/p/4cRcJDT2Hc/05:54
PragmaticAnarchibefore i start my journey turning my discusting zl1 android experience into a lovely linux in my pocket one logic indicates i ask first: does anyone know if a ports was atemptted for the leEco le pro 3 (zl1)?06:05
PragmaticAnarchiduckduckgo and xda reported nothing, so this is my last resort.06:06
T4<abhishek_0> there was one guy on it, but never finished it06:09
T4<meierrom> I'm not aware of any such port. Since lineage OS is supporting it and you can give a try. Won't be easy tho but it never is! 👍07:34
T4<meierrom> [Edit] I'm not aware of any such port. Since lineage OS is supporting it … you can give a try. Won't be easy tho but it never is! 👍07:34
Mister_Magister@meierrom fun begins when there is no lineage port or any custom android of any kind08:06
r0kk3rzpfft, its easy lots of times08:07
Mister_Magisterr0kk3rz: then halp08:07
r0kk3rzdo you need help?08:08
Mister_Magisterye i have device tree made for twrp and need to convert it to lineage and make lineage work08:08
r0kk3rzoh nice08:08
r0kk3rzwhich device?08:09
Mister_Magisterzenfone 5z08:09
Mister_Magisteryou said its easy so if you could help please08:09
Mister_Magisterin android community nobody is gonna help you :/08:09
r0kk3rzi was referring to @meierroms comment, ive never ported lineage before08:11
r0kk3rzbut xda numptys can do it so how hard can it be08:11
abransonpiggz: I think gstreamer actually supports that colour format as part of its android support, so there may be some conversion functionality in there too08:50
abransonanother alternative is that we stop misusing one format to wrap these opaque android ones, and actually label what's coming out properly08:51
abransonbut that's quite a bit more work08:51
T4<adampigg> abranson .... i only found an implementation in vlc so far ... but ill look in gst too08:53
T4<adampigg> i piped it through the nv12 function but got color issues08:54
abransonpiggz: https://github.com/GStreamer/gst-plugins-bad/blob/master/sys/androidmedia/gstamc.c#L216808:54
abransoni think that one was supported in gst-omx too08:58
abransonah not quite the same sorry08:59
r0kk3rzhow many color formats to they need?!09:00
abransoni think they're all variants that the hw decoders could do fastest09:01
r0kk3rzyeah no doubt theres an evolution of performance tweaks there09:01
abransonoh no wait that is the same. it tags QOMX_COLOR_FormatYUV420PackedSemiPlanar64x32Tile2m8ka as GST_VIDEO_FORMAT_NV12_64Z32, which gstreamer can convert to i420 itself.09:02
T4<adampigg> abranson, so, best option is?09:02
T4<adampigg> ah09:02
abransonpiggz: you could try that. see if replacing the I420 with GST_VIDEO_FORMAT_NV12_64Z32 actually produces the correct colour09:02
T4<adampigg> so, try and use gst version?09:03
abransonpiggz: don't touch it - just copy it as with the current i420 case (that's for mtk)09:03
T4<adampigg> inside gst-droid built in conversion?09:04
abransonyeah I think so. brainstorming a bit here though09:04
T4<adampigg> dont worry, it wont happen until tonight anyway, have a think09:04
T4<adampigg> so, gst-droid handles i420 and nv12 internally if not using droidmedia.....copy the i420 case and send it to another function? or add it to the if() and handle it the same?09:07
abransonadampigg: if you properly tag the frames coming out as GST_VIDEO_FORMAT_NV12_64Z32 then you won't have to use another function. gstreamer will insert a colour converter to give the browser the i420 it wants.09:08
T4<adampigg> ok09:10
abransonso for a test, just replace everything i420 in gstdroidvdec.c with that format, and it should declare it as a source then gstreamer will do the rest09:13
T4<abhishek_0> i am not getting any of this, but when you have solution for the colour issue, ping me :P09:14
abransonhmm, the current buffer copy does cropping too though. that might be tricky09:17
abransonthat format looks crazy: "NV12 with 64x32 tiling in zigzag pattern"09:17
abransonno wonder that function was so long09:17
Mister_Magisteris there somebody that knows about android porting?09:20
T4<adampigg> abranson...yeah, i wondered/about the cropping, and assumed youd taken it into account :D09:23
abransonadampigg: well that's brainstorming for you :D09:28
T4<adampigg> abranson, feel free to brainstorm for the next 6 hrs, when i will get to try it10:18
T4<abhishek_0> :wondering: if we can automate the hal builds on gitlab-ci12:22
T4<abhishek_0> because why not, thinking of if we can take off all the load from the local machine12:29
r0kk3rzbecause generally once stuff is working the hal doesnt change much12:31
r0kk3rzwith the exception of droidmedia maybe12:32
r0kk3rzotherwise you'd need to pack all the android source and build stuff into a container12:33
T4<abhishek_0> probably not very efficient , ,, or i can just setup a runner12:36
T4<abhishek_0> or maybe not,, $$12:36
r0kk3rzgenerally my idea was just to run mic against whats in OBS12:37
r0kk3rzand set that up for all devices in OBS, so we can do bulk image updates12:38
T4<abhishek_0> we can create a gitlab org12:39
r0kk3rzyeah something like that12:39
T4<abhishek_0> i can set it up ,,, and we can add devices ,,, like lineageos does12:40
T4<abhishek_0> i have updated the docker image, builds fine12:42
r0kk3rzyou just used zypper to update?13:14
Mister_Magisterr0kk3rz: sfossdk?13:19
T4<abhishek_0> r0kk3rz yeah13:49
T4<meierrom> @Mister_Magister [is there somebody that knows about android por …], Mister_Magister: lineage OS people have likely more experience in Android porting than members of this group :)13:52
Mister_Magister@meierrom people there are more likely to help even if they dont have knowledge than anybody form android community13:53
Mister_Magistertrust me13:53
Mister_Magisterthats why im asking13:53
* vknecht hesitates to ask los questions too ;-)14:11
wolfquinnoob question again; im getting the issue; (error: in sync --fetch-submodules: revision refs/tags/android-4.4.2_r2 in platform/prebuilts/tools not found) when trying to repo sync with (—fetch-submodules), im trying to sync hybris 11.0, and the kernel/device of gtelwifi(sm-t560)im using ubuntu chroot with Platform SDK of SailFishOS. and my manifest is the following:<?xml version="1.0"14:33
wolfquin  <project path="device/samsung/gtelwifi"14:33
wolfquin    name="gtelwifi/android_device_samsung_gtelwifi"14:33
wolfquin    remote="github" />14:33
wolfquin  <project path="kernel/samsung/gtelwifi"14:33
wolfquin    name="gtelwifi/android_kernel_samsung_gtelwifi"14:33
wolfquin    remote="github" />14:33
wolfquinthis is an unofficial build of LineageOS14:34
maltry without --fetch-submodules14:36
wolfquin10.4Mfatal: Couldn't find remote ref refs/tags/cm-11.0-XNPH44S-bacon-5fa8c79c0b14:37
wolfquinshould i just ignore it?14:38
wolfquinit seems it isn't downloading the kernel nor the device, must be something wrong with my manifest?14:39
elros1try to set correct revision in your local manifest14:43
malwolfquin: define revision for the projects in local manifest, in this case cm-11.014:43
T4minlexx was removed by: minlexx14:43
wolfquintrying now, everything seems to work so far, at 24%.14:50
T4<adampigg> abranson, on way home, any more brainstorming for me ;)14:53
taixzomal: matissewifi?15:32
maltaixzo: I still haven't fixed the scaling issues15:44
abransonadampigg: only that you can probably use a videocrop element to do the cropping15:49
piggzabranson: say for the moment I ignore cropping....im in gstdroidvdec.c, where am i looking ?17:09
piggzhas anyone replced a battery in a TBJ ?17:18
T4faenil was added by: faenil17:22
*** ChanServ sets mode: +v T417:24
elros1ohyash: quite smooth sfos on A31 https://raw.githubusercontent.com/elros34/freetab_8014/master/droid/Screenshot-18-09-18-18-06-13.png18:12
piggzmal: do you know where in gstdroidvdev i should tag the video as GST_VIDEO_FORMAT_NV12_64Z32 ?18:34
piggzim thinking in gst_droidvdec_data_available18:36
malpiggz: seems like a good place, somewhere before to color conversion function probably, you probably need also something in the color conversion so it will not fail with unknown format18:42
piggzmal: yes, i have that already18:43
piggzmal: im doing it wrong somewhere ... hitting an assert18:50
piggzgst_video_frame_map_id: assertion 'info->finfo->format == meta->format' failed18:50
malpiggz: fix call of gst_buffer_add_video_meta ?18:51
malit sets format to GST_VIDEO_FORMAT_I42018:52
piggzthats what i did18:52
piggz  gst_buffer_add_video_meta (buff, GST_VIDEO_FRAME_FLAG_NONE,18:52
piggz      GST_VIDEO_FORMAT_NV12_64Z32, width, height);18:52
piggzin 2 places18:52
malwhere is the assert?18:54
piggzmal: elsewhere in gstreamer i think18:57
piggzmal: assert is here https://github.com/GStreamer/gst-plugins-base/blob/1.14/gst-libs/gst/video/video-frame.c#L9019:26
malpiggz: the droidvdec says it's caps are for i420 so maybe that is the reason the other parts think it outputs only that19:28
piggzyes, i thought something like that19:28
piggzneed abranson!19:28
r0kk3rzpiggz: stuck?19:34
malpiggz: check dec->out_state->info19:36
malin gst_droidvdec_data_available19:36
malthat is GstVideoInfo struct19:37
Mister_Magisterdr_gogeta86_: ping19:41
piggzmal: i cant change that member as its const ... i'll print it out...19:49
piggzmal: format is 2, ie GST_VIDEO_FORMAT_I42019:55
malyep, so the expected format is that19:55
piggzmal: so, im back to the original....how to convert GST_VIDEO_FORMAT_NV12_64Z32 to GST_VIDEO_FORMAT_I42020:01
r0kk3rzshuffle the bits around :) just like a rubics cube20:02
piggzr0kk3rz: yeah, shuffe the bits around in a fscking zigzag pattern :/20:04
r0kk3rzyeah, like a rubics cube20:14
r0kk3rzget it looking right first and then we can try and optimise it20:16
r0kk3rzbecause it sounds like an 10000 step process to unzig the zags20:16
piggzr0kk3rz: bollocks to it ..... LLS works fine20:26
malpiggz: it's just that droidvdec is limited, maybe lls does it via built-in gstreamer codecs and stuff20:35
r0kk3rzlls would be outputting native buffers20:35
malit probably leaves the conversion to other parts of gstreamer20:36
piggzi think it is browser that mandates i420 isnt it?20:37
r0kk3rzi dont think it converts it at all in normal case20:38
piggznow then, if i could justre-write www.videolan.org/developers/vlc/modules/codec/omxil/qcom.c20:40
piggzits certainly smaller than that huge function on stackexchange!20:40
r0kk3rzyeah it has functions20:41
r0kk3rz"nothing to be understood here, don't try" :D20:42
r0kk3rz"Second wtf: why this format at all" :D:D20:43
piggzi know, genius isnt it20:43
r0kk3rzthis lets you know you're dealing with some heavy black magic20:44
T4VeryOriginalUsername was removed by: VeryOriginalUsername20:48
piggzr0kk3rz: in that vlc code, what would pitch refer to?20:48
r0kk3rzits read out of the inputted picture_t, whatever that is20:51
piggzsome function of width and pixel size20:53
r0kk3rzyes ok, dot pitch then20:54
r0kk3rzor maybe https://en.wikipedia.org/wiki/Stride_of_an_array20:55
piggzah, stride i can work wth...20:56
malpiggz: that vlc function looks much more reasonable20:57
piggzmal: certainly20:57
* piggz struggling to concentrate, when bttf-pt2 is on!20:58
piggzso, the vlc function has a single pitch/stride .... i think we have 3 seperate strides, for y,u,v21:01
piggzmal: r0kk3rz: thats me for tonight ... i got a purple screen only21:31
piggzi tried as per https://bpaste.net/show/4b6a7d59caa7 , line 365 and  132221:32

Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!