libera/#maemo-leste/ Thursday, 2021-08-12

leltallero opened an issue: https://github.com/maemo-leste/gtk/issues/1 (Where `GtkWidgetTapAndHoldFlags` is (was) defined?)00:55
leltallero opened an issue: https://github.com/maemo-leste/bugtracker/issues/568 (Where `GtkWidgetTapAndHoldFlags` is (was) defined?)09:12
lelparazyd transferred an issue: https://github.com/maemo-leste/gtk/issues/1 (Where `GtkWidgetTapAndHoldFlags` is (was) defined?)09:12
lelMerlijnWajer created a repository: https://github.com/maemo-leste/libresource10:58
lelMerlijnWajer created a repository: https://github.com/maemo-leste/ohm-plugins-misc12:12
WizzupCan anyone with a n900 confirm that the leds from mce still work with latest image + upgraded to -devel stuff?14:02
Wizzupuvos: ping14:40
Wizzupfreemangordon: btw: https://wiki.merproject.org/wiki/Nemo/Audio/MainVolume15:05
Wizzupalso https://wiki.merproject.org/wiki/Nemo/Audio#Mainvolume15:05
Wizzupuvos: ohmd seems to react to hp plug just fine on n900, so that's good news16:07
Wizzupit also has support for fmradio, so that'll be fun if that "just works"16:07
Wizzupwith the pulseaudio alsa-config for the n900 (instead of UCM) I can also switch audio for the apps to earpiece, headphones or speakers16:07
WizzupIt looks like the volume handling is a bit different, so we'll have to decide if we use their approach (mainvolume plugin) or the old maemo method16:08
Wizzupit also nicely attempts to disable the fmradio broadcast when hp is plugged in16:14
Wizzupcool16:14
Wizzupanyone with a sailfishos phone here?16:21
WizzupI was wondering what this would yield: rpm -qf /etc/pulse/xpolicy.conf16:21
Wizzupuvos: have you seen toyed with this before? https://wiki.maemo.org/Community_SSU/Features/hildon-desktop16:37
Wizzup(some of those would be good to try on software gl too: what if all transitions are set to 0 ms, blur disabled, etc)16:38
bencohI'm not certain the user can really disable everything using transitions.ini16:50
bencohbut that would be cool16:50
Wizzupprobably not16:51
Wizzupbut we control it, so if it makes a difference we could consider some of that16:51
bencohfor software egl I guess it would16:51
bencohfor n900, I doubt it16:51
Wizzupyeah16:51
Wizzupthere is also some stuff about xdamage events wrt rotation16:51
bencohreducing animations on n900 does not necessarily speed it up16:52
bencohit's funny / silly, but if you reduce animations too much on n900/fremantle, it feels kinda jerky16:52
Wizzupmhm16:53
uvosWizzup: hmm?19:45
Wizzupuvos: what part?19:45
uvosping19:45
uvoswhat do you want19:46
WizzupI don't remember :)19:46
uvosoh19:46
uvosok19:46
Wizzupmade a lot of progress on audio front though19:46
uvosragarding the transistions ini19:46
uvos i do have the roataion animation disabled19:46
uvosbut really reduceing those timers to 0 is not so usefull19:46
Wizzupdoes it get rid of the noise upon rotation?19:46
uvosno thats the kernel19:47
Wizzupok19:47
uvosthat landed with the new omapdrm code19:47
uvosin 5.10 i think19:47
Wizzupok19:47
uvoshildon is quite inefficant anyhow and will go through the motions of the animation either way19:47
uvos(ie it will create and distroy the surfaces)19:47
uvosso reducing the timer is effectively pointless19:47
Wizzupok, we could probably write some code to skip that, but I am not sure if that helps muhc if we still 3d render everything (which I guess we will)19:48
uvossure19:48
Wizzupright19:48
Wizzupit's also not relevant, we mostly have working 3d for our purposes19:48
uvosthe main problem is that hildon is composing everything19:48
uvosfor no real reason19:48
uvosimo19:48
Wizzupdo you mean upon expose, or always19:48
uvosallways19:49
Wizzupand is that indeed what affects llvmpipe performance and such? I didn't measure19:49
uvosyes19:49
uvoswithout this the apps would be renderd by x19:49
uvosand not though llvmpipe at all19:49
Wizzupcan clutter do that?19:49
uvosas is the architecture forces all of the apps to be renderd to textures19:49
uvosand then these textures are fed to the 3d engine (or llvmpipe)19:49
uvoscan clutter do what?19:49
uvosnot compose?19:50
uvosthats not related to cluter per say19:50
WizzupI mean, wouldn't everything still be a 3d render?19:50
uvosno19:50
Wizzuphmm, so if we want to pursue a llvmpipe hildon-desktop with no eye candy we could maybe attempt that at some point19:50
Wizzup(skip animations, disable compositing)19:50
uvosok so heres the deal:19:50
uvosso right now hildon asks x to put all of the xwindows into offscreen texutres which it feeds clutter and then transforms and renders all the windows itself.19:51
uvosthis rendering is done by h-d via cluter via gl calls19:52
uvosthis is what is slow19:52
Wizzupright19:52
uvosregardless of animations19:52
uvosthis makes all the animations easy and allows the task switcher to show live tiles19:52
uvosin rality h-d could just let x render everything, and any thing it wants to show (the title bar, hildon home etc)  could be real x windows that contain clutter surfaces instead of just clutter surfaces that hildon renders19:53
uvoshildon could still display all the same animations19:53
uvosby asking x for a single buffer before the animation19:53
uvosand then performing the animation with that buffer19:54
uvosthen the rendering would never pass though gl19:54
uvosexcept durin an animation19:54
uvosthe only thing you would have to sacrifice is the live switcher tiles (they would have to be screenshots of the app in its last focused state)19:54
uvosthis would then work fine on x11 with no accell19:55
uvosas x11's own rendering is very cpu optimized it should be fast(ish) even on n90019:55
Wizzupright, but rendering seperate windows and such would be a lot of (re) work19:56
uvosright19:56
uvosit makes no sense to implement this kind of thing19:57
Wizzupyeah19:57
uvosfrom a manpower perspective19:57
uvosim just saying the way they did it is kind dumb19:57
uvosas this btw also wastes a screen sized buffer worth of ram for every app thats open19:57
uvoswhich im sure hurts bad on n90019:58
uvosthats 1.5mv19:58
uvos*mb19:58
uvosper app19:58
uvosjust down the drain19:58
uvosanyhow cool progress on the audio stuff19:59
Wizzupyeah, hopefully it'll all come together soonish20:01
WizzupI think most of the sw is now packaged and built in the CI20:01
Wizzupif I understand some of this correctly, UCM support probably isn't all that problematic20:01
konais there some way to get the virtual keyboard to send keystrokes direct to the terminal?23:16
konaor a way to send characters without a CR/LF for use in e.g. vi?23:19
konai guess the latter is what i am looking for since my pinephone doesn't seem to support BT kb.23:19
stano_hi kona23:21
konahi!23:21
stano_can we pm?23:21
konayes, thanks for asking23:21
uvoskona: you cant use the vkb in immidate mode23:29
uvoskona: however you can send text with cr/lf to xterm23:30
uvoskona: just type the text and click above the vkb to close it23:30
konaoh, hmm.23:30
konaawesome, thanks!23:31
uvosyou can also get subtaly different results (sometimes usefull) by opening the vkb with volume up23:31
uvosinstead of a tap on the input field23:31
konaoh, that is great, too23:32
uvosi do have to say that the vkb of maemo is simply bad23:33
uvosits a weakness that stems from the fact that the devices we mostly use have hardware keyboards23:33
konait used to work pretty well on the 81023:34
konabut i guess that's like, a decade ago :)23:34
uvoseven though i never used a n810 i know the vkb still sucked even then, at least in implementation, because the current behind the scenes implementation is explicitly also for the n810 (ie its the same)23:35
konaoh, hmm.23:35
konayeah, i usually used the hw kb23:36

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!