libera/#maemo-leste/ Friday, 2025-03-28

Wizzupfreemangordon: https://phoenix.maemo.org/job/libcal-source/65/console works with git.maemo.org00:33
Wizzupreconfiguring the core jobs00:34
Wizzupfreemangordon: I set up push mirroring for conversations and hildon-meta and have a programmatic way to set it up now, and it seems to work fine01:17
WizzupI will set it up for all the other repos tomorrow01:17
dsc_:)01:19
Wizzuphttps://git.maemo.org/leste/jenkins-integration/commit/aff5d1a6e3db3c000801508fcde814ed5f74044102:01
Wizzupfreemangordon: do we want to migrate anakin-linux / lime2-linux / linux-beryllium / rockchip-linux ?02:27
Wizzupcloning all upstream-forks now too, but let's (re)move some of these to leste-deprecated or so02:46
sicelowe're not building beryllium images anymore?09:07
siceloah, last build over 3 years ago and no one has complained ... a bit sad, since the device is very well supported by mainline. iirc recently camera started working too. anyway, if no one has it and maintains it, it's fate is sealed09:09
halftuxWizzup: The magic for debian rpi and boot partition gets installed by the raspi-firmware package and it can self trigger a script or it gets triggerd when kernel gets updated10:11
halftux/etc/initramfs/post-update.d/z50-raspi-firmware10:11
halftux/etc/kernel/postinst.d/z50-raspi-firmware10:12
halftux/etc/kernel/postrm.d/z50-raspi-firmware10:12
halftux/var/lib/dpkg/info/raspi-firmware.postinst10:12
halftuxI guess the boot partition gets mounted to rootfs /boot/firmware10:13
halftuxetc/fstab:4:LABEL=RASPIFIRM /boot/firmware vfat rw 0 210:13
halftuxthe scripts will copy new files to /boot/firmware10:15
halftuxsicelo: the firmware and bootloader for rpi needs to be in the vfat because a CPU inside the GPU starts a binary blob from a small ROM which then loads the GPU firmware from the first fat partition. The GPU loads then the linux kernel from the fat partition into the RAM. After this the ARM CPU is switched on and booting with the loaded kernel. So the GPU is acting like a BIOS.11:06
halftuxhere is a better explanation https://bits4device.wordpress.com/2017/03/06/raspberry-pi-boot-process/11:09
siceloah yeah, I remember that now. never had an RPi. RPi4 still works that way, or it's up to the 3?11:11
halftuxrpi 4 is similar but has now also an eeprom11:12
Wizzupsicelo: not since beowulf I think11:35
Wizzupsicelo: which device is that, the pocophone?11:35
Wizzuphalftux: thanks for all the resources, so it seems like we should just ensure the right labels get set?11:36
halftuxyes for the updates we should have the same labels. For the first image build I don't know if the raspi-firmware package can be installed11:39
Wizzuphm, why not?11:46
Wizzupfreemangordon: push mirrors are now set up for core repos, this means that git.maemo.org will keep github in sync, but it's a one way sync: it's possible that if you now push to github our instance will remove it11:47
Wizzupall upstream repos push back now  too11:54
halftuxWizzup: I can't judge I have not enough knowlegde about image builder. When the partiton gets created from one changeroot it could work. However I could try to build a new image. So which bit I must tweak to get first partition mounted on /boot/firmware instead of /boot? Is it somewhere in the sdk library in the helper script?12:48
Wizzuphalftux: just for my understanding, is there still a place for /boot as well?12:49
halftuxOr could I change it in the board config or board shell script12:49
halftuxboot must be on the second partition the kernel gets installed there. So in principle you have the kernel twice12:50
Wizzupso there are different partitions for /boot and /boot/firmware12:51
Wizzupbtw, I wonder if we should just switch to debos or whatever gets used these days and make that works with maemo/devuan12:51
Wizzuphttps://github.com/go-debos/debos12:51
halftuxyes /boot is a folder in rootfs and /boot/firmware the mount point for first vfat partition12:51
Wizzupah ok, so /boot does not have its own partition then12:52
halftuxand the script also checks if something is mounted on /boot/firmware12:52
halftuxcorrect12:52
Wizzupok12:52
Wizzupcan we collect some of that info on an issue here? https://git.maemo.org/leste/bugtracker12:53
WizzupI will look at the image-builder changes that we can make, although I feel like perhaps in the near future we should switch to using debos or something more 'standard' for building images12:53
Wizzupthe image-builder has a certain way it wants things to go (boot on separarte partition, root partition and nother else)12:54
Wizzupso that's not ideal12:54
Wizzupbut I love the idea of using debian kernel and firmware packages12:54
Wizzupand this looks pretty simple https://github.com/go-debos/debos?tab=readme-ov-file#simple-example12:55
Wizzuphttps://github.com/go-debos/debos-recipes/tree/main/rpi6412:55
halftuxso should I still fill a bug? Is it worth, chance to get it changed is low and maybe makes no sence when we step back from image-builder. If you still want one what should be the title?12:59
WizzupI think just something like 'building raspi images again'13:02
Wizzupand yes issue is very helpful13:02
Wizzupand I will look at changing the image builder, but longer term, I think we want to do whatever debian is doing13:02
halftuxok13:03
WizzupI really want raspi images and the research here is super helpful13:04
halftuxmaybe it is also a solution to mount vfat partition to /boot and to /boot/firmware it is not nice but I think it could work13:04
halftux...when this is possible with fstab13:05
Wizzupright, something like this could be done13:09
Wizzupwe can run shell scripts in various parts of the process13:09
Wizzupit's just finnicky13:10
Wizzupdid you manage to build an image without the additional firmware partition, manually copy some things and get that to work?13:10
WizzupI'd be interested in that13:10
halftuxyes I used a tiny raspi3 chimaera image copied /boot/*kernelfiles from linux-image-arm64 to vfat and also firmware files from https://github.com/raspberrypi/firmware/ (not tried files from raspi-firmware, but should work)13:17
halftuxmodified config.txt (sorry for spam):13:18
halftuxarm_64bit=113:18
halftuxenable_uart=113:18
halftuxupstream_kernel=113:18
halftuxdevice_tree=bcm2837-rpi-3-b-plus.dtb13:18
halftuxkernel=vmlinuz-6.1.0-32-arm6413:18
halftuxinitramfs initrd.img-6.1.0-32-arm6413:18
Wizzupdoes linux-image-arm64 get installed in our image or did you also install it manually?13:19
Wizzup(also; this would be great to get in the issue report)13:19
halftuxI need to look into my new build image. If it does.13:20
halftuxBut it does with the generic-arm64 build so I guess yes13:21
Wizzupah sorry, this was the arm64 build?13:21
WizzupI thought you said the raspi3 image, sorry13:21
WizzupI have a bunch of raspis here and I figured if I can try it here it might be easier to look at the image-builder things13:21
halftuxI used raspi3 image but kernel from generic-arm64 build13:22
halftuxnow I edited the raspi3 script remove kernel build and put the /bin/true kernel from generic build into raspi3 script13:23
halftuxoh I forgot to edit beowulf.config to add linux-headers-arm64 and linux-image-arm64 to raspi*13:33
Wizzupdid you edit the beowulf config or the chimaera one?13:43
halftuxoh sorry for confusion I now try daedalus befor I build chimaera but there I did not edit config everything manual13:46
halftuxand you need to copy all dtb files from /usr/lib/linux-image-6.1.0-32-arm64/broadcom/ to vfat bootfs13:56
Wizzuphalftux: ok, can you pls put this in the git.maeom.org issue14:00
WizzupI'll dig up my rpis14:00
Wizzuphey I have a raspi3 here :)14:02
Wizzuptwo atually14:02
halftuxwhen I log into git.maemo.org with github account I get Internal server error 50014:09
halftuxshould I create an account?14:09
halftuxok I managed it with the second try14:13
Wizzuphalftux: did you log in with github?14:13
siceloWizzup: +1 for debos14:13
halftuxyes14:14
halftuxI will wait for image build and then writing the issue with the new steps.14:22
Wizzuphalftux: ok, thanks, and please do, the issue will help keep track and give context14:24
arno11i got the server error 500 too, but now able to log in to GH and git.maemo.org from n900 using new dillo :D15:29
arno11it is fast. we really need to build the new dillo imo15:29
arno11however, GH UI is not really usable with new dillo but forgejo is quite ok15:32
Wizzup:)15:33
Wizzuparno11: I can make a repo for dillo, should it be upstream forks or extras?15:33
Wizzupprobably extras I guess15:34
Wizzuphttps://git.maemo.org/leste-extras15:34
Wizzuperr https://git.maemo.org/leste-extras/dillo15:34
arno11cool15:38
arno11will try to add it this evening15:38
WizzupI added you as collaborator to it15:39
Wizzupusually the easiest way is to take debian's packaging and add it to the new source15:40
Wizzupand then update the changelog etc15:40
arno11ok np15:42
arno11(btw i tried to add a random comment to an issue from dillo, it works fine. display is ugly but fast15:44
Wizzup:)15:45
arno11i also finally find the motivation to debianize maemo-gpt. still some code to clean but almost ready now for extras15:49
Wizzupcool, hopefully it isn't too hard to figure out the packaging15:49
Wizzuponce you get used to it, it's mostly just copy pasteing some boilerplate15:49
arno11i meant *found. debian stuff is already ok15:51
Wizzup:)15:52
halftuxthere is already a rpi image issue from you Wizzup. I will comment to that. restore raspberry pi images #75815:58
mkfgot my n900 rooted \o/16:04
mkfs/n900/v20/g16:04
mkfnext is getting mainline linux running on it16:05
arno11oh nice16:19
arno11Wizzup: btw i still need to port pcsx and pico to daedalus (quite trivial) but i still need a fix for osso-games-startup (issue #763) for drnoksnes16:25
Wizzupsounds good16:28
sicelonice. upower upstream mostly approves of our patch ... there are a few modifications to make, but otherwise there will be minimal adjustments :-)16:43
Wizzupsweet!16:46
Wizzupmkf: great16:46
Wizzuphalftux: ok, pls comment on https://git.maemo.org/leste/bugtracker/issues/758 indeed :)16:46
halftuxdone!16:50
halftuxI tried to log in with github there was that error and I tried again and there was no need of a password but asking if I would connect to an existing account and I said no16:54
halftux^^ignore that :-)16:55
Wizzupok :)16:56
halftuxWizzup: but I posted the instruction hope everything is clear16:58
Wizzuphalftux: this is perfect, thank you17:06
halftuxcompared to chimaera there are some issues in daedalus I made a second comment.17:13
Wizzupstrange that usb keyboard wouldn't work, maybe different debian kernel version17:14
freemangordonWizzup: what about CI?17:15
freemangordonwhat repos will it build?17:15
Wizzupgit.maemo.org17:16
Wizzupfor everything except extras and l10n17:16
Wizzupchanges on git.maemo.org to core and upstream-forks are auto pushed to github upon commit/push, and also synced every 8 hours17:17
freemangordongreat17:18
Wizzupthis is done using an api key tied to my account, there was no way to do that with an api key tied to an orga17:18
Wizzupbut I think that's fine17:18
WizzupI can this locally:17:19
WizzupI ran*17:19
Wizzupsed -i 's/git@github.com:maemo-leste\//git@git.maemo.org:leste\//' */.git/config17:19
halftuxWizzup: your right to many keyboards on my table it is working.17:20
Wizzup:D17:20
WizzupI guess that means hdmi wake up also works then17:20
halftuxyes think so but I also disabled the lock screen17:23
halftuxafter wake up mouse is in the left upper corner and does not move only after a mouse click17:24
Wizzuphalftux: oh yes the lock screen might be the issue here, you need to press the 'power button'17:43
Wizzuponly after click could be a power feature17:44
Wizzuppower saving17:44
Wizzuphalftux: building an image with your changes on phoenix now, it doesn't have the final steps yet, but I at least made all your mods17:51
halftuxok nice hope it will work made a small untested change to daedalus.config that cmdline.txt and config.txt gets copied. Therefore I added copy-root-overlay. I just building it now with this change. But just for record check if the two files are on the bootfs partition.17:55
halftuxnot daedalus.config I mean raspberry-pi3.sh17:56
WizzupI think we can make a change to set labels on some of the partitions in the devuan sdk for image building, do you know if we need to use dos or gpt for these labels?17:58
halftuxI don't know but I would use dos for rpi18:00
Wizzupok, and we can set labels that way?18:07
Wizzupbecause that is what we need right?18:07
halftuxwe don't need labels we can use partitions in fstab18:09
halftuxso debian are using labels in cmdline.txt and fstab18:09
halftuxyou are building chimaera raspi3 image right?18:10
halftuxdid you edited the right config for kernel usage?18:11
halftuxmaybe I missed something in regard of the labels but I think we are fine if we are using /dev/mmcblk0p1 and /dev/mmcblk0p1 unless we want to support more options as only sd cards18:25
Wizzuphalftux: oh, I thought the postinst scripts/triggers you mentioned look at the labels, but if they just look at things being mounted, it might be ok18:26
Wizzup18:10 < halftux> you are building chimaera raspi3 image right?18:26
Wizzuphm... yeah, oops, let me cancel18:26
Wizzuphttps://phoenix.maemo.org/view/Images/job/leste-image-raspi3-64bit-daedalus/1/console18:27
halftuxhmm script uses /usr/bin/systemd-detect-virt18:33
halftuxdo we have that? systemd*18:33
halftuxfindmnt --noheading --first-only --nofsroot --output="$1"18:36
halftuxROOTPART="$(root_info source)" || ROOTPART=/dev/mmcblk0p218:36
Wizzupprobably not, but I'll need a local image to be able to do more analysis18:36
Wizzupthat's why I'm building it in ci/cd18:36
halftuxdon't know what ROOTPART will be18:37
halftuxI think when everything fails ROOTPART=/dev/mmcblk0p218:38
halftuxelif ! mountpoint -q /boot/firmware ; then18:39
halftuxI guess we need to try if it fails or not18:39
Wizzupit looks like weblate can't make merge requests for translations on our own git service22:41
Wizzupat least not hosted weblate22:41
sicelommm, basic skim says it should22:46
Wizzuphm?22:46
sicelohttps://docs.weblate.org/en/latest/vcs.html#vcs-gitea (i guess forgejo should work same)22:46
Wizzupyes, this is only for gitea.com22:46
WizzupI have no way to provide any credential22:47
Wizzupthis has to be set in the admin cfg of the weblate22:47
Wizzupwe can maybe have it auto push to a branch22:47
Wizzupbut merge requests (not auto push) does not seem like it is an option22:47
Wizzupsee https://docs.weblate.org/en/latest/admin/config.html#std-setting-GITEA_CREDENTIALS22:47
siceloit reads to me like ... you add the API key per your last link. and it'll make a Merge Request23:03
siceloit only pushes if you're using 'pure' git23:03
Wizzupright, but there is no way to set any of this23:11
Wizzupthis is done per weblate instance23:11
halftuxWizzup: image build faild but it was building the image until the end :(23:23
Wizzupugh23:38
Wizzupbuilding another one23:39

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