libera/#maemo-leste/ Wednesday, 2023-12-13

siceloWizzup: another PR for gps-nokia-n900 ... https://github.com/maemo-leste/gps-nokia-n900/pull/208:43
siceloplease review and merge if acceptable :-)08:43
Wizzupyeah I guess we can use this code unconditionally now10:02
Wizzupsicelo: if it wasn't clear, I merged and built11:46
siceloYay! Thanks.12:45
uvosWizzup: other than that you have to reopen whatever window with the drop down menu, sphone should have no problems with dynamicly added backends14:43
freemangordonguys, do we have any application that runs on n900 and requires 24bpp?14:49
freemangordonif not, I think it makes sense to default to 16bpp there14:50
uvosi mean given enough swap anything should run on n90014:50
uvosif you enjoy correspondence chess with your phone14:51
freemangordonwell... pushing a button and waiting a minutes is not exactly "running"14:51
freemangordonI mean "runs properly"14:52
freemangordonalso, if anyone is missing some application, it can always switch to 24bpp14:54
freemangordonor fix the application :)14:54
Wizzupuvos: ah ok15:01
Wizzupfreemangordon: fine by me @ 16bpp15:01
uvosbut making the gui follow along with sphone core would be a valid improvement15:02
uvosim fine with it too for n900, dont think the tradeoff is sane on mapphones or faster devices15:02
freemangordonagree15:02
freemangordonbut on n900 we are just torturing the device for no particular reason15:03
freemangordonon faster device I suspect we will gain almost nothing, perhaps 2% more battery life15:04
uvosat the very least composition is pretty memory intensive even at d4 scale (ie takes several mb with a couple of windows open)15:05
uvos16bpp helps with that15:05
uvosbut anyhow15:05
freemangordonright15:06
freemangordonbut I suspect we'll have issues with browsers and whatnot15:07
freemangordonI wonder if we somehow could cheat using composition15:07
freemangordonlike, h-d can use 16bpp context15:07
freemangordonthe others 24 or 3215:07
freemangordonbut honestly, no idea how to achieve that15:07
freemangordonanyway, XV support in mafw is back, runtime detected with fallback XV->GLES2->GL15:09
uvosno way of doing that without majorly hacking xorg15:09
freemangordonwith that (and 16bpp) even n900 is able to play 480p BBB with no issues in OMP15:09
uvossince h-d output buffer has to be in the depth of x and x will report its depth to all its clients15:10
freemangordonok, but in theory you can request drawable that's in different depth, no?15:11
uvossure, but you cant draw to the xwindow without conversion, theres an extension that allows x to do this but thats probubly really slow15:12
uvosyou would be converting 32bits to 16 for h-d to do its rendering and then having x convert to 32 for output15:12
uvosthats maddness15:12
freemangordonbut anyway we are doing blits through GPU15:13
uvosso i dont see how you could do this without poor performance atm15:13
freemangordonthat'd do that conversion for free (I guess)15:13
freemangordonbut yeah15:14
freemangordonnot sure what the gain would be15:14
freemangordon2MB saved from scanout buffers?15:14
freemangordonmakes no sense15:14
uvosa fun(ish) idea would be to have 2 xservers15:15
WizzupI think we can just try 16bpp on n900 and if it turns out it's not feasible we flip it back15:16
freemangordonmhm15:16
uvosand vt switch for 32bit requireing apps15:16
Wizzupwe don't -need- firefox to work anyway for example15:16
uvosfirefox works15:16
uvosit just is slow in 16bit15:16
freemangordonno, it does not15:16
uvos(presumably dose conversion)15:16
uvosfreemangordon: hmm it works here for sure (desktop)15:16
freemangordonheh15:16
freemangordonon n90015:16
uvosXephyr :1 -ac -screen 800x600x1615:17
Wizzupuvos: or xpra maybe15:17
uvosrun firefox on that15:17
uvosworks15:17
uvosits really slow however15:17
freemangordonuvos: the least issue on n900 running FF would be the screen depth15:17
uvosfreemangordon: sure15:17
uvosfreemangordon: im just saying it works in 16 bit at least on llvmpipe15:17
uvosmodesetting/llvmpipe15:18
freemangordonwe can run it on d4 without llvmpipe15:18
uvosyes none of this is practical15:18
freemangordonmhm15:18
freemangordonbut still, it is about n90015:18
uvosim just stateing 16bpp is not fundamentally broken in ff15:18
freemangordonWizzup: did you create new config with powervr.ini?15:19
uvosfirefox on Xephyr :1 -ac -screen 800x600x16 <- actually that no longer works15:20
uvosbut remoteing firefox 78 on d4 to a 16bit xephyr works15:21
uvosso they broke this since15:21
uvosnice15:21
Wizzupfreemangordon: yes15:22
Wizzupfreemangordon: do you want me to pkg it?15:22
Wizzupor try powerv.d ?15:22
freemangordonyes, maybe powervr.d is better place15:22
freemangordonbut that should be part of mapphone config, no?15:23
freemangordonno separate package15:23
freemangordonleste-config that is15:23
Wizzupleste-config-mapphone iirc15:26
Wizzupshould I try and see if it also works when it is in powervr.d ?15:26
freemangordonyes, please15:26
freemangordonbut name ti something like...dunno, ...15:27
freemangordonsgx545.ini or something15:27
freemangordonor mapphone.ini15:27
Wizzupso we still want this?15:27
Wizzup# cat hildon-desktop.ini15:27
Wizzup[hildon-desktop]15:27
WizzupWSEGL_UseHWSync=015:27
freemangordonnot sure what the format of the files in that directory is15:27
freemangordonno15:27
freemangordonthis is not used at all15:28
freemangordonWSXXX is gone long ago15:28
Wizzupmaybe that's just a stray file on my phone :)15:28
freemangordonI think h-d installs it15:28
freemangordonWizzup: so, for now I am done with multimedia stuff15:29
freemangordonOMP ha some glitches in video playback window, but I'll leave them to either me, when I want to play again with it, or someone else15:30
freemangordonissue is that background is not cleared properly in some cases15:30
freemangordonso, I'll give it a couple of days for bugs to appear and will move to stable (mafw, tracker, omp)15:31
freemangordonok?15:31
Wizzupyes15:41
WizzupI think with powervr.d/mapphone.ini it doesn't work15:46
Wizzupthe corruption is back15:46
Wizzupso I guess we need powervr.ini15:46
freemangordonmhm15:53
tmlindno ants so far, nice :)16:22
arno11freemangordon: happy you're convinced by 16bpp on n900 now :P18:56
arno11btw i tried an old img to compare with my main config and that's pretty similar in term of performances19:11
arno11excepting touch screen responsiveness19:12
arno11*(main config: 16bpp, boost freq off, light trans, new ddx, u3 card)19:24
arno11btw light transitions have few/no effect on scrolling but a huge effect to open some menus or apps19:34
Wizzupstill doesn't make sense to me19:45
arno11me too19:45
arno11maybe some animations are not working properly and slowing down some apps on startup ?19:48
arno11i think i should definitely make a video...19:50
Wizzupfreemangordon: building leste-config for pwoervr.ini now19:51
bencohsilly question, but how transitions are supposed to affect scrolling?19:53
bencohI mean, how do you test that part?19:53
bencohyou can't really scroll during transitions (?)19:53
arno11i mean there are not only transitions settings in transitions.ini19:53
bencohah, I see19:54
arno11and i think transitions.ini tweaks are just workarounds: the issue is probably elsewhere19:56
arno11but that's over my skills19:57
uvostmlind: "uvos: does modprobe modprobe cpufreq-dt-platdev fix it for you?" <-- yes it dose20:17
uvosstrange why is it not loaded on boot20:18
uvostmlind: oh btw, i installed leste on a new d4 and discovered that https://github.com/tmlind/droid4-kexecboot/tree/master/2022-12-05 now has up + down on the kbd swapped to be left and right, which isent a problem but is a bit strange, was that on purpose?20:22
Wizzupuvos: depmod -a?20:44
uvosWizzup: i mean the kernel install performs a depmod20:51
Wizzuptry it and see, when I did cross compiles I ran into that problem sometimes20:53
WizzupI'm not saying it will fix it, but it very well might20:53
freemangordonuvos: so, is it possible to override default kbd backlight brightness without recompiling mce?21:23
uvosfreemangordon: yes21:31
freemangordonhow?21:31
uvoshttps://github.com/maemo-leste/mce/blob/9af7139e61e5b3ece906b2509b7497a123abf7f1/config/mce.ini#L36621:32
freemangordonthanks21:32
freemangordonhmm: Dec 13 22:34:55 localhost alarmd[2766]: /dev/rtc0: open -> Permission denied21:38
uvosmapphone?21:41
freemangordonyes21:42
freemangordond421:42
uvosi presume alarmd wants to use the rtc to setup the rtc alam21:42
uvosso this is immaterial as the driver dose not support this feature21:42
Wizzupwould that give perm errors or something else21:42
freemangordonok, but it tries to open it every now and then21:42
freemangordondef it is perm error21:43
freemangordonit is root:root21:43
uvosopenting the rtc should be fine21:43
uvosthe ioctl should fail21:43
freemangordonopen -> Permission denied21:43
freemangordonI don;t see how user can open deviice that is root:root 60021:43
Wizzupwe can set up a udev rule for it21:44
freemangordonI think we shall21:44
uvossure, this causes the error but fixing this wont do anything, just change the error and presumably cause alarmd to waste time21:44
freemangordonit already wastes time21:44
freemangordonalso, d4 is not the only one with rtc clock21:44
freemangordonn9 has one as well21:44
freemangordonthat supprts wake-up21:44
uvosi mean hw wise the d4 supports wake-up too21:44
uvosbut sure21:45
freemangordonalso, why do you think omap4 does not support rtc wake-up?21:45
uvosjust make sure alarmd knows not to crash when its ioctl fails21:45
uvosfreemangordon: its not omap4 its cpcap21:45
uvosand it supports it fine21:45
uvosits just the driver dosent21:45
freemangordonwill fix it if it crashes21:45
freemangordonok, we may want to implement support at some point21:45
freemangordonuvos: btw, how's your idle power usage after latest tracker fixes? back to normal?21:46
uvosdunno messing with the kernel rn21:46
WizzupI think it is21:46
freemangordonWizzup: shall I open an issue for that (rtc)?21:46
freemangordonor you will remember it21:46
Wizzupissue is better21:47
freemangordonI'd rather remember it and pester you from time to time :)21:47
Wizzupalso fine21:48
freemangordonbtw it seems on arch they change owner of rtc to root:audio 66021:48
Wizzupwith the battery you made for me I get 2 days or battery life with no problem21:48
Wizzupwith the tracker bug it was way less21:48
freemangordongreat21:48
siceloi wonder how 'audio' relates to rtc0 :p21:49
Wizzupfreemangordon: btw I think the udev rule I would add would be mapphone specific21:50
Wizzupwhat I am understanding here is that you want leste in general to change /dev/rtc? perm bits21:50
freemangordonyes21:56
freemangordonbecause we have /dev/rtc on n900 as well, no?21:56
freemangordonwhy it should be mapphone specific only?21:56
Wizzupfreemangordon: shouldn't be, I just didn't know the perm problem existed everywhere22:21
uvosso anyone know how to activate the freqencies in scaling_boost_frequencies enable?23:00
uvosso the only writeable sysfs files i have are scaling_max_freq scaling_min_freq and scaling_setspeed23:03
uvoswriteing something from scaling_boost_frequencies into scaling_max_freq wold be the obvious aproch but this dosent work for freqencies not listed in scaling_available_frequencies which dosent include boost23:04
arno11uvos: when boost freqs are added in dts, a new boost folder appears in /sys/devices/system/cpu/cpufreq/23:09
arno11*boost file23:10
uvosok yeah23:10
uvosand writeing there insta hanged by device23:10
uvosso i gues it works23:10
Wizzupdepending on what you write :D23:11
uvosdo you know how to select which boost freqencies become avialble?23:11
arno11just set boost to 1 and boost freqs are automatically available23:11
arno11and then you can set max freq23:12
uvoshmm but just writeing a 1 there hanged my device23:12
uvosi never got to the later part23:12
arno11ah23:12
uvosoh well23:13
uvosi gues it works23:13
uvosthis n900 just dosent like anything over regular clocks23:13
arno11ah ok23:13
uvoslemmy try mapphone23:13
uvosseams to work23:20
arno11cool23:20
uvosbut i was only able to achive a minimal overclock 1.3ghz, vs my previous 1.4 ghz overclock via modifing the dts23:21
uvosso thats a bit wierd23:21
uvosanyhow23:21
uvoslemmy merge in .67 and do a release23:21
arno11maybe the voltage is a bit different ?23:21
uvosshould be unchanged voltage from 1.2ghz in both cases23:22
arno11ok23:22
uvosso not sure rn23:22
uvosbut anyhow23:22
uvosit works in principal23:22

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