| - S0rin (QUIT: Ping timeout: 250 seconds) (~S0rin@user/s0rin) | 00:15 | |
| + S0rin (~S0rin@user/s0rin) | 00:18 | |
| violet | hmm my reform is not showing HDMI as an output in `swaymsg -t get_outputs` when i have it plugged in | 00:37 |
|---|---|---|
| violet | i wonder if this is like a kernel version or device tree thing. im on a pretty old one of each | 00:37 |
| violet | Linux reform 5.12.0+ #1 SMP PREEMPT Thu Jun 10 15:07:50 CEST 2021 aarch64 GNU/Linux | 00:38 |
| violet | i dont really know how to update though | 00:38 |
| violet | it seems to be booting off the SD card and then pivoting root to the internal nvme | 00:38 |
| noam | If your fstab has /boot from the SD card, `apt upgrade` should work (assuming you're on Debian) | 01:11 |
| noam | otheriwse, you might need to copy the kernel over from NVMe to the SD card after updating it on the normal file system | 01:11 |
| violet | im fully `apt upgraded` but yeah the SD is not mounted at /boot | 01:12 |
| noam | Yeah, I'd just mount it as /mnt or some such and copy the kernel over from /boot | 01:12 |
| noam | ...with the caveat that I'd keep the old one there just in case, so that you can revert if something goes wrong | 01:12 |
| violet | all i have in /boot is flash.bin, flash-rescue.bin and flash-rescue.sh | 01:12 |
| violet | theres no kernels there | 01:13 |
| noam | ...okay I have no idea then O_o My reform doesn't actually run Linux, so I was just inferring based on the Reform boot process and Linux bootup | 01:13 |
| minute | violet: did you get like an old stock or used reform with system image v2? | 01:28 |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:50) | 01:41 | |
| violet | its a used reform yeah | 01:43 |
| violet | i have no idea what system image version it is | 01:43 |
| violet | they just kinda like, gave it to me with the sd card in there and the old install on the nvme with their user wiped | 01:45 |
| - mjw (QUIT: Ping timeout: 252 seconds) (~mjw@gnu.wildebeest.org) | 01:45 | |
| - bkeys (QUIT: Remote host closed the connection) (~Thunderbi@45.134.140.137) | 01:58 | |
| - nsc (QUIT: Ping timeout: 246 seconds) (~nicolas@200-98-142-46.pool.kielnet.net) | 02:37 | |
| + nsc (~nicolas@242-48-142-46.pool.kielnet.net) | 02:39 | |
| + sl (~sl@contrib.inri.net) | 03:17 | |
| - nsc (QUIT: Ping timeout: 246 seconds) (~nicolas@242-48-142-46.pool.kielnet.net) | 03:39 | |
| + nsc (~nicolas@133-48-142-46.pool.kielnet.net) | 03:41 | |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 04:24 | |
| - natalie- (QUIT: Remote host closed the connection) (~natalie@user/natalie) | 04:24 | |
| - sl (QUIT: Quit: Leaving...) (~sl@contrib.inri.net) | 05:08 | |
| josch | violet: this sounds like you are using sysimage-v2. The disadvantage with that was, that you cannot really use apt upgrade to get the latest kernel and that the boot process does not use the normal Debian things like an initrd. This is all fixed with sysimage-v3. Unfortunately there is no automated way to go from sysimage-v2 to v3 but if you do not yet have any important data on your system maybe just | 05:50 |
| josch | flash sysimage-v3 on your sd-card and start from scratch? | 05:51 |
| violet | id rather not start from scratch but id be fine with doing a manual migration | 05:58 |
| violet | im going to start from scratch when i get around to putting gentoo on here but i have a whole line of yaks i need to shave before i get around to that | 05:59 |
| josch | violet: here are some notes about how to do it manually: https://source.mnt.re/reform/reform-system-image/-/issues/15#note_1601 | 06:00 |
| josch | violet: in addition to that, make sure that the reform repo is in your sources.list, apt install the latest kernel image and reform-tools | 06:02 |
| josch | make sure that your /etc/fstab contains the correct entries depending on where your / and /boot are | 06:03 |
| josch | { echo LINUX_KERNEL_CMDLINE=\"console=ttymxc0,115200 console=tty1\"; echo LINUX_KERNEL_CMDLINE_DEFAULTS=\"ro no_console_suspend cma=512M pci=nomsi\"; } > /etc/default/flash-kernel | 06:04 |
| josch | printf "pwm_imx27\nnwl-dsi\nti-sn65dsi86\nimx-dcss\npanel-edp\nmux-mmio\nmxsfb\nusbhid\nimx8mq-interconnect\n" > /etc/initramfs-tools/modules | 06:04 |
| violet | thanks | 06:04 |
| josch | echo "MNT Reform 2" > /etc/flash-kernel/machine | 06:04 |
| josch | those should be the bare necessities if i didn't forget anythhing | 06:05 |
| violet | whats the package name for the kernel image | 06:05 |
| josch | violet: linux-image-arm64 | 06:05 |
| josch | violet: before rebooting, you could run "sudo reform-check" to see whether the state you are in looks "normal" enough | 06:06 |
| violet | oh cool | 06:06 |
| josch | violet: your / is on nvme and your /boot on sd-card? | 06:10 |
| violet | my / is on nvme and there is nothing mounted on /boot | 06:10 |
| josch | where is your /boot content? | 06:10 |
| josch | it cannot be on nvme because u-boot cannot boot from that (yet) | 06:10 |
| violet | but the sdcard is in fact what is doing the boot because it doesnt boot when the sdcard is out. and i checked the sdcard and it has the kernel and dtbs and the reform-boot-medium file and the rest of that bootstrapping file system | 06:10 |
| josch | okay, then you need to create a new /etc/fstab entry for your sd-card and make sure it is mounted before installing thhe kernel | 06:11 |
| violet | theres just no fstab entry to put the sdcard on /boot | 06:11 |
| violet | yeah | 06:11 |
| violet | ok thats taken care of | 06:14 |
| violet | i dont have a reform-check command | 06:21 |
| violet | but i do have reform-tools installed | 06:21 |
| josch | violet: do you have the reform apt repository configured? | 06:22 |
| violet | i have deb https://mntre.com/reform-debian sid/ | 06:22 |
| josch | deb [trusted=yes arch=arm64] https://mntre.com/reform-debian-repo reform main | 06:22 |
| violet | reform-tools is version 1.0-7 | 06:22 |
| violet | ah | 06:23 |
| violet | should i delete the other line or add this under it | 06:23 |
| josch | replace it | 06:23 |
| violet | oh cool yup there's updates | 06:23 |
| violet | can i put sysimage-v3 on the emmc and leave the sd card alone to fall back on | 06:33 |
| violet | id rather get my sd card slot back for other purposes anyway | 06:34 |
| josch | yes you can | 06:36 |
| josch | but first you should partition your emmc accordingly | 06:36 |
| josch | the easiest way to do so is to run reform-flash-rescue | 06:36 |
| violet | oh does the reform-system.img not have the partition table in it | 06:37 |
| violet | (idk if thats even relevant i just downloaded it off the gitlab Ci) | 06:37 |
| josch | the image can be directly flashed to your sd-card so it includes a partition table | 06:38 |
| violet | oh i cant flash that to eMMC though? | 06:39 |
| josch | why? | 06:39 |
| violet | sorry, im asking if i can or not | 06:39 |
| violet | because you mentioned that it could be flashed to sd-card, im wondering if that image in particular is only for the sd-card | 06:40 |
| josch | yes, the image is only for sd-card as it includes an /etc/fstab with entries that reference the sd-card | 06:40 |
| josch | but theh reform-flash-rescue script runs "reform-boot-config --emmc emmc" which fixes this | 06:41 |
| violet | oh ok | 06:41 |
| Boostisbetter | does anyone know of how vaguely fast eMMC memory is compared to RAM, SD Card, NVME, etc. | 08:41 |
| josch | Boostisbetter: https://mister-muffin.de/p/kJs8.png | 08:44 |
| Boostisbetter | honestly, that isn't that slow for a sd card | 08:45 |
| Boostisbetter | is eMMC markedly faster? | 08:46 |
| josch | Boostisbetter: that is eMMC | 08:51 |
| josch | Boostisbetter: you asked for eMMC not the sd-card | 08:51 |
| Boostisbetter | Ahh ok, thank you! | 08:51 |
| Boostisbetter | That performance though is pretty similar to an sd card though right? | 08:51 |
| Boostisbetter | I'm curious if using the eMMC is something that is advisable, or if you are just as well off using the SD card. | 08:52 |
| - klardotsh (QUIT: Ping timeout: 264 seconds) (~klardotsh@98.97.36.213) | 09:19 | |
| violet | woo i am now booting from eMMC and on linux 6.3 | 09:39 |
| violet | josch: thanks for the direction | 09:39 |
| + mjw (~mjw@gnu.wildebeest.org) | 10:34 | |
| Boostisbetter | Violet: grats! So no nvme drive for you? | 10:36 |
| minute | i just paid $11786 for 50x 16GB ddr4 chips for ls1028a to a broker and hope that it won't be my first big instance of being scammed :3 | 10:51 |
| sknebel | Boostisbetter: mostly anecdotally from my experience with other SoMs, but eMMC seems to be less likely for weird failures than SD cards. performance is usually similar, since they are technically similar. | 10:52 |
| minute | (200x mt40a4g8nea-062e:F) | 10:59 |
| violet | Boostisbetter: im on NVMe for the rootfs, i just mean the boot is starting from eMMC instead of SD card now | 11:22 |
| violet | minute: hot DAMN nice, hope that works out | 11:23 |
| + lifning (~lifning@user/lifning) | 11:29 | |
| Boostisbetter | Violet: ah OK. Nice! | 11:43 |
| josch | violet: it worked?! | 12:09 |
| violet | yeah! | 12:17 |
| violet | so what i did is i flashed the reform-system.img to the eMMC and fixed up my fstab to mount it on /boot. then i installed linex-image-arm64 and linux-headers-arm64. then i ran reform-boot-config --emmc nvme | 12:20 |
| violet | i had to flash u-boot manually because `reform-flash-uboot` donwloaded a flash.bin with a different hash than `reform-flash-uboot emmc` expected (weird but ok) | 12:21 |
| violet | i decided to just not use the downloaded one at all and use a flash.bin that i built a couple days ok myself from the reform u-boot repo | 12:21 |
| violet | oh before any of that i also installed flash-kernel because that wasnt installed. and i did the changes you suggested to the flash-kernel configs | 12:23 |
| josch | nice! | 12:24 |
| josch | there are not many success stories like yours | 12:24 |
| josch | at least now it's recorded for any people who want to do that in the future :) | 12:24 |
| violet | hope it helps | 12:25 |
| violet | oh i also did not have to remove the old sway/mesa like in the issue you linked because my system happened to already be on the debian packages for that stuff | 12:28 |
| violet | which was not my doing so i dont really have any more context than that haha | 12:29 |
| violet | and i managed to switch my machine to bookworm a couple weeks ago from sid since i hadnt updated after the bookworm release, so that's neat too | 12:30 |
| - qbit (QUIT: Quit: WeeChat 3.8) (~qbit@h.suah.dev) | 12:54 | |
| josch | violet: i'm working on creating a mirror that follows stable so you can switch to that soon | 13:08 |
| josch | that mirror will then also have a gpg signature so you do not need trusted=yes | 13:11 |
| - mjw (QUIT: Killed (NickServ (GHOST command used by mark_!~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae))) (~mjw@gnu.wildebeest.org) | 13:41 | |
| * mark_ -> mjw | 13:41 | |
| + mark_ (~mjw@gnu.wildebeest.org) | 13:42 | |
| - Boostisbetter (QUIT: Ping timeout: 246 seconds) (4a410829d7@irc.cheogram.com) | 13:49 | |
| - mtm (QUIT: Ping timeout: 246 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 14:03 | |
| + qbit (~qbit@2602:ff16:3:0:1:3a0:0:1) | 14:07 | |
| + Boostisbetter (4a410829d7@irc.cheogram.com) | 14:39 | |
| minute | i am wondering if we strictly need support for the SEC engine on ls1028a. because i could immediately get chips that have the export controlled SEC engine disabled | 14:50 |
| minute | it looks like these have the armv8 cryptography extensions anyway, which includes AES... | 14:51 |
| minute | josch: do you have an opinion on that? | 14:51 |
| minute | this is an open question that is blocking me from being able to just manufacture all of the remaining ls1028a modules https://mastodon.social/@mntmn/110667215463111850 | 14:52 |
| minute | hmm, i could simulate this by benchmarking TLS and LUKS on my existing module with the SEC driver on and off, perhaps | 14:54 |
| [tj] | minute: if you knew that aes-gcm etc was disabled you could benchmark with and without before buying | 14:55 |
| + bkeys (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 15:07 | |
| + chomwitt (~chomwitt@2a02:587:7a06:dd00:1ac0:4dff:fedb:a3f1) | 15:08 | |
| Boostisbetter | minute: Is the concern primarly about performance | 15:13 |
| Boostisbetter | minute: Is the concern primarly about performance? | 15:14 |
| minute | we rolled out a new Documentation section on our website just now https://mntre.com/community.html | 15:14 |
| minute | Boostisbetter: performance of TLS and LUKS | 15:14 |
| minute | Boostisbetter: i am not sure if it is impacted | 15:14 |
| - bkeys (QUIT: Ping timeout: 240 seconds) (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 15:40 | |
| + bkeys (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 15:43 | |
| Boostisbetter | l | 15:44 |
| Boostisbetter | sorry mistype. Ok I hope are able to get in touch with them. | 15:45 |
| minute | somebody sent me this https://patchwork.ozlabs.org/project/uboot/patch/20200602150904.1997-1-michael@walle.cc/#2457448 | 15:49 |
| josch | minute: answering that sort of question is way beyond my skillset, unfortunately. If AES and friends are "export controlled" then why can others freely sell hardware with it? I thought that these algorithms being export controlled was a thing from the early 2000s but not affecting today's hardware/software. | 15:52 |
| + bkeys1 (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 15:56 | |
| - bkeys (QUIT: Ping timeout: 246 seconds) (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 15:57 | |
| * bkeys1 -> bkeys | 15:57 | |
| minute | josch: i also don't get it | 16:00 |
| josch | what an annoying issue :( | 16:01 |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 16:02 | |
| minute | josch: especially not clear if it is an issue in practice | 16:03 |
| josch | yup -- who knows whether linux makes even use of this | 16:04 |
| minute | there's a driver. i think it's comparable to CAAM in imx8mq | 16:07 |
| - bkeys (QUIT: Ping timeout: 246 seconds) (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 16:08 | |
| + bkeys1 (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 16:08 | |
| * bkeys1 -> bkeys | 16:11 | |
| - bkeys (QUIT: Read error: Connection reset by peer) (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 16:36 | |
| + bkeys (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 16:36 | |
| sevan | josch: alrighty, I'll take a look at the reform repo tonight. | 16:41 |
| + bkeys1 (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 16:46 | |
| - bkeys (QUIT: Ping timeout: 245 seconds) (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 16:47 | |
| * bkeys1 -> bkeys | 16:47 | |
| minute | josch: we should come up with a standard way of figuring out which reform processor module is active that can be used by reform-tools etc. | 16:48 |
| minute | josch: i guess via device tree in sysfs? | 16:49 |
| minute | > userland should stick to /proc/device-tree | 16:49 |
| bluerise | ack | 16:50 |
| minute | good news: we don't need any special module loading or handling on a311d. that just works now | 16:51 |
| minute | the main thing i need to solve now is tweaking something here (and in reform-display-config) so that on a311d, the correct .dtb is chosen https://source.mnt.re/reform/reform-tools/-/tree/main/flash-kernel | 16:52 |
| minute | (and later for ls1028a) | 16:52 |
| minute | rpi cm4 doesn't use u-boot so i'll let someone from the community figure out how to integrate cm4 | 16:53 |
| minute | ah echo "MNT Reform 2 HDMI" > /etc/flash-kernel/machine;; | 16:54 |
| minute | so /etc/flash-kernel/machine is the source of truth for flash-kernel | 16:55 |
| minute | oh, this has two identical ttymxc0 entries https://source.mnt.re/reform/reform-tools/-/blob/main/u-boot-menu/reform.conf | 16:56 |
| minute | root@reform:~# cat /proc/device-tree/model | 16:57 |
| minute | MNT Reform 2 with BPI-CM4 Module | 16:57 |
| minute | aha, and the dtbs are registered in /usr/share/flash-kernel/db/all.db | 16:59 |
| minute | i guess these were registered in the official flash-kernel repo? (by vagrantc?) | 17:06 |
| minute | so for now i will add the new entry to /etc/flash-kernel/db | 17:07 |
| bluerise | minute | rpi cm4 doesn't use u-boot so i'll let someone from the community figure out how to integrate cm4 | 17:07 |
| bluerise | but you can run it with u-boot? | 17:08 |
| bluerise | end even EFI/edk2? | 17:08 |
| bluerise | that's what we do | 17:08 |
| bluerise | imho: U-Boot decides which platform it is on, then picks the right DTB from the disk to boot | 17:08 |
| bluerise | all of that is pretty standard | 17:08 |
| bluerise | the only thing that you might need to do is have linux put the right u-boot to the disk? | 17:09 |
| minute | bluerise: ok, i haven't invested much time in how these things work on cm4 | 17:10 |
| minute | bluerise: so i'll say that rpi cm4 is more like a community supported platform | 17:10 |
| minute | bluerise: i will focus more on bpi | 17:10 |
| - Boostisbetter (QUIT: Ping timeout: 240 seconds) (4a410829d7@irc.cheogram.com) | 17:11 | |
| bluerise | ack | 17:11 |
| minute | oviously i will accept integration MRs! | 17:11 |
| bluerise | But I won't use the CM4 :D | 17:11 |
| minute | ah :D | 17:11 |
| bluerise | at least not the rpi4 one | 17:11 |
| minute | hehe | 17:11 |
| bluerise | I'll put in something else | 17:11 |
| minute | btw allwinner just released an 8 core a55 thing | 17:11 |
| minute | i am already addicted to cortex-a7x though | 17:12 |
| minute | ok, back to the task. so flash-kernel does the right thing now on a311d, except i need to tweak the kernel commandline | 17:12 |
| minute | i am wondering where the initial @@LINUX_KERNEL_CMDLINE@@ comes from | 17:15 |
| minute | ah, /etc/default/flash-kernel | 17:15 |
| minute | ah, this is where the machine is first set https://source.mnt.re/reform/reform-system-image/-/blob/main/reform2-imx8mq/mkuserland.sh#L28 | 17:16 |
| minute | ok cool, cool, cool. basically then i only need a MR to https://salsa.debian.org/installer-team/flash-kernel/-/merge_requests | 17:20 |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 17:22 | |
| josch | minute: i have nothing to add to what you wrote above | 17:24 |
| josch | minute: i already created a flash-kernel MR for the pocket reform: https://salsa.debian.org/installer-team/flash-kernel/-/merge_requests/48 | 17:24 |
| minute | josch: yep, seen :3 | 17:25 |
| minute | josch: thanks for weighing in | 17:25 |
| josch | minute: as said in that MR, the model name in theh device tree should be accepted upstream before adding it to flash-kernel | 17:25 |
| minute | afaik in this case the model name was accepted upstream | 17:25 |
| minute | (bpi case) | 17:25 |
| josch | minute: i think you still wanted to change the model name for the a311d dtb upstream? | 17:25 |
| josch | ah | 17:26 |
| josch | well then that's thaht | 17:26 |
| minute | ah, well, theoretically we could, but unfortunately i didn't write down what we wanted to call it instead | 17:29 |
| minute | ok, so flash-kernel ~almost~ succeeded. the only issue is that this u-boot has a default (wrong) $fdtfile set | 17:29 |
| + Boostisbetter (4a410829d7@irc.cheogram.com) | 17:30 | |
| - mark_ (QUIT: Ping timeout: 245 seconds) (~mjw@gnu.wildebeest.org) | 17:30 | |
| josch | does it work with upstream u-boot or is there again a custom u-boot fork? | 17:37 |
| minute | josch: upstream actually | 17:41 |
| minute | so i guess i can just tweak the bananapi-cm4-cm4io_defconfig in u-boot | 17:42 |
| josch | vagrantc: talking about the undistributable proprietary blobs again: support for the mechanism to load the non-free blob from a package might still be useful even if the package containing the blobs cannot be shipped by non-free-firmware | 17:49 |
| josch | vagrantc: because MNT is allowed to distribute them, we could ship the blobs in the MNT Debian package repo | 17:49 |
| josch | vagrantc: so by adding support for that mechanic, even if it's not useful for Debian, would make it easier for downstream distros | 17:50 |
| bluerise | minute: For the RK-Based one it's nice because they are mostly only defconfig+dts nowadays, I'd actually (at least for rockhip) create one defconfig and a dts that includes the cm4/cm3/whatever dtsi | 17:58 |
| bluerise | the cm4io is probably just a dts that includes a dtsi? | 17:59 |
| bluerise | might make it easy to upstream to u-boot :) | 18:02 |
| minute | yep | 18:02 |
| + grubaroni (~user@user/grubaroni) | 18:06 | |
| minute | trying to build the first a311d system image... | 18:12 |
| - robin (QUIT: Quit: Leaving) (~robin@user/terpri) | 18:21 | |
| + robin (~robin@user/terpri) | 18:30 | |
| minute | ugh, the debian kernel doesn't even have the rtw88 sdio driver yet...... | 18:33 |
| minute | ... but it was relatively easy to drop in the patch series | 18:45 |
| vagrantc | josch: sure | 18:45 |
| vagrantc | josch: dropped imx8mq from arm-trusted-firmware in debian :/ ... so will also have to carve out dummy space for that too | 18:46 |
| + klardotsh (~klardotsh@98.97.36.213) | 18:55 | |
| minute | josch: i wonder if it would make sense to split out the kernel package from reform-debian-packages. right now we have to rebuild a lot of stuff if there is a kernel patch added etc | 19:05 |
| josch | vagrantc: no rush -- personally i'd switch hthe imx8mq u-boot to upstream once display support is available there as well | 19:05 |
| josch | vagrantc: display support is really needed for stuff like u-boot-menu | 19:05 |
| josch | minute: we can split it out but build.sh contains code that only rebuilds the stuff that has a new version | 19:06 |
| josch | minute: if a source package didn't change in unstable, then it will not be rebuilt | 19:06 |
| minute | oh ok | 19:06 |
| minute | josch: are the versions compared to what is already in the repo or how does it know that? | 19:07 |
| josch | yes, it compares the versions in unstable to the version in the repo | 19:08 |
| minute | ok, thx | 19:09 |
| minute | asking because this is building a lot of stuff (gst and so on) https://source.mnt.re/reform/reform-debian-packages/-/jobs/1329 | 19:09 |
| josch | oh i thought you meant locally! | 19:09 |
| minute | no ;__; | 19:09 |
| josch | on the CI it always gets rebuilt | 19:09 |
| josch | uff... i think i messed up somethhing with my keyboard when i upgraded my lpc | 19:10 |
| minute | oh no, what's wrong with it? | 19:10 |
| josch | my h key is triggering multiple times when i press it :( | 19:10 |
| josch | so i often have to backspace the additional h letters... | 19:10 |
| minute | hmm, i had this recently with a space key, it was just an aged solder joint, so touching it up made it work perfectly again | 19:11 |
| josch | that is done from theh bottom of theh keyboard? | 19:11 |
| minute | and someone else here fixed their repeating letter (another one) by tweaking the debounce code in the keyboard fw | 19:11 |
| minute | josch: yep, each keyswitch has 2 solder joints on the bottom | 19:11 |
| josch | okay, i will try that during the weekend | 19:11 |
| josch | i'm using the keyboard brace suggested here: https://community.mnt.re/t/reform-mod-idea-internal-keyboard-brace/1373 | 19:12 |
| josch | and i noticed too late that when i closed the acrylic bottom, one of the internal usb cables was stuck between the acrylic plate and the keyboard brace | 19:12 |
| josch | and that pressed the keyboard upwards so that it would be bent up | 19:12 |
| josch | i fear the bending must've been too much | 19:13 |
| minute | hmm, maybe there was already a cold solder joint and it cracked a little bit, so the contact is noisy | 19:15 |
| minute | my issue with the packages is that they are building since over 30 minutes and haven't even reached the kernel yet, but i need to merge the kernel and deploy a new kernel package to be able to really test the new image build | 19:17 |
| josch | minute: if you just want the kernel you can quickly disable all the other packages either by commenting out the code in build.sh or by running chmod -x patches/* | 19:19 |
| josch | patches/ffmpeg is already disabled that way | 19:19 |
| minute | oh, so it is gstreamer that is taking so long? | 19:20 |
| josch | the patches are actually scripts and if they are missing the executable bit, then they do not get applied | 19:20 |
| minute | if i were to merge that, the old version of those packages would be kept in the repo, yes? | 19:20 |
| josch | minute: the reason it is taking so long is gobject-introspection | 19:20 |
| josch | minute: any package using gobject-introspection cannot be cross-built | 19:20 |
| josch | so what the script does instead is to use qemu to do an emulated native build | 19:21 |
| minute | oooof | 19:21 |
| josch | and that's why it takes so super long | 19:21 |
| minute | i would really prefer this to be split into a different repo / build process | 19:21 |
| josch | reasons for why anything using gobject-introspection cannot be cross-built are here: https://wiki.debian.org/CrossBuildPackagingGuidelines#gobject-introspection | 19:21 |
| minute | otherwise it will be hard to maintain this process in the next weeks when several new cpu modules have to be supported | 19:21 |
| josch | minute: splitting this up is no problem and i do not think such a split would be a bad idea because the build process is very different from the rest | 19:21 |
| minute | ok cool | 19:22 |
| josch | minute: you just have to adjust your cronjob that assembles the repo on mntre to now somehow combine the packages from two repositories | 19:22 |
| minute | i am also wondering if we still need the xwayland patches, as they're now a year+ old | 19:22 |
| minute | josch: ok | 19:22 |
| josch | minute: or maybe things should be kept in the same repo but the scripts should be split? | 19:23 |
| minute | josch: how would that work? different pipelines or something? | 19:24 |
| josch | the pipeline jobs could be split | 19:24 |
| minute | yeah | 19:24 |
| minute | sounds good if that works in practice | 19:24 |
| josch | i think it should work (but i have to try) to define two jobs in ./.gitlab.yml -- one for the normal packages and one for the kernel | 19:27 |
| josch | then when you run the pipeline you can cancel the jobs you don't want | 19:28 |
| josch | or you can manually comment out one line in ~/.gitlab.yml temporarily until you are happy with the result | 19:28 |
| josch | this would have the advantage that a common repo would be built and you do not have to somehow (i would not know how) merge two repos into a single one | 19:28 |
| minute | yeah | 19:36 |
| minute | sounds good | 19:36 |
| minute | ok, the next slow build is box64 | 19:36 |
| vagrantc | josch: now would be a good time to at least get major changes that trigger NEW processing ... | 19:36 |
| minute | would be awesome if that was packaged for real in debian | 19:36 |
| vagrantc | josch: i think the main quirk left with mainlining the display patch is to turn it off before booting the kernel (so that the kernel can initialize the eDP state or whatever) | 19:37 |
| vagrantc | we'll only be able to kind of half-assedly do it in debian ... you still will have to bring your own binary blobs | 19:38 |
| minute | weird, the box64 build just hangs at 42% https://source.mnt.re/reform/reform-debian-packages/-/jobs/1330 | 19:38 |
| josch | minute: yes, i need to upload this soon then it can be dropped from the pipeline | 19:40 |
| minute | josch: nice | 19:41 |
| josch | i couldn't have done this a few weeks ago when i finally got stardew valley to run because Debian was in full freeze | 19:41 |
| josch | now that stardew valley works i'm actually using it and it doesn't make sense to package something one is not using :r | 19:41 |
| - S0rin (QUIT: Ping timeout: 245 seconds) (~S0rin@user/s0rin) | 19:44 | |
| minute | oh yeah :3 | 19:44 |
| + S0rin (~S0rin@user/s0rin) | 19:46 | |
| minute | ok, kernel package upgrade on a311d reform worked well | 19:50 |
| josch | grrrr my stupid remote keyboard... | 19:50 |
| minute | josch: what's up with it | 19:50 |
| + mark_ (~mjw@gnu.wildebeest.org) | 19:50 | |
| josch | minute: if you ever sell a wireless version of the reform keyboard, you will have my money (again) | 19:50 |
| minute | oh! | 19:51 |
| minute | well, then we should do that ^^ | 19:51 |
| josch | haha ;) | 19:51 |
| josch | i'm happy with your a311d progress -- i'm happy that now i'm not the only one anymore who understands how all the bits and pieces of sysimage-v3 are working together :D | 19:52 |
| minute | oh yeah, this has been very enlightening! | 19:52 |
| - bkeys (QUIT: Remote host closed the connection) (~Thunderbi@static-198-54-130-101.cust.tzulo.com) | 20:05 | |
| minute | odd, there's some bug in gitlab, it just stops streaming the log after some point | 20:07 |
| minute | the server is actually building the kernel but i can't see that in the gitlab job | 20:07 |
| josch | that's not a bug but a feature | 20:07 |
| josch | gitlab has a maximum size of log it keeps | 20:08 |
| josch | minute: that is what the ./filter-output script is for | 20:08 |
| josch | it makes sure that it displays up 60% of the maximum allowed size and then stops | 20:08 |
| josch | that way, when something fails, the bottom 40% of the allowed log size can be filled with the last lines from the real log | 20:09 |
| josch | and the real log is stored as output.log | 20:09 |
| josch | i don't know if you as the gitlab server admin can increase the log size limit | 20:09 |
| - chomwitt (QUIT: Remote host closed the connection) (~chomwitt@2a02:587:7a06:dd00:1ac0:4dff:fedb:a3f1) | 20:10 | |
| minute | ohhh ok josch | 20:10 |
| minute | this is configurable on the build runner machine | 20:11 |
| minute | for the next build i will upgrade it to 10MB | 20:11 |
| josch | you are building a linux kernel | 20:12 |
| josch | 10 MB are not enough :D | 20:12 |
| josch | minute: the last log was 101 MB: https://source.mnt.re/reform/reform-debian-packages/-/jobs/1325/artifacts/browse | 20:13 |
| minute | oh!! | 20:13 |
| josch | one practical reason for the limit is probably to prevent filling a browser window with 100 MB of log files | 20:13 |
| minute | i shall increase it to 256 MB :D | 20:13 |
| minute | ok, so the rtw88 driver has issues with this old kernel | 20:14 |
| josch | minute: 6.4 is already in experimental so you could try s/unstable/testing/ in build.sh | 20:15 |
| josch | errr | 20:15 |
| josch | s/unstable/experimental/ | 20:16 |
| - mjw (QUIT: Killed (NickServ (GHOST command used by mark_!~mjw@gnu.wildebeest.org))) (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 20:18 | |
| * mark_ -> mjw | 20:19 | |
| + mark_ (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 20:19 | |
| minute | > Job succeeded | 20:21 |
| minute | josch: livi also takes quite some resources to build, are you using it as your video player? | 20:23 |
| josch | nope, i'm a clapper user | 20:23 |
| minute | ok, i will comment it out then for now | 20:24 |
| josch | makes sense | 20:24 |
| josch | also, i cannot package box64 yet | 20:24 |
| minute | :0 | 20:24 |
| josch | i have been running the version from git which includes important fixes for arm64 | 20:24 |
| minute | oh? | 20:24 |
| josch | and there hasn't been a new release since march | 20:24 |
| minute | oh :( | 20:24 |
| josch | so that would be broken | 20:24 |
| josch | actually... | 20:24 |
| josch | let me upload the old version | 20:24 |
| josch | because we have to wait for NEW processing anyway | 20:25 |
| minute | josch: objections? https://source.mnt.re/reform/reform-debian-packages/-/merge_requests/7 | 20:28 |
| josch | well.. this still includes my two unfinished wip commits | 20:29 |
| josch | did you try this out on the pocket already? | 20:29 |
| josch | then i'll finish those | 20:30 |
| minute | ah yeah, sorry i didn't realize until a second ago that this was based on the pocket reform branch | 20:30 |
| minute | but the pocket is not deployed yet, so it won't hurt anyone | 20:30 |
| josch | did you try the kernel on the imx8mq as well? | 20:31 |
| josch | just to make sure it doesn't accidentally break that? | 20:31 |
| minute | no, i shall do that now | 20:31 |
| josch | brb bringing baby to bed | 20:31 |
| minute | i will copy the .debs to my personal reform and install it there | 20:31 |
| minute | hmm, so it boots fine, but display is shifted slightly to the right... grr | 20:39 |
| minute | omg shit i know why | 20:40 |
| minute | sorry for swearing | 20:40 |
| minute | josch: good that you asked for a test on imx8mq | 20:41 |
| minute | the problem is that i had to adjust display panel timings for a311d because their dsi driver does some wrong calculations | 20:41 |
| minute | so i will have to temporarily make up a new panel variant and put that in our a311d dts... | 20:42 |
| josch | fun! :D | 21:09 |
| - mtm (QUIT: Ping timeout: 246 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 21:19 | |
| minute | did that, offset is still there. so that must be the BURST_MODE in sn65dsi86 driver then... | 21:39 |
| sevan | new GPUs for MI4! https://blog.imaginationtech.com/imagination-gpus-now-support-opengl-4.6 going to need a risc-v cpu now though. :) | 21:58 |
| josch | hrm... i think i need to order some more cells from eremit | 22:02 |
| josch | my battery meter showed 35% remaining and then my reform switched off with two of the cells at 2.6 and 2.7 V respectively | 22:02 |
| sevan | 10 months on, mine sit around 3.7 fully charged. really like that they charge quickly. Useful now that I take the machine out to the park (London is a dangerous dump) | 22:05 |
| - jacobk (QUIT: Ping timeout: 260 seconds) (~quassel@47-186-122-163.dlls.tx.frontiernet.net) | 22:28 | |
| Boostisbetter | josch: My cells were about a year old. I have only had one fail, but just sharing as that is my experience. | 22:31 |
| minute | tried to implement a new "burst-mode" dts flag for ti-sn65dsi86 that is set for a311d but not for imx8mq | 22:53 |
| sevan | heh, with the latest cycle, cells range from 3.3 to 3.6 :) | 23:03 |
| minute | ok, display is fine again on imx8mq | 23:03 |
| josch | \o/ | 23:03 |
| sevan | proceeding with the latest updates despite " #1040178 - linux-kbuild-6.3: missing Breaks: linux-headers-6.3.0-1-common", seems to be related to nvidia kms? | 23:05 |
| sevan | Unpacking reform-handbook (1.1) over (1.1) ... < need a version bump? | 23:05 |
| minute | ok, and display is also fine on a311d with the same kernel! | 23:07 |
| minute | josch: the a311d+imx8mq kernel build is fine now i think. but will sleep over it and test a bit more after sleep | 23:16 |
| minute | pic https://mastodon.social/@mntmn/110669204450596858 | 23:21 |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 23:25 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
