libera/#maemo-leste/ Tuesday, 2021-11-23

uvoslooks like all gtk3 applications also segfault00:02
freemangordonthis is something to do with mesa00:04
freemangordonbecause I am sure those were working00:04
freemangordonat least firefox-esr00:04
uvoshave the devuan equivalent to the buster-debug repo above handy?00:05
uvosff-esr is a different version in devuan than buster00:05
HuGoDrOcHaalguem do brasil ?00:05
HuGoDrOcHa?00:05
uvoswell nvm00:06
freemangordon0xaf250706 in dri2CreateScreen (screen=<optimized out>, priv=<optimized out>)00:06
uvosyeah it segfaults in glx00:06
freemangordonhmm, can we tell it to use gles, not glx?00:07
uvossure maybe00:07
uvosbut it should not segfault00:07
freemangordonyeah00:07
uvosit should use sw rendering instead00:07
uvosin ddk1.9  it did use gles00:07
freemangordonwell, maybe it is pvr_dri at fault00:07
uvosnot sure what changed here00:07
uvosmaybe something reports glx being avialable that dident before00:08
Wizzupright that's quite possible00:08
uvossec ill start the server with -glx00:08
freemangordonyeah, that helps00:09
uvosyeah00:12
uvosfullscreen is wonky tho00:12
uvosbut you know allready00:12
uvosok00:13
uvosSection "Module"00:13
uvos        Disable "glx"00:13
uvosEndSection00:13
uvos^^^ we can add that to xorg.conf for now00:13
uvoswe want to keep it really (to stop stuff from using glx)00:13
uvosbut ofc sefault is still a bug00:13
uvosscrolling in ff is great ;)00:14
Wizzupnow I need to try it as well00:15
Wizzupuvos: so you have just that as conf?00:15
uvosyeah00:15
freemangordonwhat?00:15
uvosfreemangordon: hmm?00:16
freemangordonwhat conf for FF TS scroll?00:16
uvosfreemangordon: there is an evvar00:16
WizzupI think the config is more like don't make firefox segfault00:16
uvossec00:16
Wizzupoh00:16
Wizzupright00:17
freemangordonah00:17
Wizzupnvm me00:17
WizzupI'll shut up ;)00:17
uvosexport MOZ_USE_XINPUT2=100:17
freemangordonok, fullscreen if fixed too, going to push00:17
uvosalso apz.allow_zooming00:20
uvos(to make zooming smartphone style)00:20
freemangordonok00:20
uvos+ something else for smartphone style selection00:21
uvosbut cant find it right now in my about:config00:21
uvosWizzup: yeah no kidding about maep00:24
uvosthats insane improvement00:24
uvosfreemangordon: really incredible work :)00:24
freemangordon:)00:26
freemangordonand it will get better with DRI2 doing double-buffering00:26
freemangordonfor some reason omap driver assumes this to be tripple-buffering00:27
freemangordonbut anyway00:27
freemangordonit needs to be fixed00:27
uvosfreemangordon: there is also something up with suspending composing00:27
freemangordonuvos: compositing is not accelerated still00:27
uvosfreemangordon: right00:27
freemangordonalso XV00:27
uvosfreemangordon: but i mean suspending it00:27
uvosdosent work00:27
uvosin hildon00:27
freemangordonuvos: sec to push fixed version00:28
freemangordonand then we'll search for bugs00:28
uvosthe display just stops updateing if you ctrl-shift-n toggle it00:28
freemangordonyes, I know00:28
uvosok00:28
freemangordonthat's the fix I am about to push :)00:28
uvosok ok00:28
uvos:)00:28
uvossicelo: btw now that randr rotation working on n900 is really imminant00:30
uvossicelo: it would be great if you fixed the compatbile or what the probing problem was with the new st accell driver + the n900 accell00:31
uvosi dimmly remember this as just being a missing compatible00:31
Wizzupfreemangordon: building the src with fix00:37
freemangordoninstalling ATM for a final test before asking you to rebuild DDX :)00:38
freemangordonah, well00:38
Wizzuphehe well it's -experimental00:39
freemangordonyeah00:39
freemangordonbut still00:39
freemangordonyeah, that change fixed it00:42
WizzupI still see some weird corruption in surf I think, but it's not a lot01:16
Wizzupseems to be in the same place and it 'follows' the surf page around01:16
Wizzupso it's in the 2d content I'd say rather than display flickering, so that's not the same issue01:32
Wizzupthis may be incidental but ofono seems better with 5.1502:19
freemangordonWizzup: yeah, there seems to be some corruption from time to time, but it is not flickering07:37
freemangordonasctually there *is* some flickering when 2d scrolling, but I hope that will disappear when we enable tripple-buffering07:41
freemangordonok, we *cannot* use dri3, because of the rotation, not until there is a way to request TILER BOs when using GBM07:50
parazydWizzup: I imported libllvm8 to experimental, will try the upgrade now.10:17
parazydWizzup: Let me know about pvrsrvinit, as now you put pvrsrvctl in /usr/bin, and the initscript does not use that.10:18
Wizzupfreemangordon: yeah, it seems to be in the pixmaps themselves10:46
Wizzupof surf in this case10:46
Wizzupparazyd: does it not?10:46
uvosbtw with /etc/init.d/powervr being in default atm10:46
Wizzupparazyd:10:47
Wizzup    ebegin "Starting PowerVR"10:47
Wizzup    /usr/bin/pvrsrvinit10:47
uvos(i had to compile my own pvrsrvinit)10:47
uvosas parazyd mentions10:47
parazydWizzup: Yeah there's no such binary10:47
Wizzupidgi?10:47
parazydWhat does your `dpkg -S pvrsrvinit` say?10:47
uvosparazyd: its this https://github.com/IMbackK/pvr-omap4/blob/master/pvrsrvinit.c10:47
Wizzuppvrsrvinit is not a package10:47
Wizzupit's in sgx-ddk-um-tools10:47
parazydWizzup: It's not10:47
uvosWizzup: your pacakge contains pvrsrvctrl10:47
uvosnot jit10:47
uvosinit10:47
uvoswe dont need ctrl at all10:47
Wizzupwell I was going to check but mce turned off the device for battery guard even though it was at 40-50% and charging10:48
parazyduvos: Thanks10:48
Wizzupso you'll have o wait a bit longer10:48
uvoshehe shade10:49
uvoswierd10:49
Wizzupuvos: pvrsrvctrl --start --no-module is not what we need you say?10:49
uvosno10:49
Wizzupthis literally worked for me yesterday10:49
parazyduvos: That should just work with gcc, no extra libs?10:49
uvosthat works10:49
uvosWizzup: but its a big binary blob we just dont need10:49
uvosparazyd: -ldl10:49
parazyduvos: ok10:49
Wizzupok but hang on10:50
Wizzup1. it works now with the stuff I packaged10:50
Wizzup2. you want an alternative10:50
Wizzupthat's fine10:50
uvosno the init script calls pvrsrvinit10:50
uvosthat binary isent in ddk1.1710:50
Wizzupbut /usr/bin/pvrsrvctl exists and the init scripts call the right pkg afaik10:50
uvosit was in 1.910:50
WizzupI purged everything 1.9 related and mine boots to accelerated h-d10:50
uvosas it stands right now the package provides pvrsrvctrl but the init script tries to call pvrsrvinit10:51
uvosi thats how it is on my d4 atm10:51
Wizzupthen why does it work for me10:51
Wizzupweird10:51
Wizzupmaybe it's fmg's init code10:51
uvosparazyd: also make sure its in sysinit10:51
Wizzupuvos: yeah my droid says 53% charging, I really can't understand why it did reboot10:51
parazyduvos: mhm10:51
Wizzupparazyd: should I not fix the package then?10:52
parazydI'm on it10:52
uvosWizzup: so there is the problem that the adc output is quite noisy10:52
Wizzupall we needed to change was 3 chars it seems like :P10:52
uvosWizzup: upower might signal that the voltage is to low to early10:53
uvosbut i have never seen it10:53
Wizzupparazyd: are you adding another binary?10:53
uvosreally hard to debug to if thats the reason :\10:53
parazydWizzup: We don't need pvrsrvctl10:53
parazydInstead we replace it with https://github.com/IMbackK/pvr-omap4/blob/master/pvrsrvinit.c10:53
Wizzupuvos: wtf: Nov 23 10:47:48 localhost mce[3354]: Requesting shutdown from powerkey.c: generic_powerkey_handler(); action: 210:54
WizzupI'm really quite sure I didn't do that lol10:54
uvoshmm10:54
Wizzupparazyd: well this works, but if you want to try something else, ok10:54
Wizzupit's already packaged and we'd just have to change the three chars10:54
Wizzupbut yeah ok, we'd have to compile it though10:54
WizzupI'm fine either way10:54
uvosWizzup: could be a hw glitch even (some moisture)10:55
Wizzupparazyd: so the alternative is to reply 'init' with 'ctl' in the init script10:55
Wizzupss/reply/replace/10:56
parazydThe point is to replace a binary blob with 20 lines of C10:56
Wizzupwhat about the other ~3-4MB of blob :P10:56
Wizzupbut ok, whatever10:56
parazydNothing10:56
parazydWe liberate what we can10:56
Wizzupthe same code is also in the DDX fwiw10:57
uvosyeah but that dosent work10:57
uvosits to late10:57
Wizzupso why does my droid4 init11:05
Wizzuplol11:05
WizzupI don't get it11:05
WizzupI don't have /usr/bin/pvrsrvinit11:05
Wizzupmaybe the X code does work for me?11:06
uvosWizzup: so the code in the ddx "works" but its not sufficant because 1. there is a race with h-d, h-d might sometimes end up on llvmpipe 2. it dosent do anything for other drm applications we might want before x starts, like a bootlogo or charge-mode11:07
Wizzupok, so I haven't seen the race yet11:07
Wizzupoh yeah there was the chargemode thing11:07
uvosparticularly we need pvr to be init'ed before charging-sdl if we want charging-sdl to work on acellerated drm and not directfb11:07
Wizzupit doesn't use acceleration atm though right?11:08
uvosits on directfb so n11:08
uvosbut it works fine on ddk1.17 drm i tested it before there11:08
uvosso its just a matter of flipping it over11:08
uvosthen its accellerated and can turn of the display11:08
uvosso we def want that11:08
Wizzupok11:09
Wizzupso the next thing is either n900 with uImage from 5.15 or u-boot mainline11:10
Wizzuphttps://maedevu.maemo.org/images/n900/tools/ u-boot-2020.12-pali.bin this one is mainline u-boot afaik11:12
Wizzupas in IIRC it should just work11:12
WizzupPali: iirc all your n900 work is in v2021.10 - right?11:17
Wizzupprobably not in u-boot v2021.0711:17
Paliyes, they should work11:17
Wizzupboth? ok11:17
Paliyes11:17
parazydWizzup: We should package it like this: https://github.com/maemo-leste/pine64-uboot11:22
freemangordonuvos: I don't see a way h-d to start without X :)11:40
Wizzupparazyd: there are various places where it can be flashed, with attached kernel and stuff, so it's not easy11:40
Wizzupparazyd: I'm just going to make mine work for now11:40
freemangordonbut I agree that it is racy on startup because could have no pvr modules loaded11:41
Wizzupparazyd: I'm going to test 2021.10 now11:43
parazydSure11:43
parazydI still don't have a clean upgrade path11:43
Wizzup:)11:43
Wizzupplease keep at it :D11:43
Wizzupalso try to have cloudgps or other extras installed11:43
WizzupI had to force remove cloudgps since it depended on some libgles sgx thing11:44
parazydxserver-xorg-video-omap doesn't want to upgrade11:44
Wizzup(maybe we just rebuild it)11:44
parazydThere's still some of these somewhere: https://github.com/maemo-leste/pvr-omap4/blob/master/debian/control11:44
parazydBut I can't find where11:45
Wizzupok 2021.10 boots for me and it boots leste as well11:46
uvosi mean it might make some sense to stop recomending people attach kernels, surely we can just have people move whatever fremantle kernel they want to use somewhere else?11:49
uvosthe other option is to have the install script for uboot dd the partition to an image, extact the kernel if any and reattach it11:49
Wizzupwe might need attached dtb on n900?12:08
WizzupI guess I should check12:11
uvosgood question, yeah we do rn12:11
parazyduser@devuan-droid4:~$ glxgears12:12
parazydLoadLib: Couldn't load libpvr_dri_support.so: libpvr_dri_support.so: cannot open shared object file: No such file or directory12:12
parazydWizzup: ^12:12
parazydSeen this?12:12
uvosparazyd: GLXgears?12:13
Wizzupparazyd: why are you trying glx?12:13
Wizzupglx is not supported and also it's not gles12:13
Wizzupes2gears_x11 is what you want I think12:13
parazydah sry, I wanted to try es2gears indeed12:13
uvosit should work on llvmpipe in theroy12:13
Wizzupbut, h-d is a much better test12:13
Wizzupuvos: yes but we don't want that :)12:14
uvosbut its broken atm (we allready know)12:14
parazydbut the same12:14
parazyduser@devuan-droid4:~$ es2gears_x1112:14
parazydLoadLib: Couldn't load libpvr_dri_support.so: libpvr_dri_support.so: cannot open shared object file: No such file or directory12:14
parazydlibEGL warning: DRI2: failed to create dri screen12:14
Wizzupso did you reboot to hildon-desktop and new ddx?12:14
parazydh-d is slow, so I'd say swrast12:14
Wizzupwhat is the state of your system atm12:14
parazydWizzup: Obviously12:14
parazydI have slow h-d running12:14
Wizzupwasn't clear to me12:14
Wizzupok, first thing to check is /tmp/Xorg.0.log12:14
Wizzupsearch for any errors12:14
uvosalso check you have the envvar12:14
Wizzupit's in the package12:15
uvosok12:15
WizzupI added it last night, it should be there12:15
Wizzupbut we'll find out soon enough from Xorg.0.log12:15
parazydhttp://sprunge.us/oYJaL712:15
parazydI have the envvar12:15
uvoslooks fine12:15
parazyduser@devuan-droid4:/tmp$ env | grep pvr12:15
parazydMESA_LOADER_DRIVER_OVERRIDE=pvr12:15
uvosso libpvr_dri_support.so exists right?12:15
uvos /usr/lib/arm-linux-gnueabihf/libpvr_dri_support.so.1.17.494895712:16
uvos+ link12:16
parazydso.1 and so.1.17.4948957 exist12:16
parazydBut no .so itself12:16
Wizzupinstall -dev maybe12:17
Wizzupmaybe that's it12:18
Wizzupsgx-ddk-um-dev12:18
Wizzup# dpkg -S /usr/lib/arm-linux-gnueabihf/libpvr_dri_support.so12:18
Wizzupsgx-ddk-um-ti443x: /usr/lib/arm-linux-gnueabihf/libpvr_dri_support.so12:18
Wizzuphmm12:18
Wizzupdid you change anything? let me apt upgrade12:18
parazydNothing related to the libs12:19
parazydWe just don't have those links12:19
Wizzupyeah it's gone12:19
WizzupI just upgraded and it's gone12:19
Wizzupso you did somethin that broke it12:20
Wizzup# dpkg -S /usr/lib/arm-linux-gnueabihf/libpvr_dri_support.so12:20
Wizzupdpkg-query: no path found matching pattern /usr/lib/arm-linux-gnueabihf/libpvr_dri_support.so12:20
parazydok, that's probably an apt thing then12:20
Wizzuphuh?12:20
Wizzupis "pvrsrvinit: Use version for symbol loading.12:21
Wizzup" also related to this?12:21
parazydno12:21
parazydWell, yes12:21
parazydBecause the .so weren't installed12:21
WizzupI am pretty sure they are on my device12:21
parazydJust give me a minute12:21
Wizzupmaybe you adding a Makefile broke it12:21
Wizzupsure, just being helpful here :p12:22
Wizzupas in they were there in ec28dc199f5666885c5756704dc67a6874c917ab12:22
uvosWizzup is in slightly beligerent mood :P12:23
Wizzupnah it's fine to liberate more stuff I'm just pointing out it used to work12:24
WizzupI don't see what caused the problem either from the git history, so it's some weird debian build thing I'm sure12:25
Wizzupprobably dh install does something else now12:25
Wizzupmeanwhile I almost have a droid4 kernel pkg with n900 dtb and stuff in it12:25
Wizzupwill try that out in a bit, hope I can pass the dtb using u-boot this time12:26
WizzupPali: do you boot with appended dtb still with zimage, or do you pass it separately for n900?12:28
PaliI have tested only 2.6.28 kernel which do not use DTB12:28
Wizzuphm, ok, ideally for leste we would use a separate dtb, not attached12:29
PaliU-Boot should support both options without issues12:29
Wizzupok12:29
Wizzupusually I think u-boot needs an addr to load those at, I'll try to figure it out12:30
Palion other ARM boards I'm using U-Boot with separate DTB and it is working fine12:30
Paliyes, you need to specify address where is DTB loaded in memory when calling U-Boot bootz command12:30
Palialso now U-Boot for n900 supports bootz command, so there is no need to generate uImage from zImage12:31
Paliand booting 2.6.28 kernel (stock zImage) with bootz command in n900 qemu is part of u-boot automated tests12:32
Wizzupgreat12:32
Pali... so this ensures that upstream u-boot does not break this booting12:32
Wizzupdo you know what addresses are sane?12:32
PaliIIRC there is kernel documentation for it12:33
PaliI will try to find it12:33
Wizzupappreciate it, thanks12:33
Palihttps://www.kernel.org/doc/html/latest/arm/booting.html12:34
PaliSetup the device tree --> A safe location is just above the 128MiB boundary from start of RAM.12:34
PaliCalling the kernel image --> The zImage may also be placed in system RAM and called there. The kernel should be placed in the first 128MiB of RAM.12:35
Wizzupok12:36
PaliLoad initramfs --> A safe location is just above the device tree blob which itself will be loaded just above the 128MiB boundary from the start of RAM as recommended above.12:36
PaliAnyway, there is standard U-Boot variable ${fdt_addr_r} which should specify address where DTB should be loaded12:37
PaliI think that n900 does not set this variable in include/configs/nokia_rx51.h (during compile time)12:38
PaliSo it would be a nice to define this variable and send patch12:38
PaliAnd ${fdtfile} variable should contain name of (default) DTB file12:39
PaliSimilarly ${kernel_addr_r} is address for zImage12:40
Wizzupok, I'll try to figure it out and see if I can send a patch12:40
Palin900 uses custom variables ($kernaddr, $initrdaddr, $scriptaddr), so aliases for standard variables could be useful12:41
Wizzupoh, so the addrs are already there :)12:42
parazydWizzup: Ok, when we convert it from native to quilt, it works.13:00
Wizzuplmao what13:00
parazydhttps://github.com/maemo-leste/sgx-ddk-um/commit/dc4ff73d0e4f83a66af6f4716672be0f31fc8baa13:00
parazydgbp on a native package cleans "foo.so" files13:00
Wizzupjfc13:00
parazydOn a quilt package it doesn't.13:00
Wizzupwhy does it not do that for the ones I did?13:00
parazydI don't even wanna understand it.13:00
uvosfreemangordon: so x hangs in ospoll_wait until OMAPDRI2SwapComplete13:00
Wizzupbecause the ones I made worked13:00
uvosfreemangordon: not sure why yet13:00
parazydWizzup: Probably because there was no Makefile.13:00
uvosbut that seams wrong13:01
parazydWizzup: It also works only with `dpkg-buildpackage`. It's just gbp that's broken13:01
Wizzupblergh13:01
Wizzupparazyd: what packages do you install to try to uprade to the new stuff?13:01
uvoslol debian packaging13:01
WizzupI want to try on my n90013:01
parazydWizzup: N900 is not ready yet.13:02
parazydBut you could try on a bionic or something13:02
parazydWizzup: This is the least steps I could do to perform the upgrade: http://sprunge.us/tASSBg13:02
freemangordonuvos: hmm, this sounds ok to me13:04
Wizzupparazyd: what do you mean 'n900 is not ready yet' ?13:05
Wizzupdo you mean dep wise?13:05
freemangordonit should do nothing until page flip occurs13:05
parazydWizzup: hildon-meta mostly13:05
freemangordonthe problem to me is that client does nothing as well int the meantime, instead of rendering13:05
Wizzupwell, I can ignore that for now as we need someone to test kernel and packages on n90013:05
uvosfreemangordon: right because x i hanging i suspect13:05
Wizzupand I don't have a bionic with me13:05
uvosie client waits on some x call13:06
freemangordonah13:06
freemangordonyeah, right13:06
freemangordonmikes sense13:06
freemangordon*makes13:06
parazydWizzup: So just try to install the t343x libs and upgrade the ddk13:06
uvosim not _sure_ this is whats happening but it looks suspicous13:06
Wizzupk13:06
parazydThat's enough, along with upgrading mesa13:06
uvosill try on bionic sec13:09
parazydHold on for 5 more minutes so the latest libs are in the repo13:11
parazyduvos: http://sprunge.us/8hERem13:13
parazydAfter adding experimental, this worked13:13
parazyduvos: Also I recommend having latest devel updated to avoid other package noise13:13
uvosparazyd: ok will do13:13
uvosbut have to charge the thing first13:14
parazyd:)13:14
Wizzupcan hildon-meta depend on libegl1?13:21
Wizzupthen maybe apt dist-upgrade could work?13:21
parazydPerhaps the ddk should?13:29
parazydhttps://github.com/maemo-leste/hildon-meta/blob/maemo/beowulf-experimental/debian/control#L13913:30
WizzupI think mesa provides it13:30
Wizzupthe ddk does not provide this anymore afaik?13:30
uvosyes13:31
uvosits mesa13:31
uvosnothing should depend on any of the ddk libs at all anymore13:32
Wizzupthen they will not get pulled in ;)13:32
uvosexcept mesa itself and stuff that wants to use pvr2d13:32
Wizzupbut yeah13:32
uvosthe device meta pacakge should pull it13:32
Wizzupit's more about making the upgrade smooth/working13:32
Wizzupright13:32
uvosnothing else should13:32
uvosexcept video-omap (our only pvr2d user)13:32
parazydYeah so I'm saying13:34
parazydSince hildon-meta-droid4 has xserver-xorg-video-omap13:34
parazydShould xserver-xorg-video-omap then have the sgx-ddk-um deps?13:34
uvosi gues13:34
uvoslong term it would be better if we split hildon-device specific stuff and just device specific stuff13:35
parazydok, I'm not sure where to put libegl1 then13:35
parazydPerhaps hildon-meta-$device indeed13:35
uvosso hildon-desktop itself needs libegl13:36
uvosso dose any other gles user13:36
uvosthey should pull mesas libegl13:36
freemangordonwait, clutter chould already depend on that13:36
parazydIt doesn't have a libegl1 dep13:36
uvosthen idealy mesa should pull sgx-ddk-um (but only on omap)13:36
parazydOnly libgles2-mesa13:36
freemangordonand that pulls the other stuff, no?13:37
uvosthats wierd pacakging on debains part13:37
uvostho13:37
uvossince you cant use libgles without egl13:37
parazydSo maybe we should be more explicit with clutter deps?13:37
freemangordonif it doesn't, then debian packaging is broken13:37
freemangordon*mesa debian packaging13:37
uvosparazyd: sure having clutter depend is a workaround13:38
uvosbut really this is upstreams fault if its as descibed above13:38
uvoshow dose it work on pp rn?13:39
parazydah wait13:39
uvossomhow egl must be installed there13:39
parazydI think if we rebuild clutter13:39
parazydWe might be in luck13:39
parazydBut we'd have to build it in experimental to test13:39
parazydI'll try that13:40
Wizzup13:34 < parazyd> Should xserver-xorg-video-omap then have the sgx-ddk-um deps?13:49
Wizzupdoesn't it have that already?13:49
Wizzupit depends on -dev for build and it should depend on -libs at runtime13:49
Wizzupif that's not the case it needs to be fixed13:49
Wizzupuvos: why should ddx not pull sgx-ddk-um? it needs the libraries13:50
uvos?13:51
uvosit should13:51
freemangordonEXA needsw libs13:51
uvos[13:32] <uvos> nothing should depend on any of the ddk libs at all anymore13:51
uvos[13:32] <uvos> except video-omap (our only pvr2d user)13:51
freemangordon:)13:51
Wizzupagreed, ok13:51
parazydThe issue is it can't pull them in directly13:51
parazydSince we have omap3 and omap413:51
parazydSo we need to resolve that someplace else (hildon-meta)13:52
uvoswhy13:52
uvosthe package is in n900 and in droid4 section13:52
uvosit will pull what its on13:52
parazydWizzup made it use a virtual13:53
uvosok13:53
parazydhttps://github.com/maemo-leste/sgx-ddk-um/blob/maemo/beowulf-experimental/debian/control#L1413:53
Wizzupright, yes, the pkg specific -meta should pick the exact version it wants13:53
Wizzupwhere version = omap version13:53
parazydhttps://github.com/maemo-leste/xf86-video-omap/blob/maemo/beowulf-experimental/debian/control#L2813:53
uvosim not sure why you want to do it like tis13:53
uvosinstead of using the sections13:53
uvosbut if you do then sure13:53
uvosmeta then needs to pull it13:53
Wizzupwell we would have to introduce new sections13:53
Wizzupand that is not an easy upgrade path13:54
uvosWizzup: why?13:54
Wizzupunless you want to use n900 for omap3 and droid4 for omap413:54
Wizzupuvos: because people would have to change sources.list13:54
uvosno not that13:54
uvosok sure we might have some problem in the future if we add a non-mapphone non-n900 omap device13:55
uvosreally the current sections arnt so great anyways maybe we sould break at some point13:55
uvosneeds to be well thought out tho13:55
parazydI agree we could work on this a bit13:55
uvosi mean bionic is droid4 bionic13:55
uvosrealy there sould be omap4 mapphone bionic or something like that13:56
Wizzupuvos: like xt610 :)13:56
uvosheh13:56
uvosyeah13:56
Wizzupso I went for the virtual like that because we had no easy other path, but yes we need to work on the components13:57
uvosyeah sure makes sense13:57
uvosjust makes the packaging a bit ugly rn13:57
uvosas you have discoverd :)13:57
WizzupI thought the -meta-$device could just pull the right libs13:57
Wizzupdoes that not work?13:57
uvosit dose13:58
uvosbut some package needing some other package but not directly depending on it is a bit ugly13:58
Wizzupyeah13:58
uvosyou might want to install hildon without going for the full meta13:58
Wizzupwell that will break a lot more currently13:58
parazydWe can have a big break, but then we need to clean up everything at once.13:59
parazydtbh it sounds better than patching solutions14:00
parazyd(And really it's not a break, just changing sources.list and dist upgrade)14:00
uvosyeah proubly fine if a short blogpost acompanies it to tell people what they need to do14:01
parazydExactly14:01
Wizzupwhy not do it with the buster+114:02
parazydThat'd work14:04
parazydThat's the time when we rebuild everything anyway :D14:04
WizzupI prefer that since it's obvious they should probably read the news update or flash a new image, which is better than folks not knowing what's up and upgrading and things breaking14:04
parazydFor this, I'd like that we design some doc/spec on what the repo layouts should look like next14:05
parazydAnd do some prior investigation to know if it will work (and scale)14:05
Wizzupyes we can make an issue for it14:05
Wizzupwith chimaere milestone14:05
Wizzup(sp)14:05
Wizzuplol14:05
Wizzupok, will try n900 momentarily14:05
Wizzupfreemangordon: you said portrait doesn't work yet right?14:06
sicelohe said so, yes14:10
Wizzupsicelo: thanks for refreshing my memory :p14:14
parazydSorry I dunno how to get a clean upgrade14:29
Wizzupwhat is the problem that you run into that doesn't work with dist-upgrade14:30
parazydThings don't want to upgrade unless I manually install libegl114:30
parazydAnd even that implies that hildon-meta is removed14:31
Wizzupwhat is the verbose tree / error?14:31
parazydbbl I have to work on other things14:31
parazydWizzup: I advise to try it with a new image14:31
parazydI outlined the steps to you and uvos earlier14:32
parazyd(See the sprunge)14:32
Wizzupyes, I wanted to offer help not pick up the task ;)14:32
parazydI don't know how to solve it14:32
parazydSo somebody does have to pick it up14:32
crabhi again, what is the current kernel version i should be running on n900 assuming im up to date?14:36
crabis it > 5.1.21 ?14:37
Wizzupno, it is 5.1.21 until we finish the 5.15 upgrade path14:38
WizzupI am going to try to do that today on my n90014:38
crabaah cool.14:38
crabthanks14:38
crabi was just wondering if the reason my gui doesnt work is maybe because im installed on an sd card14:39
craband kernel wasnt getting updated somehow, but it seems like it is in that case. :)14:39
Wizzupyour gui (still) doesn't work?14:40
Wizzupapt-get -s -o Debug::pkgProblemResolver=yes upgrade14:44
Wizzupthis seems very helpful14:44
crabaah thanks14:51
crabalso ill have to remember to sort out my uboot settings when that new kernel does get upgraded14:51
crabDependencies are not satisfied for hildon-connectivity-wlan:armhf < 1.4+2m7 -> 1.5+2m7 @ii umU Ib >14:53
crabhmm.14:53
Wizzupany context other than that line?14:57
crabKeeping package hildon-connectivity-wlan:armhf14:58
crabCalculating upgrade... Done14:58
crabThe following packages have been kept back:14:58
crab  hildon-connectivity-wlan14:58
Wizzupdoes 'dist-upgrade' help?14:59
Wizzupapt dist-upgrade14:59
crabnah14:59
crabthat says shes all good :)15:00
crabshall i try dist upgrade with apt-get and the debug ProblemResolver?15:00
Wizzuphuh, that's wierd, upgrade complains and dist-upgrade says there is nothing to do?15:01
crabheh15:01
crabsame outcome15:01
crabyeah looks like it15:01
crabanyway dont let me distract you from more important work15:02
Wizzupanyway, your UI is broken?15:02
crabmaybe it'll all come out in the wash when the next raft of updates are released,15:02
craband if not ill just do a fresh install15:02
crabyeah15:02
crabthe text part of the boot process is fine15:02
crabthen it all goes black screen on me15:02
crabbut fortunately i can ssh in15:02
Wizzupdoes /var/log/Xorg.0.log say anything useufl15:03
Wizzuperr15:03
Wizzupnot that15:03
Wizzup/tmp/Xorg.0.log15:03
crabthere are a few errors about missing font paths15:04
crabbut it looks reasonably ok15:04
crabi seem to remember something about "ofono"15:04
Wizzupwait, you have -devel enabled?15:05
crabsome errors during the boot process about that15:05
crabaah yeah i prob. do :)15:05
crabhow can i check?15:05
crabis that a sources.list thing?15:06
crabi cant see it there15:06
WizzupI wonder how ofono got installed15:06
crabim not convinced it is15:07
crabhmmm15:08
crabsome bits of it appear to be though15:08
crabspecifically libgofono and ofono-scripts15:09
Wizzupwe *really* need a good page on how to provide info about failures15:09
WizzupI don't know what to suggest atm15:09
crabdont worry :)15:10
crabthanks for help15:10
craband good luck with the real work.15:10
uvosWizzup: crab: i mean provide tail -n 2000 /var/log/daemon.log is a good start15:15
uvoswould be cool to have icd2 working in the emergency runlevel for this15:16
crabok ill pastebin that up if its really eating you, but i will give the rest of the channel a chance to stop me so we can save you from yourself15:16
craboh sorry, i didnt pay attention to who was typing...15:16
crab:P15:16
uvoscontrary to popular belief this chat is not simply Wizzups different personalitys chating with eatch other15:17
crabwatch it! im in here under two irc clients as discussed! ;)15:17
crabdaemon.log is empty. :|15:17
uvosreally15:17
uvosthats strange15:17
crabpossibly because i forwarded to my syslog server :P15:18
Wizzupuvos: yes but I think we should split up daemon.log, I have syslog rules to do it per daemon15:18
Wizzupbut yeah only on my d4 atm15:18
crabneither of you worry about it.15:19
uvosWizzup: sure thats fine15:19
crabi got a weird feeling it'll sort itself a few apt's down the line15:19
craband if it doesnt its probably time to clean the machine up anyway15:19
crabits got maemo, a strange gui-less debian install,15:19
crabmaemo leste15:19
craband possibly some other guff on here15:20
uvosWizzup: but thats not as conviniant if some user who knows little about linux just saies: black screen, dosent work or?15:20
craband none of them satisfy me like maemo-leste *even without the gui* ;)15:20
uvosi mean leste with no gui is just devuan no :P15:20
uvosbut yeah15:20
crab:)15:20
crabit has a more up to date kernel than the debian thing on here15:21
craband more up to date userland15:21
crabso its all good15:21
Wizzupuvos: right, we also need a set of instructions15:21
crabalso currently ranked 42 on multirpg15:22
crab:)15:22
Wizzupuvos: parazyd: hm looks like the ddx somehow pulls in ti434x15:23
WizzupI have ti343x installed nad it wants to remove it15:23
Wizzupthat sucks...15:23
Wizzupis that because it resolves it to ti434x at build time or something15:24
sicelosounds like some package has wrong dep(s) then :-)15:34
uvosrr:7 https://pkgmaster.devuan.org/merged beowulf Release15:34
uvos  Certificate verification failed: The certificate is NOT trusted. The certificate chain uses expired certificate.  Could not handshake: Error in the certificate verification. [IP: 152.228.204.144 443]15:34
uvoswhats wrong here15:34
uvosthe oldstable change?15:34
uvos(upgrading a old bionic to latest devel)15:34
uvosah was my fault15:37
Wizzuptime?15:42
uvosyes15:43
uvosi saw 2021 in date and thoght that ment ntp worked15:43
uvosbut i forgot fake-hwclock exists15:43
freemangordonWizzup: hmm, I think I know what happens15:52
freemangordonit was build against ti434x I guess15:52
freemangordonso, I think ddx should have dependency to ti434x removed during build and a manual dependency to ddk-sgx-um virtual package15:54
freemangordonor ddk-sgx-um-libs or whatever the package is called15:54
Wizzuphow would do we that in the ci16:00
Wizzupfreemangordon: btw, how do you boot your 5.15 n900 kernel image?16:00
Wizzupany specific boot args?16:00
WizzupI have the serial here but no lab power supply, so I can't check serial for another two weeks :D16:01
bencohnah, who needs a lab psu anyway ... just rip open a usb cable and add a voltage divider with two resistors :*16:04
bencoh(mostly kidding ... but it would probably work)16:04
Wizzupmaybe the kernel I built just doesn't work for the n90016:07
uvosparazyd style upgrade works fine on bionic16:11
uvosand the new ddk setup works fine on bionic in general16:11
Wizzupgreat16:14
Wizzupweird that if hildon-meta-droid4 depends on libegl1 (with specific ersion) it still does not get installed16:15
freemangordonWizzup: I used uImage16:21
freemangordonWizzup: why CI? debian packaging should be fixed to exclude ti434x frompackage dependencies16:21
freemangordonand in debian/control a manual dependency to virtual package shall be added16:22
freemangordonor just shlibs:Depends can be removed, but that's too drastic IMO16:22
freemangordonlemme see if I can fix the packaging16:23
Wizzupfreemangordon: isn't that what I did?16:23
Wizzupfreemangordon: ok thanks16:23
freemangordonno idea, sorry, did you do something I missed?16:23
Wizzupfreemangordon: ok, I can't get the zImage to boot on the n90016:23
Wizzupfreemangordon: no, please just take a look16:23
freemangordontry uImage16:23
Wizzupyeah the ci doesn't build a uimage16:23
Wizzupbut I'll make one16:23
freemangordonyeah16:23
freemangordondo you want me to provide mkimage command16:24
freemangordonI mean - the paramters16:24
WizzupI think it is this mkimage -A arm -O linux -T kernel -C none -a 80008000 -e 80008000 -d zImage uImage16:24
Wizzupbut that's just what's in my history :)16:24
Wizzuplet me check n9xx-linux16:24
freemangordonmkimage -A arm -O linux -T kernel -C none -a 80008000 -e 80008000 -n zImage -d zImage uImage16:24
freemangordon-n can be different/missing ofc16:25
Wizzupyeah16:25
freemangordonWizzup: what is the virtual package name?16:31
freemangordonfound it sgx-ddk-um-libs16:33
Wizzupfreemangordon: right16:33
Wizzupand they both 'provide' it16:33
freemangordonah, it is already a dependency16:34
* freemangordon did some mogic, lets see if it is going to work16:34
Wizzupcool16:35
freemangordon*magic* even :)16:35
Wizzupbrb, I'll try the uImage when I get back16:35
freemangordonDepends: libc6 (>= 2.4), libdrm-omap1 (>= 2.4.33), libdrm2 (>= 2.4.36), xorg-video-abi-24, xserver-xorg-core (>= 2:1.18.99.901), sgx-ddk-um-libs16:47
freemangordon:)16:47
freemangordonWizzup: https://github.com/maemo-leste/xf86-video-omap/commit/6a1eb3bf227ff69326895d11fdcf44ccb6dbca7c16:49
freemangordonplease rebuild when you have time16:49
Wizzuprunning17:00
freemangordonWizzup: is it fixed now?17:13
Wizzuplet me see17:14
Wizzupoh right17:14
WizzupI need to fix the uImage17:14
Wizzup(since it's powered off atm)17:15
mighty17[m]freemangordon:do you mind me packaging your mesa in pmOS?17:20
Wizzupfreemangordon: hm seems to just reset after a bit (kernel)17:20
freemangordonmighty17[m]: not sure I understand the question17:21
freemangordonwhy shall I mind?17:21
freemangordonit is not 'mine' really17:21
mighty17[m]asking for permissions is always better :D17:21
freemangordonok, you have my permission if you (and pmos guys) promise to fix the bug there if any :p17:22
mighty17[m]well exporting =pvr is a not a bug, its a feature17:23
bencoh:]17:23
freemangordonmighty17[m]: also, 'my' mesa was a bit older, Wizzup rebased on some recent version17:23
mighty17[m]we've been using 20.3.2 since ages :P17:23
freemangordon21.2.5+pvr1-1+2m717:24
freemangordonmost-probably has PP fixes/improvements17:24
freemangordonbut up to you ofc17:24
Wizzupfreemangordon: when you build n900 kernel, what config do you use?17:26
WizzupI wonder if perhaps some defconfig or built in command line change is the problem here17:26
freemangordonthe same as for d417:26
freemangordonomap2plus-defconfig17:26
Wizzupblergh17:26
Wizzupok17:26
Wizzupand you also do appended dtb I assume17:27
freemangordonyes17:27
Wizzupwell with 5.1 my n900 boots to a llvmpipe desktop at least17:28
Wizzupso it's not completely fuba17:28
Wizzupfubar*17:28
Wizzupfreemangordon: can you send me your uimage + mods (assuming the mods are different)?17:31
WizzupI took the droid4 img and did 'cat zImage omap3-n900.dtb > ../zImage' and then a mkimage, and it just doesn't boot17:31
mighty17[m]<freemangordon> "most-probably has PP fixes/..." <- PP = pinephone?17:32
freemangordonyes17:32
freemangordonWizzup: "mods"?17:33
freemangordonmodules?17:33
Wizzupmodules17:33
freemangordonsec17:33
Wizzupalt you can try the one we build in CI, but I suspect that's more work for you17:33
freemangordonjust lemme verify it still boots before sending to you17:34
Wizzupok17:35
freemangordonhmm, maybe I shall provide you the .config files as well17:38
freemangordonI am not 100% sure I didnt; change anything17:38
Wizzupif it boots I won't need te config files, they are usually built in17:38
freemangordonok17:38
Wizzupbtw, yes, looks like the ddx fix worked17:40
Wizzup(wrt deps)17:40
freemangordongreat17:40
Wizzupfreemangordon: do you use 'run sdboot' to boot the kernel?17:44
freemangordonyes17:44
Wizzuptrying to find what I am doing wrong, not finding it :)17:45
freemangordonbtw, this kernel *does not* boot to h-d17:45
freemangordonbut boots to console17:45
Wizzupfor me it resets after like 10-15s17:45
freemangordonah, ok17:45
Wizzupand I see nothing on the screen17:45
freemangordonok, boots here, sec to provide archive17:45
Wizzupbtw, if it does not boot to h-d, is it an older commit than the droid4-linux branch we have then?17:46
freemangordonhttp://46.249.74.23/leste/n900/n900.tar.gz17:47
freemangordonno, it is on leste/droid4-pending-pvr-omapdrm-v5.1517:47
freemangordonnot saying it is kernel fault17:47
freemangordonfor sure it is missing nokia modules17:48
freemangordonand maybe ofono hangs because of that17:48
freemangordonwild guessing, didn;t investigate what happens17:48
Wizzupso it's on this: https://github.com/maemo-leste/droid4-linux/commit/32fd293562d705670dfb3522e19e533e98729a9e17:48
freemangordonyes17:48
Wizzupty I wil try this17:48
freemangordonlast commit is 32fd293562d705670dfb3522e19e533e98729a9e17:49
Wizzupcool17:49
freemangordonbut maybe I have some config changes17:49
freemangordonwifi related for sure17:49
freemangordonI don;t think wl1251 is enabled by default17:50
Wizzupstill the one we have in the CI doesn't even get to a screen, probably not even to openrc17:51
Wizzuplet me try now..17:51
freemangordonhmm17:51
freemangordon"Nov 23 18:44:20 localhost /etc/init.d/mce[2268]: ERROR: mce failed to start"17:51
freemangordonthat's why it doesn;t boot17:52
Wizzuplol, your kernel also resets for me17:53
Wizzupwhat u-boot do you have?17:53
freemangordonhmm17:53
freemangordonlemme check17:53
WizzupI built one today, 2021.1017:53
freemangordonsome old one17:53
freemangordonbut not fremantle17:53
Wizzupprobably https://maedevu.maemo.org/images/n900/tools/u-boot-2013.04-2.bin17:53
freemangordonlike, from the times I played with uboot17:53
freemangordonno, it is few months old17:54
freemangordonhave to reboot to verify17:54
freemangordonsec17:54
uvosmce --verbose --verbose?17:54
uvoswrt it not starting17:54
Wizzupwith old u-boot it also resets with no info :(17:55
freemangordonuvos: still, have to reboot17:55
freemangordonalso, this device hasn;t been updated for ages17:55
freemangordonWizzup: :(17:55
freemangordonthat's bad17:56
Wizzupyeah, blind debugging is no fun17:56
WizzupI have the serial right here, but no the lab psu17:56
uvosthe n900's hard to acess serial port is a constant problem :\17:56
freemangordonI know (blind debugging) :)17:56
Wizzupuvos: yes we had an idea about making some serials but you didn't like the design right :P17:56
uvoswell you dont need to me to rubber stamp your designs do you :P17:57
freemangordonWizzup: my uboot is built on 31 oct 202017:57
WizzupI think I'll take a break and see if I can think of a method to debug this later17:57
freemangordonbut I don;t think that's relevant17:58
Wizzupright, same problem with our 'standard' 2013 uboot17:58
freemangordonsomehitn broken on uSD card?17:58
Wizzupvery much doubt it18:00
freemangordon"/lib/rc/sh/openrc-run.sh: 1: export: directory:: bad variable name"18:00
freemangordonyeah, SW on this device is old :)18:00
WizzupI'll be home on 7 december, then I can try it with psu18:00
Wizzupexactly two weeks from now ;)18:01
freemangordonnice18:01
Wizzupmaybe it's some watchdog kicking before the module is loaded ?18:01
freemangordonnew rule of thumb - never go abroad without lab psu :p18:01
WizzupI had one, I left it in sofia!18:01
Wizzuplol18:01
freemangordonah, I see18:02
freemangordon"1 upgraded, 0 newly installed, 0 to remove and 663 not upgraded."18:02
Wizzuplol18:04
freemangordonyeah, this will take time18:05
Wizzupfreemangordon: I wonder if it is the watchdog being '=m'18:31
WizzupCONFIG_OMAP_WATCHDOG=m18:31
WizzupCONFIG_TWL4030_WATCHDOG=m18:31
freemangordonWizzup: maybe18:44
freemangordonif you boot from a slow uSD that could be a problem I guess18:44
Wizzupn9xx-linux $ grep WATCH ./arch/arm/configs/n900_defconfig18:59
WizzupCONFIG_WATCHDOG=y18:59
WizzupCONFIG_OMAP_WATCHDOG=y18:59
WizzupCONFIG_TWL4030_WATCHDOG=y18:59
freemangordonhmm19:01
freemangordonwell, try it19:01
sicelowl1251 should be enabled by default for n900. i committed that a while ago19:02
Wizzupfreemangordon: yeah will do in a bit19:12
freemangordonWizzup: who provides mesa.sh?19:23
freemangordonfound it19:24
freemangordonWizzup: we need nokia-modem enabled too19:26
Wizzupthat is for your problem, right?19:41
WizzupI'll enable it a bit later once I have it booting19:41
Wizzup(your problem = not booting to h-d)19:41
uvostmlind: btw 5.15 pwrdm is unhappy on bionic19:47
uvospwrdm state mismatch(l4per_pwrdm) 3 != 1 repeats fairly often19:48
uvosrelated to the cpcap regulator swap?19:48
freemangordonWizzup: not sure why h-d is not booting19:54
freemangordonwhy xsession needs ofono?20:05
Wizzuppinentry probably20:05
uvosi think we added it for sphone20:05
Wizzupfreemangordon: wait are you seeing X starting, but hildon not?20:05
uvosbut sphone dosent need to be started after ofono anymore20:05
freemangordonWizzup: yes20:05
Wizzupmaybe see if some pin entry thing is running20:06
freemangordonbecause I have no ofono installed ;)20:06
freemangordonand because of that, xsession fails to start20:06
uvosjust remove it20:06
uvosits not nessecary anymore20:06
freemangordonbut why it is there in the first place? I upgraded20:06
uvosi think parazyd added it for sphone20:07
uvosas i said20:07
freemangordonok, please remove it20:07
uvosi dont have commit access20:07
uvosso parazyd ^^^20:07
Wizzupso this could be all the ui update problems maybe as well20:08
uvosyeah if its dependes20:08
uvosinstead of merly after ofono20:08
uvosin the initscript20:09
uvosthen yeah20:09
uvosthats a bug20:09
uvoswhat pacakge is the init scipt in question in?20:10
uvosWizzup: its you accutally20:12
uvoshttps://github.com/maemo-leste/maemo-system-services/commit/5b164f9a8dc0bf34abb4120348c1aa0c69a88c8820:12
uvosapearantly not for sphone either20:12
uvosScripts that get installed to xsession (like connui-conndlgs) require20:13
uvosofono to be running, or the entire xsession will block indefinitely20:13
uvos(on startup-pin-query in this case).20:13
uvoswe need to do soemthing about that ^^^^20:13
uvosofono cant be mandatory20:13
WizzupI think it was not on purpose that this went to table20:13
Wizzupwell later on it will have to be20:13
Wizzupbut not now20:13
uvosok20:13
uvosno20:13
uvosofono cant be mandatory20:13
Wizzupit should be depending on the installed packages, but let's not go into that now please20:14
uvosok20:14
uvosanyhow please remove this commit20:14
uvosWizzup: a clean ish solution for now20:16
uvosis for that init script to not need ofono but be after ofono20:16
Wizzupwon't that do the same?20:17
uvosno20:17
uvosand whatever xsession script blocks should check pidof ofono20:17
Wizzupfreemangordon: are you sure this is causing your problem btw?20:17
uvosand then exit immitaly20:17
uvosif ofono is not running20:17
Wizzupi.e. can you remove ofono from the /etc/init.d/xsession20:17
freemangordonyeah, already did20:17
freemangordonnow trying to boot20:17
Wizzupand that's the problem?20:17
Wizzupok20:18
freemangordonno idea, still trying to boot20:18
freemangordonhmm, actually device got powered off a couple of times, exactly as you explained20:18
freemangordonon the 3th try it booted fine20:18
freemangordonyep, h-d is up20:19
uvossweet random reset timeing bug :P20:19
freemangordonyeah, maybe the issue is with WDs being modules20:21
Wizzuplet me try wd now20:27
siceloi've never understood why retries make a difference with the booting :-/20:37
Wizzupif it's timing related20:37
uvosalso second boot is warm20:37
uvos(ie registers are not in reset state nesscarly)20:37
Wizzupfreemangordon: how long before you see something on console btw?20:38
Wizzupit's not resetting now at least with me compiling watchdog in20:38
siceloit's not a warm/cold boot issue. sometimes cold boots, then warm won't boot20:38
siceloanyway, i guess serial is the only way to really find out20:39
uvosWizzup: maybe the problem is that it fails to init the mmc card20:39
uvosthat would explain the reset20:39
uvoscant find the modules20:39
uvosand the no reset now20:39
Wizzupit doesn't reset now with watchdog compiled in20:39
uvosbut still no boot20:39
Wizzupyes maybe20:39
Wizzupbut why would that be the case for me and not for fmg?20:39
uvossdcards take different amounts of time to come up20:39
uvosfmgs might come up faster20:39
uvosor smth20:39
Wizzupcan't have nice things, ever :D20:40
Wizzupwait it boots now I think20:41
uvosmaybe build in the dss modules20:41
uvos /omapdrm20:41
uvosthen it might light the display20:41
uvoshaving a fully static kernel around is very usefull for debuging this kind of issue gennerally20:42
Wizzupyes it booted to 5.1520:42
uvosgreat20:42
Wizzuplooks like my X is trying to use fbdev20:43
uvosstill have -video-sgx installed?20:43
WizzupI think I removed it20:43
Wizzuppurging it...20:44
Wizzuphm, same error20:45
uvosxorg.log?20:45
* Wizzup removes xserver-xorg-video-fbdev20:46
Wizzuphm, maybe not20:46
Wizzupuvos: same as before basically, no mention of omap20:46
Wizzuphm:20:48
Wizzupunrelated but:20:48
Wizzup# /usr/bin/pvrsrvinit20:48
WizzupPVR:(Error): SrvInit: PVRSRVInitSrvConnect failed (129) [0, ]20:48
Wizzupfailed to initialize server20:48
Wizzup(it might have already initialised)20:48
Wizzupyeah it did I think:20:49
Wizzup[  481.334197] PVR_K: UM DDK-(4948957) and KM DDK-(4948957) match. [ OK ]20:49
Wizzupfreemangordon: do you have xorg conf on n900?20:49
WizzupIt tried to load everything but omap it seems20:50
freemangordonWizzup: yes, I have20:53
freemangordonit is there from when I played with omap driver20:54
freemangordonI guess20:54
Wizzupfreemangordon: care to share the config?20:55
freemangordonWizzup: https://pastebin.com/0f2JWWEz20:56
freemangordonbut i guess if you copy it from d4 it will work20:57
Wizzupwe have none at d420:57
Wizzup[  1032.943] (EE) AIGLX error: dlopen of /usr/lib/arm-linux-gnueabihf/dri/omap_dri.so failed (/usr/lib/arm-linux-gnueabihf/dri/omap_dri.so: cannot open shared object file: No such file or directory)20:57
Wizzupmesa eh20:57
Wizzupweird, even with the env var it doesn't work20:58
Wizzuptime to compare versions20:58
freemangordonhow's that (none on d4)?20:59
freemangordonhow does it know it shoul dload omap and not modesetting?20:59
WizzupI don't know, maybe I removed modesetting?21:00
freemangordonI am almost sure d4 has omap.conf21:03
rulehthere is /usr/share/X11/xorg.conf/99-omap.conf21:03
rulehxorg.conf.d21:04
Wizzupruleh: oh, yeah, I forget about these new paths... why do we have config in /usr/share again :-)21:04
rulehI think it is supposed to beĀ  something like vendor supplied stuff goes to usr/share and admin configured stuff into etc or so21:05
Wizzupfreemangordon: so it doesn't segfault for you as soon as you touch the screen ? :P21:06
Wizzupruleh: right, yeah21:07
freemangordonno21:08
WizzupI'll install debug symbols21:08
Wizzupslowly getting there at least21:08
Wizzupfreemangordon: https://dpaste.com/2PVNJCCMC21:10
Wizzupmaybe I have old libdrm_omap21:10
tmlindWizzup: have you checked the kernel defconfig has been flipped over to 1.17? it defaults to 1.14 i think21:10
Wizzupno, same version21:10
Wizzuptmlind: it's the same kernel as the droid4-linux one, just with watchdog built in21:11
tmlindok21:11
Wizzupfreemangordon: hm I think maybe the init doesn't happen via init script21:11
Wizzupfreemangordon: ah: [  390.615112] cma: cma_alloc: reserved: alloc failed, req-size: 375 pages, ret: -1221:12
tmlinduvos: that pwrdm mismatch is a bit of a mystery, probably some glitch between the cpuidle and prm driver, harmless afaik21:12
freemangordonWizzup: oh, now I remember, I increased CAM to 32 MiB :)21:13
freemangordon*CMA21:13
Wizzuphehe21:13
Wizzupok21:13
Wizzuplet me do that21:13
Wizzupany reason to make it even higher?21:13
freemangordonWizzup: I don;t think we shall increase it that much21:15
Wizzupok21:15
Wizzuprebuilding kernel21:15
freemangordonI have a kernel fix in mind, unfortunately Tomi refused to accept the idea21:15
uvostmlind: ok, dont see it on d421:15
WizzupI think 32M is fine (for now)21:15
uvostmlind: seams to apear all the time on my bionic21:16
uvostmlind: dosent have any obvious negative effect yeah21:16
freemangordontmlind: is there any negative effect of increasing CMA on n900?21:17
Wizzupuvos: ok questions are being asked internally about gitorious21:25
uvosthank you :)21:25
tmlindfreemangordon: well how much does increasing help? n900 runs out of memory easily..21:27
freemangordonwell, seems I didn't ask my question correctly :). Lemme rephrase - what should be the recommended value? given that we need at least 3 800x480x4 buffers for Xorg only21:28
tmlindnot sure, sounds like some experiments are needed to find the usable minimum size21:29
freemangordonyeah. well, I'd rather write that fix21:30
tmlindsounds like no help increasing it beyond the necessary minimum?21:30
Wizzupuvos: they might hand me all the gitorious data and task me with putting it online, we'll see21:31
freemangordonbut, I think we need contiguous memory for VRFB, so maybe that patch is not that helpful21:31
Wizzupfun...21:31
freemangordon*will not be that helpful21:31
freemangordonyeah: "...must be allocated as a contiguous memory segment."21:34
tmlindmaybe play with cma=size[MG] option a bit?21:34
freemangordontmlind: well, no hurry, I am just thinking a bit21:35
freemangordonwe need at lest 6MiB just for the framebuffers21:36
freemangordonbut for others, I still think it makes sense to not use CMA21:36
freemangordonnot sure what was needed by IVA21:36
uvosThis is a hack which is required until: http://lists.x.org/archives/xorg-devel/2013-February/035449.html is merged21:36
uvosheh ok mesa21:36
freemangordonuvos: hmm?21:38
Wizzupalso need to check my powervr.ini -- we don't have that in any of our new pkgs atm I think21:38
freemangordonwe don;t need it21:38
freemangordonanyway, I am afk21:39
freemangordonttyl21:39
Wizzupfreemangordon: wait21:39
uvosfreemangordon: so21:39
Wizzupfreemangordon: it works :)21:39
uvosglx segfaults21:39
Wizzupjust wanted to share that21:39
uvosbecause dri2CreateScreen blindly assumes the extension  __DRI2_CONFIG_QUERY exists21:39
* tmlind zzz21:40
Wizzupsicelo: around?21:41
freemangordonuvos: ah21:44
freemangordonuvos: ok, will implement that21:44
freemangordonWizzup: cool!21:45
Wizzupfreemangordon: great work21:45
Wizzupuvos: I think what we should do is integrate gl4es to do glx for us perhaps21:45
siceloWizzup: yes.21:45
Wizzupsicelo: do you have n900 pm stuff around>?21:46
WizzupI'd like to try it on 5.1521:46
Wizzuplike, to make it enter off mode, like a droid4-pm script21:46
uvosyeah thats a good idea maybe xorg not ideling the display ever was all there was too it21:46
uvoswrt it not idelin21:47
sicelono. i do have tmlind's patch though. had forward ported it to the last kernel i tried (5.12/5.13)21:47
uvosg21:47
Wizzupsicelo: right that's for reading the blockers21:47
Wizzupbut first we need to run the script and measure21:47
WizzupI had an adapted version at some point21:47
Wizzupbut I probably nuked it21:47
sicelothere's the one in wiki. i was using that21:47
sicelon900 wiki, i.e.21:47
WizzupI don't think so21:48
sicelommm, let me also check in my uSD card. i remember using a script similar to the d4 one, but can't find it now21:48
Wizzupwe also need to update idle_uarts (just change 4 to a 3 iirc)21:49
Wizzupwe also don't mount debugfs by default21:50
sicelohehe, and then my card won't work in laptop :-/21:53
Wizzupanyway this is fine, I can port the blocker patch and work on a pm script21:53
Wizzupbasically everything that blocks idle is loaded atm :P21:53
Wizzupsicelo: rebuilding with deeper idle patch, will let you know how far I get22:01
sicelogreat. :-)22:05
sicelowhat patch is that?22:05
* calebtheythem[m] uploaded an image: (74KiB) < https://libera.ems.host/_matrix/media/r0/download/connolly.tech/xcEvGlcqFCwBKoZGQRJsZnfC/IMG_20211123_210549.jpg >22:06
siceloseems my card is toast. now as soon as i put it on laptop, i get  'mmcblk0: mmc0:aaaa SB16G 14.8 GiB (ro)' - that (ro) at the end :-(22:06
calebtheythem[m]well i gave it a go (hopefully image sends), maemo boots on the OnePlus 6 with a hacked up pinephone rootfs22:06
calebtheythem[m]touch seems to work - i was getting haptics feedback when touching the screen initially but it doesnt seem to do anything22:07
sicelohehe, looks interesting!22:07
calebtheythem[m]ah it is just rotated22:08
calebtheythem[m]do you guys have rndis gadget support? OR how do you usually debug?22:09
Wizzupsicelo: this https://github.com/maemo-leste/n9xx-linux/blob/maemo/beowulf-devel/debian/patches/0001_deeper_idle.patch22:10
Wizzupcalebtheythem[m]: yeah usbnet usually22:11
Wizzupcalebtheythem[m]: we have a daemon that also loads it automatically on plug, but it uses a whitelist...22:11
Wizzupcalebtheythem[m]: really cool btw!22:11
WizzupI like the glow that comes from it22:11
uvostklock looks really bad @hdpi22:12
Wizzupyes22:12
Wizzup:D22:12
calebtheythem[m]it's only 1080p XD22:12
Wizzupcalebtheythem[m]: n900 is 800x48022:12
siceloah ok. that's the one i also had (patch)22:13
Wizzupdroid4 is a bit higher22:13
Wizzupsicelo: right it just applied22:13
uvos960x54022:13
calebtheythem[m]Wizzup: what's the daemon? I can do some tweaking22:13
calebtheythem[m]I use some custom busybox ramdisk which brings up rndis, it mostly works as long as maemo doesn't tear down the interface22:14
calebtheythem[m]interfaces*22:14
Wizzupcalebtheythem[m]: let me try to remember :P22:15
calebtheythem[m]cheers22:15
Wizzuphttps://github.com/maemo-leste/ke-recv/blob/master/src/udev-helper.c#L6722:16
calebtheythem[m]Wizzup: ah I see, so that brings up rndis?22:19
Wizzupcalebtheythem[m]: well, I believe that will launch a set of scripts, in particular a binary (let me find the name) that sets up usbnet22:20
WizzupI think it's this one: /usr/sbin/hildon-usb-gadget-network22:20
calebtheythem[m]ah right22:20
calebtheythem[m]I think I have some script somewhere which brings it up, I'll try dropping that in for now22:21
Wizzupshould work too22:21
Wizzupthere is a script that calls that binary and also sets up the IP I think22:21
freemangordoncalebtheythem[m]: what's with the date of that thing?22:22
freemangordonomg:22:23
freemangordonOcta-core (4x2.8 GHz Kryo 385 Gold & 4x1.7 GHz Kryo 385 Silver)22:23
calebtheythem[m]freemangordon: the bloody RTC on this thing is readonly22:23
Wizzupfreemangordon: no keyboard though22:23
Wizzupiiuc22:23
calebtheythem[m]it reports the time since the battery was last removed (relative to epoch)22:24
Wizzuplol22:24
freemangordonand? you cannot set the date?22:24
calebtheythem[m]i guess some the lockscreen clock got confused22:24
calebtheythem[m]you can't write to the RTC22:24
siceloiirc this is the fastest/most powerful phone that's mainlined22:24
freemangordonyeah, and is relatively new22:25
calebtheythem[m]in postmarketOS we have a script which "solves" this by storing the now-rtc offset and updating the time22:25
freemangordon201822:25
Wizzupfreemangordon: anything in particular that should be in powervr.ini ?22:25
uvoshildons gona fly on that22:25
freemangordonI imagine leste is liek butter on this22:25
uvosunfortionatly it dosent scale22:25
freemangordonWizzup: we don;t need powervr.ini :)22:26
uvosso you need 7nm fingers to go with your 7nm process node22:26
freemangordonuvos: I think we can easily scale it22:26
uvosx can scale it22:26
Wizzupcalebtheythem[m]: btw mce will disable input devices if the device is considered in 'locked' mode, but since it vibrated it shouldn't be locked22:26
uvossure22:26
freemangordonit is clutter behind the scenes after all22:26
uvosbut it looks bad22:26
uvosx can just do this too22:26
freemangordonanyway, I am again afk22:27
uvosie blinear scale22:27
freemangordonttyl22:27
uvos*bilinear22:27
Wizzupsicelo: I get this atm: https://pastebin.com/raw/ZUCUJ3uX22:29
Wizzupsicelo: I don't remember if I got that list reversed or not22:29
WizzupI will have to recheck with some specific tests22:29
Wizzupotherwise the blockers are mostly mmc22:30
Wizzupsicelo: oh wait it doesn't actually read the live values..22:33
Wizzupok, break time :)22:35
Wizzupsicelo: yeah ok so that list needs a reversed(), just confirmed with otgusb22:47
Wizzuphttps://dpaste.com/A7PRA4QLH22:49
Wizzupoh mmc1 is the card, not wifi, nevermind me, anyway, you get the point22:49
Wizzupthese are the currnet blocking bits, maybe tmlind has some ideas tomorrow:22:51
WizzupST_SDRC22:51
WizzupST_OMAPCTRL22:51
WizzupST_I2C122:51
WizzupST_MCSPI422:51
WizzupI will also just boot to busybox a bit later to see what happens if we load almost nothing22:51
Wizzupbut not today22:51
* Wizzup afk22:51
sicelonice stuff22:52
siceloi'll definitely need to get a new sd and get my hands dirty again :-)22:54
sicelonice to see wifi appears to already idle?22:55
Wizzupyeah not too surprising22:57
siceloTRM says SDRC = SDRAM Controller. i suppose that won't really ever idle so much23:00
freemangordonand now we can finally enable -depth 16 :D23:01
sicelomcspi4 ... isn't that spi? where our wifi is sitting. can't recall if we have anything else on spi bus23:02
freemangordonhmm, where is display?23:03
siceloah, yes23:04
Wizzupa lot seems to be on spi if I read /sys correctly23:05
Wizzupoh sorry, not spi23:05
Wizzuptouchscreen is on spi23:07
Wizzuppanel as well it seems23:08
uvossudo lsof /dev/input/*23:08
uvosmaybe something holds the input device open23:08
Wizzup&mcspi4_pinsis wifi in dts23:08
Wizzup&mcspi4_pins is wifi in dts*23:09
Wizzupsicelo: and it doesn't always block23:09
Wizzupso yes, wifi idles23:10
WizzupST_SDRC23:10
WizzupST_OMAPCTRL23:10
WizzupST_I2C123:10
Wizzuponly have these now23:10
Wizzuplooks like st_i2c1 is twl23:11
uvoscheck whats on i2c1 with ic2cdetect23:11
uvosrmmod whatever is controling that device23:11
uvosno idea about the others23:11
Wizzupwhich could be a lot23:11
Wizzupalso mce and ke-recv hold input open23:11
uvosthey should23:12
uvosnon touchscreen ones23:12
Wizzupyes, keypad, power button and and gpio_keys23:12
Wizzupkeypad is a bit weird IMHO23:12
Wizzupmce     2854 root   12r   CHR  13,64      0t0  412 /dev/input/event023:12
Wizzup/dev/input/event0:TWL4030 Keypad23:12
uvosWizzup: trying to idle the keypad is a useless endeavor23:12
uvosWizzup: the kernel holds any device with any KEY _watever event open for itself regardless23:13
Wizzupanyway I'm really going to take a break now23:13

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