libera/#maemo-leste/ Wednesday, 2022-10-12

mighty17[m]<buZz> "mighty17: why not maemo leste on..." <- didnt we already do that :P, with d4 rootfs that is07:52
mighty17[m]and please just use galaxy tab 2, the code names are a long list (P3100, P3110, P3113, P5100, P5110, P5113 etc)07:52
freemangordonmighty17[m]: what was the missing extension?07:52
mighty17[m]<buZz> "even has a 3G modem :P" <- no support on mainline iirc,  it is xmx626207:53
mighty17[m]intel and usb :P07:53
mighty17[m]freemangordon: for wlroots? bgra07:53
mighty17[m]gimme a sec, I'll send the patch i used07:53
freemangordonyes, wlroots. could you provide the exact name?07:53
mighty17[m]https://github.com/MightyM17/wlroots/commit/bfbe46c8d9b742d640dcf16994919d912adf32f207:54
mighty17[m]`EXT_read_format_bgra`07:54
freemangordonugh, gnome crashed :(07:55
mighty17[m]ouch :( what changes did ya do?07:56
freemangordonnothing07:56
freemangordonit happens once in a while07:56
freemangordonwhen I click somewhere07:56
mighty17[m]:(07:57
freemangordonEXT_read_format_bgra, right?07:58
mighty17[m]<freemangordon> "EXT_read_format_bgra, right?" <- yes08:19
sicelofreemangordon: you might want to keep a tab on [PATCH 1/4] ARM: dts: omap3-n900: fix LCD reset line polarity email ( https://marc.info/?l=linux-omap&m=166557193823911&w=2 )13:34
freemangordonI do :)13:35
sicelo;-)13:49
buZz:)13:56
Wizzupsicelo: oh, sweet, does that also fix the idle issues?15:44
Wizzupfreemangordon: I see a bo cache, does it help?15:46
buZzhmm, maybe i should change the three-phone-stand into some 'rack' and add a fan :P16:08
siceloWizzup: what does? the panel stuff? i don't think so. it's just about changing between HIGH & LOW for reset pin. i thought fmg should keep note because recent mail mentions VRFB, although that's in the omapfb driver, as oppossed to omapdrm16:10
WizzupIirc I tried a similar patch from sre back when I was doing off mode tests and it significantly increased power usage16:11
Wizzupbut it might have been different16:11
siceloah yes ... i recall your mail :-)16:11
siceloactually it's the same thing in the new commit, only this one puts it in dts, while sre put it in the driver.16:12
buZzi'll be probably be getting a spyder4 on saturday :) hopefully it'll work well16:14
buZzmight have to kill hildon and use fluxbox or something, perhaps16:15
WizzupI don't think the droid 4 even allows for calibration16:15
Wizzupthis is a hw feature usually16:15
buZzi think it does? i just mean the normal icc profiles16:16
buZzwell, i'll discover it saturday either way16:16
Wizzupmhm16:18
buZz> According to section "Loading ICC profiles" in https://wiki.archlinux.org/index.php/ICC_profiles, loading ICC profiles for deskop environments not providing color management can be done in X11 using XICCD.16:19
buZzbut yeah, not sure16:19
Wizzupprobably won't perform16:20
Wizzupbut who knows16:20
buZzyeah doesnt take a lot of effort, plus i still needed to calibrate my new tv here, its a bit annoying to drag windows across desk and see em change color :P16:31
buZz'new tv' from the giveawayshop, same model i already had but 39" instead of 32"16:32
buZzehhhhh lol, i now have hdmi output on d4 + mouse connected, it works16:39
buZzbut16:39
buZzthe mouse cursor is on LCD, not HDMI :D16:39
buZzamazing ^_^ nice that it does show easily that dual video output is possible in the hw16:40
buZzthis was after restarting hildon-desktop after ; xrandr --output HDMI-1 --same-as DSI-1 --mode 1280x72016:41
sicelodual video output?16:46
buZzyeah , both the LCD and the HDMI were enabled, and displaying different pixels16:51
buZzthe mousecursor was visible only on LCD , the hildon output was on HDMI16:52
Wizzupfreemangordon: I am resurrecting leste-image-sunxi builds17:15
buZzman, i want https://www.chip1stop.com/USA/en/product/detail?partId=SOCI-0000003&mpn=SC0FQAA-B-000  so i can compile arm on arm faster ;)17:15
buZzWizzup: w00t!17:15
Wizzupfreemangordon: what should we use as kernel? omap?17:16
WizzupI suppose not really17:16
buZzi have a cubieboard2 here to test17:16
buZzmaybe just mainline? iirc works pretty well17:16
WizzupbuZz: I am swimming in sunxi boards @ my amsterdam apartment17:16
WizzupbuZz: yes, but I would prefer to use our omap kernel17:16
Wizzupit's just that we use omap2plus_defconfig17:16
Wizzupwhich doesn't include most of sunxi_defconfig17:16
buZzdoesnt debian already have a kernel for allwinners?17:16
Wizzupso we can make -another- repo17:16
buZz:)17:16
Wizzupbut I was hoping to avoid it17:16
Wizzupor at least a different pkg but same repo or something... don't know17:17
buZzi'm SO CLOSE to a working devilutionx package to submit to -extras17:17
buZz:D17:17
buZz(that at least should work on pp, d4 and n900)17:17
buZzprobably also builds for x86 i guess17:17
Wizzupgreat17:18
freemangordonWizzup: bo cache shall improve scrollin performance a lot17:19
freemangordontry modest or adressbook after you install it17:19
freemangordonre sunxi - that's great, but I don;t think we'll be able to use omap kernel as is17:20
freemangordonthough...17:20
freemangordonwell, lets try17:20
freemangordonthough I doubt the same config will work17:21
freemangordonWizzup: I wonder if we can resurrect n9/50 builds17:31
Wizzupsure we can, but we need to test it17:31
freemangordonwell, we can, but does it make sense17:31
freemangordonI have n950 here17:31
freemangordonlemme check if it is still alive17:32
WizzupI have a n9, but not here17:32
Wizzupdo you want to try to show it at openfest?17:32
freemangordonn9?17:32
Wizzupn95017:33
Wizzupso for sunxi, we can take mainline kernel and just make it use sunxi_defconig much like pine64 kernel then17:33
freemangordonlets see if battery is still alive17:33
Wizzupprobably a lot of kernel work if required for it to just work17:35
freemangordonmhm17:35
freemangordonok, it seems to be charging17:36
WizzupI think email scroll is also faster btw17:54
Wizzupalso x term17:54
freemangordonyep, it should be17:54
Wizzupxterm in particular is very noticeable17:54
freemangordon10-15fps at least17:54
freemangordonthough I wonder why it looks like brightness is varying while scrolling18:01
freemangordonsome display technology weirdness?18:02
freemangordondoe not happen on n90018:02
freemangordon*does18:02
freemangordondidn;t check on pp18:02
freemangordonWizzup: also, please, whoever maintains startup scripts for n900, remove zram and use flash swap instead18:03
freemangordonwe are simply torturing the device with zram/zswap18:04
uvosthe problem iwth flash swap is we dont want to use emmc on sutch an old device for fear of breaking it18:05
uvosand we cant tell at build time if other swap space will be available18:06
uvosas to if we should just have nothing instead of zram idk18:06
freemangordonuvos: I used my old device for 10 years without having any issue with emmc being used as swap18:06
freemangordonneither did I hear anyone having issues18:06
uvosthats a nice anecdote18:06
freemangordonwtym?18:07
Wizzup18:06 < uvos> the problem iwth flash swap is we dont want to use emmc on sutch an old device for fear of breaking it18:07
Wizzupthis18:07
uvosim saying that your device survieing dosent mean anything18:07
Wizzupbut this is mostly a matter of not having the time t focus18:07
freemangordonfremantle uses it all the time18:07
uvossince how much swap will create writes depends on usage patterns18:07
uvoswe probubly should ship a script or something to enable the swap partition on emmc18:08
uvoswith an explenation ofc18:09
freemangordonwell, at least disable zswap by default18:09
uvosand maybe not have zram at all, but idk about that.18:09
freemangordon:nod:18:09
uvosdepends on how it beaves without any swap18:09
freemangordonwill behave better than with zswap in any case18:10
uvosthe kernel has the tendacy to get stuck in buffer flush loops when there is no swap18:10
uvoson pc18:10
freemangordonright after boot I already have 27648 KiB of zswap18:10
uvosfor this reason usually linux just hangs when it runs out memory with no swap18:10
uvosinstead of oom ever getting called18:10
uvosthis is an anoying issue since forever18:10
uvosfreemangordon: so reduce swappiness to 118:11
uvosand otherwise leave it be18:11
freemangordonuvos: trust me on that, I played months with zswap and whatever kernel parameters you can imagine back then18:11
freemangordonnothing helps18:12
uvosbuZz: Wizzup: wrt callibrating the d4 display18:13
uvosso the d4 has really lackluster color accuracy beacuse of the rgbw nature of the pannel, there is no way to get rid of that aside from turning the w subpixels off (wich in turn looses you resolution), its jus a tradeoff motorola made in favor of power consumption at the expense of color accuracy.18:14
freemangordonuvos: do you have idea why is scrolling so strangely looking on d4?18:15
freemangordonis that tearing or something else?18:15
uvosfreemangordon: it loses frames18:15
uvosthis is some hw bug18:15
uvosthe bionic dosent have it18:15
uvosthe d4 also has it in android18:15
freemangordonreally?18:15
freemangordonomg18:15
freemangordonlemme reboot there to see if it is the same18:16
uvosyes the transfer of the frame sometimes just fails18:16
uvosto the pannel18:16
freemangordonok, but I think this is another issue18:16
uvosthis causes microstuttering18:16
freemangordondo you have contacts imported in addressbook?18:16
uvoson d418:16
uvosthat you cant see on bonic (really obvious if you place them next to eatch other)18:16
uvosfreemangordon: yes18:16
freemangordonok, scroll there and tell me is this what you mean by "microstuttering"?18:17
uvosthis is a really poor test18:17
uvosas fps is really low18:17
freemangordonit is > 3018:17
freemangordonwith latest driver18:18
Wizzupuvos: it's much better with latest ddx18:18
uvosWizzup: no this is a different effec18:18
uvost18:18
uvosits not better18:18
uvosoh the fps18:18
freemangordonfps is better18:18
freemangordonyes18:18
uvosright but its still to slow18:18
uvosits best viewed on something that dose 6018:18
uvosso no i dont see it there18:18
uvosits not fully smooth as is18:19
uvosthere is allso tearing18:19
uvosbut thats some other issue18:19
uvosit looks like tiles sometimes dont get updated18:19
freemangordonno, it is the same on android18:19
freemangordonfints stutter there18:19
uvoswhat now18:20
uvosthe stutter?18:20
uvosyes thats on android18:20
uvosthe tearing18:20
uvosno18:20
freemangordonah, no, tearing is because we don;t support TE18:20
freemangordon]but that's clear18:20
uvosas i said the stutter is a hw bug in the pannel/dsi18:20
uvos*dss18:20
freemangordonbut it is not so obvious on android because it uses dark fonts on white background18:21
uvosim not sure what difference that makes18:22
uvosthe pannel also has pretty slow response times, this is affected by that18:22
uvosbut thats something else18:22
uvosso we agree what your seing is the hw bug18:25
uvosor are you seeing something else?18:25
freemangordonyes, I think I agree18:27
freemangordonin applications menu in android, when you scroll you can see how fonts become 'thinner' and even black at times18:28
freemangordonthe same happens on leste18:28
buZzuvos: afaiui , the ICC profile could still alter the colors in xorg, to have 'human eye' perceive better colors on d418:30
uvosbuZz: yes18:33
uvoshowever that requires the ddx to be able to hw accelerate the output frames color transformation18:33
uvosif you want resonable perf18:33
uvosi doubt omapddx can do that18:34
uvosmodesetting/glamor can18:34
uvosand even so18:34
Wizzupright18:34
uvosyou cant get good color repoduction on d4 its just impossible18:34
uvosunless you sacrifice brightness and resolution by turning off the white subpixels18:35
uvosinterestingly the bionic dose do this18:35
uvosi has better color repoduciton but way worse brightness18:35
freemangordonok, neither pp nor n900 suffer from that choppiness, so not much we can do there18:35
uvosbecause it bearly uses the white subpixxels18:35
uvosfreemangordon: bionic wont either18:35
uvosnor d318:35
freemangordonok18:35
buZzuvos: we'll see :)18:36
uvosbuZz: look at the bionic18:36
uvosbuZz: if you have one, thats the best you can do18:36
uvosimo i think it stucks more than d4 beacuse its so damn dim18:36
buZzi'll do pp and n900 aswell18:36
buZzbut dont have a bionic yet18:37
buZzits just, diablo looks so blue on d4, vs the others18:38
buZzthere must be something between current and 'dim but accurate'18:39
uvosyou have a hard time doing that without knowing how to change the pannel parameters18:39
uvossince the pannel gets rgb and transfroms that to rgbw18:39
buZzi'll use dispcalgui with calibration hw18:40
uvossince you only have 3 varaibles to work with but the pannel uses 418:40
uvosyou cant make the pannel do what you want18:40
uvoswithout changing the transformation parameters stored on the pannel18:40
uvos(and we dont know how or even if those can be changed)18:40
uvosbut sure getting it closes might be possible18:41
uvosbut see comment about the ddx18:41
uvos*closer18:41
buZzuvos: i think you could even have a .icc for seperate applications18:42
buZzat least, seems some tools do that18:42
Wizzupimage-builder always confuses me18:42
buZzso a mediaplayer could get 'best possible' but ereader just 'briiight'18:43
uvossure18:43
uvosalso mpv could work since you can give it color profiles to apply18:43
uvosbut if you want something global18:43
buZzat least i'll try a normal dispcalgui run18:43
uvosxorg needs support to accelerate the transformation18:44
buZzand see if it affects anything at all :P18:44
uvosfor it to be resonable18:44
uvosand i doubt omapddx has18:44
buZzhmhm18:44
buZzwe'll see :)18:44
uvosso btw18:45
uvossince gtk2 scrolling works a bit better now18:45
uvosam i the only one who hates the bouce back effect when it reatches the end of a scoll area18:45
uvosimo its works pretty poorly18:46
uvosalso on fremantle18:46
Wizzupit is the indicator you have reached the end18:46
uvosi know18:46
uvosbut it looks jaring18:46
uvosimo18:46
uvosmaybe its just poorly implemented18:46
uvosi dont hate it as mutch on ios18:46
Wizzupcould be18:46
Wizzupuvos: maep is much more smooth too18:47
Wizzupfreemangordon: ^18:47
Wizzupimpressively more smooth18:47
buZzuvos: i am MORE annoyed by the timepicker difference between qalendar and alarms18:48
Wizzupthat is just qt and gtk18:48
uvoswell thats just broken18:48
buZzthe alarm one is good, the qalendar one is hell18:48
uvoswut no18:48
buZzyeah18:48
uvosboth alarm and qalendar is qt18:48
Wizzupwhat I mean is one uses the gtk implementation18:48
Wizzupthe other has its own18:48
uvosits just the alarm implements a custom widget18:48
buZzthey do look the same18:48
uvosin maep i think its unfortionate that it dosent do kinectic scrolling18:49
buZzbut always sooo hard to pick the right time on applications18:49
uvoswell and mutlitouch18:49
uvosbuZz: yeah18:49
uvosbut this is a bug18:49
uvosnot something that works as intended thats just bad18:49
Wizzupuvos: well, it's a 2008 era app when there was no multitouch18:50
buZzhmhm, afaik on issuetracker18:50
WizzupI'm sure it can be added later18:50
uvoswell n900 had no mutlitouch18:50
uvosit was common else where18:50
uvosin 200818:50
uvosanyhow18:50
buZzcapacitive touchscreens are older than resistive ones anyway18:51
uvosnavit is also faster18:53
uvosaltho fast it is not18:53
uvos(but it isent on my desktop ether, lol)18:53
freemangordonWizzup: and NEON optimisations in gtk and gdk-pixbuf are yet to come19:00
uvoswith fallback path i hope19:00
uvosarmv7 devices with no neon are a thing19:01
uvosit was optional19:01
uvos(eg the atrix/ tegra2)19:01
freemangordondo we support such?19:01
uvosnot atm19:01
freemangordonwell, do we *plan* to support?19:01
uvosbut would be a bad move to force support for an optional extension19:01
uvosidk what everyone plans :P19:01
uvostegra2 was very popular and is very close to omap4 in terms of perf19:02
uvosso its not out of the question19:02
uvosim sure other chips of simmular vitange with no neon also exist19:03
buZzdoesnt tegra2 do normal opengl?19:07
uvosyes19:08
uvosbut thats not uncommon19:08
uvosmostly just sgx dose only gles :P19:08
buZzwell, mali too?19:09
uvospretty sure mali dose ogl, unles were talking ainchent gles1 stuff19:10
uvosif you can do gles2 you can do ogl1.x really19:10
uvosim also going to go out on a limb here19:12
uvosand say the difference between sgx 540 and sgx 545 is software19:12
uvos(the latter doing ogl2)19:12
buZzthere's a project i wanted to package but its opengles3 only :(19:15
buZzi wonder if packaging stuff with gl4es inside the package is possible19:16
uvossure19:16
uvosjust LD_PRELOAD it19:16
buZzsystemwide gl4es works, but i dont want everything to use it19:16
uvosimo systemwide gl4es is a bad idea19:17
uvosjust LD_PRELOAD it where required19:17
buZzcurrently, just apt install gl4es, gives it systemwide19:17
uvosi know19:17
buZzmaybe we should change that?19:17
uvosmaybe19:17
uvosbut then again19:17
uvostelling people to mess with envvars isent good expieriance either19:18
buZzmaybe19:18
uvosso it would need some kind of ui19:18
buZzi dont see 'linuxphone users' as beginners usually19:18
buZzheh yeah pff19:18
buZz:)19:18
uvoslinuxphone users are mostly crazy people19:21
buZzgood19:21
buZz:)19:21
buZzWizzup: ^^^ first tshirt txt for leste marketing19:21
norayrthis pmos podcast tells tdat there is tow open nource pvr driver: https://cast.postmarketos.org/static/audio/23-PVR-SGX540-Kupfer-UBports-installer-SoCs-Phone-Testing-2.opus19:31
norayri maybe missed that, do we alceady use it?19:31
freemangordonwhat driver?19:32
Wizzupthe podcast is wrong, there's only some wip stuff, unless we completely misunderstood the guys writeup19:33
WizzupI mean not to piss on his work, it'll be great to have19:34
norayroh19:34
norayrthey sounded very inspiring.19:34
Wizzupgreat :)19:36
norayreh i prefer to know the sour truth.19:36
Wizzupwell I might be wrong :)19:41
siceloit does sound like the person has a bit of experience in the field20:01
Wizzuphappy to be proven wrong :D20:05
bencoh:)20:29
buZzdoes anyone here have some idea about packaging gl4es inside a .deb for opengl-wanting programs?20:51
buZzi really want openscad on leste :)20:52
buZzinstead of doing systemwide21:00
norayrmaybe u have to package it with your game and put to non-standard location, so it won't be found by linux loader.21:01
buZzhmhm, maybe a run.sh that LD_PRELOAD=./gl4eswhateverneeded21:01
norayrthen start your binary as uvos suggested21:02
buZzyeah21:02
norayror maybe by using ld ribrary path21:02
WizzupbuZz: maybe make gl4es not systemwide and some way for pkgs to depend on it21:06
uvosthis ^^^21:10
uvosmaybe we should have gl4es that just installs the .so21:10
uvosgl4es-syswide that simlinks the .so to gl.so21:10
uvosand then just depend on gl4es and preload the so21:10
uvosotherwise packaging gl4es with a game would require you to build it for eatch package in that package21:11
uvoswhich is a mess21:12
norayru can rename the package (call it buzzgles or mgles) and hack the scripts that create deb files to install librarie to different, weird locations. and then more than one package may depend on it.21:16
Wizzupbuggles then please21:16
norayr(:21:16
buZzhehe22:07
buZzomnomnom22:07
buZzi'm trying to convince someone here to adopt a droid4 so he can get moonlight working22:09
buZzwhat was joergs nickname again23:23
buZzfreemangordon: meh, not even 144p plays well hon yt here23:29
buZzfreemangordon: can you give me the full useragent you spoof23:30
sicelobuZz: doc___scrutinizer23:30
buZzooo yes ty23:31
buZz:)23:31
uvosbuZz: something like mpv `youtube-dl -f 18 -g $(URL)` tends to work very well23:34
uvosyou can also try -vo xv, but mpv has some issues with xv and it provides only marginally better perf than gpu if at all.23:35
buZzright, but i wanted to record playing normal youtube.com in chromium23:38
buZzi noticed the same with -vo xv vs -vo gpu23:39
buZzalso -vo gpu has working hue/saturation/gamma23:39
buZzxv doesnt23:39
uvosi mean 240p works ok in firefox23:41
uvosdosent appear to be loosing any frames23:41
uvosbut fullscreen dosent work23:41
uvosand you cant interact with the controles23:41
uvosso its not usable as sutch23:41
buZzyeah i dont know, i think freemangordon said he has fluid fullscreen now on youtube with just spoofing old-iphone user agent and blocking adds23:43
uvosi spoof android23:43
uvosfullscreen not working is just a problem with ff/hildon23:44
uvostheres a race condition23:44
uvosof some kind23:44
buZzit might be because of something toggling compositing23:47
uvosno23:47
uvosits has something to do with focus out23:47
uvoswhen ff gets focus out it exits fullscreen23:47
uvosfor some reason ff asking for fullscreen also causes hildon to send focus out and then immidatly foucs in again23:48
uvosthis kicks ff back out of fullscreen23:48
uvosits a bug in hildon most likely23:48
uvosanyhow its not related to youtube23:49

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