| + jacobk (~quassel@utdpat241106.utdallas.edu) | 00:00 | |
| - mjw (QUIT: Ping timeout: 268 seconds) (~mjw@gnu.wildebeest.org) | 00:04 | |
| - chomwitt (QUIT: Ping timeout: 268 seconds) (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 00:13 | |
| minute | josch: not yet, because it's not safe yet (permissions are off), but that would be the plan | 01:12 |
|---|---|---|
| - mtm (QUIT: Ping timeout: 264 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 02:04 | |
| - jacobk (QUIT: Ping timeout: 268 seconds) (~quassel@utdpat241106.utdallas.edu) | 02:06 | |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:50) | 02:08 | |
| - nocko (QUIT: Ping timeout: 264 seconds) (~nock@user/nocko) | 02:10 | |
| + nocko (~nock@user/nocko) | 02:12 | |
| + Gooberpatrol66 (~Gooberpat@user/gooberpatrol66) | 03:24 | |
| - nsc (QUIT: Ping timeout: 240 seconds) (~nicolas@24-48-142-46.pool.kielnet.net) | 03:30 | |
| + nsc (~nicolas@i5E86127A.versanet.de) | 03:32 | |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 04:10 | |
| + jacobk (~quassel@utdpat242016.utdallas.edu) | 05:46 | |
| swivel | minute: is it still recommended that I buy my imx8 reform through CS or should I buy direct from you? | 05:50 |
| swivel | I'm in the US(cali) fwiw | 05:51 |
| + colinsane1 (~colinunin@97-113-159-4.tukw.qwest.net) | 05:51 | |
| - colinsane (QUIT: Ping timeout: 260 seconds) (~colinunin@97-113-159-4.tukw.qwest.net) | 05:52 | |
| - jacobk (QUIT: Ping timeout: 240 seconds) (~quassel@utdpat242016.utdallas.edu) | 06:02 | |
| + jacobk (~quassel@129.110.242.173) | 06:23 | |
| - colinsane1 (QUIT: Quit: bye) (~colinunin@97-113-159-4.tukw.qwest.net) | 07:03 | |
| - ggoes (QUIT: Ping timeout: 260 seconds) (~gregf@fsf/staff/ggoes) | 07:05 | |
| minute | swivel: yeah, CS has them in stock | 07:17 |
| - mtm (QUIT: Ping timeout: 268 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 07:27 | |
| + ggoes (~gregf@fsf/staff/ggoes) | 07:29 | |
| josch | swivel: i think there was somebody in the forum who wanted to sell theirs? | 07:33 |
| josch | minute: you should add the kernel version to the package name so that it becomes possible to install multiple of those at the same time | 07:33 |
| josch | minute: because you don't know which version is currently running | 07:34 |
| josch | just having a kernel package installed does not mean it's the one the user is running | 07:34 |
| + colinsane (~colinunin@97-113-159-4.tukw.qwest.net) | 07:35 | |
| + chomwitt (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 07:36 | |
| josch | minute: instead of installing to /etc/modprobe.d, consider installing into /usr/lib/modprobe.d/ | 07:44 |
| josch | what you install into the former are conf-files and removing or upgrading your package will not easily change or remove it in case it becomes necessary | 07:44 |
| josch | remember we had the problem with lots of other stuff that used to be installed to /etc by the system image and reform-tools and it was painful to change that | 07:45 |
| josch | about making the package name versioned so that multiple different modules can be installed for the various installed kernel: in that case you need to move the files that are currently shared into another package because no two packages can claim the same file | 07:47 |
| josch | unless they are Multi-Arch:same but in your case, the package version will differ so they cannot be co-installable | 07:47 |
| - colinsane (QUIT: Quit: bye) (~colinunin@97-113-159-4.tukw.qwest.net) | 07:56 | |
| + colinsane (~colinunin@97-113-159-4.tukw.qwest.net) | 07:59 | |
| - chomwitt (QUIT: Ping timeout: 268 seconds) (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 08:34 | |
| + chomwitt (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 09:18 | |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 09:32 | |
| + mjw (~mjw@gnu.wildebeest.org) | 09:58 | |
| - chomwitt (QUIT: Ping timeout: 256 seconds) (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 10:17 | |
| josch | minute: i'm working on the next reform-tools version, fixing the bug that was found in reform-flash-uboot. In the process I'm also improving support for pocket reform by implementing an additional variable that stores whether the platform can boot from sd-card. The imx8mplus is the only one that cannot load u-boot from sd-card, correct? | 10:19 |
| - mjw (QUIT: Ping timeout: 260 seconds) (~mjw@gnu.wildebeest.org) | 11:21 | |
| * Guest2951 -> mjw | 11:23 | |
| - jacobk (QUIT: Ping timeout: 240 seconds) (~quassel@129.110.242.173) | 12:03 | |
| - nsc (QUIT: Ping timeout: 272 seconds) (~nicolas@i5E86127A.versanet.de) | 12:25 | |
| minute | josch: ah yeah, good points about package name & /usr/lib, and will move firmware to a separate deb | 12:45 |
| minute | josch: about imx8mp, so far yes, rk3588 situation is not yet clear about this | 12:46 |
| minute | josch: the remaining question about qcacld: how could we make sure users have a chance to install the new version for their kernel package without forcing it on everyone? | 12:49 |
| minute | because if it's a different package for every kernel, apt upgrade won't upgrade it | 12:49 |
| + jacobk (~quassel@64.189.201.150) | 12:49 | |
| minute | then people could suddenly end up without wifi after apt upgrade and reboot | 12:50 |
| + nsc (~nicolas@i5E86127A.versanet.de) | 12:52 | |
| josch | minute: you'd do the same thing as the kernel does | 13:06 |
| josch | the kernel has one package linux-image-arm64 and that depends on the latest linux-image-${version}-arm64 | 13:06 |
| josch | the problem though (again) is that then apt will mark the old package for autoremoval | 13:07 |
| josch | not knowing that it might still be needed, depending on what kernel the user is running | 13:07 |
| josch | so the next question then becomes: how does the kernel packages guard against that? | 13:09 |
| josch | the answer is a special snippet in /etc/apt/apt.conf.d/01autoremove which special-cases linux-image-* packages | 13:10 |
| josch | you'd have to set up the same for your firmware package | 13:10 |
| + ericsfraga (~user@2.28.19.21) | 13:34 | |
| + chomwitt (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 13:54 | |
| - mtm (QUIT: Ping timeout: 256 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 14:02 | |
| + mark_ (~mjw@gnu.wildebeest.org) | 14:34 | |
| - mark_ (QUIT: Ping timeout: 256 seconds) (~mjw@gnu.wildebeest.org) | 14:49 | |
| - chomwitt (QUIT: Ping timeout: 255 seconds) (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 15:00 | |
| - blast007 (QUIT: Quit: When the pin is pulled, Mr. Grenade is not our friend) (~blast@user/blast007) | 15:06 | |
| + blast007 (~blast@user/blast007) | 15:09 | |
| + chomwitt (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 15:25 | |
| + 047AAXED9 (~Christoph@p4fe73e33.dip0.t-ipconnect.de) | 16:04 | |
| - 047AAXED9 (QUIT: Client Quit) (~Christoph@p4fe73e33.dip0.t-ipconnect.de) | 16:06 | |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 16:08 | |
| - chomwitt (QUIT: Ping timeout: 255 seconds) (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 16:10 | |
| - jacobk (QUIT: Ping timeout: 260 seconds) (~quassel@64.189.201.150) | 17:00 | |
| + jacobk (~quassel@utdpat242053.utdallas.edu) | 17:10 | |
| minute | josch: 01autoremove has a section VersionedKernelPackages that says something about "(out-of-tree) modules" | 18:13 |
| minute | josch: so if i call my module qcacld2-foobar-modules, what will happen? | 18:13 |
| - wakest (QUIT: Ping timeout: 246 seconds) (m-a7d6fe@45.77.48.108) | 18:13 | |
| + wakest (m-a7d6fe@45.77.48.108) | 18:14 | |
| minute | josch: or i just add a file with NeverAutoRemove { "^reform-qcacld2-.*" }; in my meta-package right | 18:15 |
| minute | in any case i understand now that i need a versioned metapackage | 18:15 |
| - jacobk (QUIT: Ping timeout: 268 seconds) (~quassel@utdpat242053.utdallas.edu) | 18:21 | |
| minute | ok, rebuilding those debs... | 18:28 |
| + minutepocket (~reform@p509936c9.dip0.t-ipconnect.de) | 18:30 | |
| minute | minutepocket: https://source.mnt.re/reform/qcacld2/-/jobs/3936/artifacts/browse | 18:30 |
| - Gooberpatrol66 (QUIT: Read error: Connection reset by peer) (~Gooberpat@user/gooberpatrol66) | 18:41 | |
| josch | minute: re VersionedKernelPackages, I cannot tell you unfortunately. This is not something I have done before. | 19:01 |
| + mark_ (~mjw@gnu.wildebeest.org) | 19:07 | |
| - minutepocket (QUIT: Remote host closed the connection) (~reform@p509936c9.dip0.t-ipconnect.de) | 19:28 | |
| + Christoph_ (~Christoph@p4fe73e33.dip0.t-ipconnect.de) | 19:31 | |
| minute | josch: alright! in any case, the overall strategy appears to work | 19:35 |
| minute | josch: the next question is, what's the best way to add these debs to our repo | 19:38 |
| minute | josch: also, here's my way of loading the driver on imx8mp platforms https://source.mnt.re/reform/reform-tools/-/commit/853ea13bba1d95a2034b4aa3dbacb50335e9629d | 19:42 |
| minute | how stable is uname -v as an API? :D | 19:42 |
| josch | minute: don't you want "uname -r"? | 19:45 |
| + chomwitt (~chomwitt@2a02:587:7a17:f700:1ac0:4dff:fedb:a3f1) | 19:45 | |
| minute | josch: do i? is that fine grained enough? | 19:46 |
| * mjw -> Guest2486 | 19:46 | |
| - Guest2486 (QUIT: Killed (zinc.libera.chat (Nickname regained by services))) (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 19:46 | |
| * mark_ -> mjw | 19:46 | |
| minute | perhaps it is | 19:46 |
| + Guest2486 (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 19:46 | |
| minute | right now my deb name is reform-qcacld2-6.7.9-reform2-2+reform20240327T174618Z.deb | 19:47 |
| minute | (and the .ko as well) | 19:47 |
| minute | perhaps the .ko need not be that specifically named? | 19:47 |
| josch | that's the deb name? usually the package name is separated from the version by an underscore | 19:47 |
| minute | good point | 19:47 |
| josch | you can use dpkg-name to auto-rename it | 19:48 |
| minute | here it looks a bit all over the place? https://source.mnt.re/reform/reform-debian-packages/-/jobs/3913/artifacts/browse/repo/pool/main/l/linux/ | 19:48 |
| josch | it does? where? | 19:49 |
| minute | ah hm maybe i need to look closer | 19:49 |
| josch | there are two underscores in every name | 19:49 |
| josch | the first is the package name, then the version, then the architecture | 19:49 |
| minute | > linux-config-6.7_6.7.9-reform2-2+reform20240327T174618Z_arm64.deb | 19:49 |
| minute | i wonder why this has 6.7 also in the name? | 19:49 |
| josch | because you can install linux-config for each kernel version you have | 19:49 |
| minute | mhm | 19:49 |
| josch | that's why the kernel itself also has the version in the name | 19:50 |
| minute | aha | 19:50 |
| josch | for example over here i have a package named linux-image-6.5.0-0.deb12.4-reform2-arm64 | 19:50 |
| josch | with version 6.5.10-1~bpo12+1+reform20231224T144225Z1 | 19:50 |
| josch | as an aside: | 19:50 |
| minute | ok, one really has to have eagle eyes for this :3 | 19:50 |
| josch | the uname -v output might not be good anyways because it seems to be limited to 64 characters | 19:50 |
| josch | so what i get is: | 19:51 |
| josch | #1 SMP Debian 6.5.10-1~bpo12+1+reform20231224T144225Z1 (2023-12- | 19:51 |
| minute | yeah | 19:51 |
| josch | if you need the version of the currently running kernel you can do this: | 19:52 |
| josch | dpkg-query --show --showformat '${Version}' "linux-image-$(uname -r)" | 19:52 |
| minute | ok lets see if i understand correctly | 20:02 |
| minute | josch: my deb should be called: reform-qcacld2-6.7.9-reform2-arm64_6.7.9-reform2-2+reform20240327T174618Z_arm64.deb yes? | 20:02 |
| minute | because it is specifically _for_ 6.7.9-reform2-arm64 | 20:03 |
| minute | and its version is reform2-2+reform20240327T174618Z but it could maybe also just be a date/timestamp? | 20:03 |
| minute | currently i would copy this version from the kernel package's version | 20:03 |
| minute | the version looks like the output of date -u +"%Y%m%dT%H%M%SZ" | 20:08 |
| + gustav28 (~gustav@c-fd68524e.019-141-67626730.bbcust.telenor.se) | 20:13 | |
| minute | new name: > "reform-qcacld2-${KERNELRELEASE}_${QCAPACKAGEVERSION}_arm64.deb" | 20:14 |
| minute | and meta package name: > "reform-qcacld2_${QCAPACKAGEVERSION}_arm64.deb" | 20:14 |
| minute | josch: i'm not sure if your command for getting the currently running kernel's version is correct or if i understand this wrong. on my pc "uname -r" is 6.3.0-1-amd64, but dpkg-query shows 6.3.7-1 | 20:20 |
| + Gooberpatrol66 (~Gooberpat@user/gooberpatrol66) | 20:26 | |
| minute | ok, think i figured it out | 20:30 |
| minute | josch: your query command is fine, thanks | 20:31 |
| minute | package names are now: | 20:41 |
| minute | - reform-qcacld2-6.7.9-reform2-arm64_20240404T183846Z_arm64.deb | 20:41 |
| minute | - reform-qcacld2_20240404T183846Z_arm64.deb | 20:42 |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 21:03 | |
| - colinsane (QUIT: Quit: bye) (~colinunin@97-113-159-4.tukw.qwest.net) | 21:08 | |
| josch | minute: version is the output of this: | 21:10 |
| josch | date --utc ${SOURCE_DATE_EPOCH} +%Y%m%dT%H%M%SZ | 21:10 |
| josch | you can find it in reform-debian-packages:linux/build.sh | 21:11 |
| + colinsane (~colinunin@97-113-159-4.tukw.qwest.net) | 21:11 | |
| gustav28 | hi all! I'm curious about the mnt reform but as I'm not a linux user (macos here), I have some questions. the first one is, how easy is it migrate processor module while keeping ones system disk? say that I set up my OS on a m2 ssd, is it as easy as swapping the cpu board and using another boot-SD-card for the new cpu board? or will linux not start | 21:21 |
| gustav28 | up because it still have the old drivers installed? is there a document with steps one can follow to do the swap? | 21:21 |
| vkoskiv | I did that when I swapped from the original i.mx8 module to the A311D, and it mostly just works. I forgot if there's any docs about that process, I did write some of my observations on the forum. | 21:22 |
| gustav28 | cool. does that work because all drivers are already in the (previous) kernel? or does it download stuff automatically (I find that hard to believe)? | 21:25 |
| gustav28 | *previous=already installed kernel | 21:26 |
| gustav28 | another question: will OpenGL games work out-of-the-box on linux using OpenGL ES 3.1 drivers? or do the games need to be built for OpenGL ES 3.1 to work? I'm a hobbyist game developer and while I could switch to OpenGL ES it would be nice if I didn't have to do much to keep working | 21:28 |
| vkoskiv | In linux, drivers are either built into the kernel, or as loadable modules. I think the system image ships all the drivers for the different SoMs, so those should be okay. | 21:42 |
| vkoskiv | The only thing that requires attention is the bootloader configuration, since that differs between SoMs | 21:42 |
| gustav28 | and the bootloader config usually resides on the sd-card? | 21:45 |
| vkoskiv | Often, yes. On my i.mx8 I had it on the internal EMMC, and that might be possible with some of the other modules too, but on A311D the current recommendation is to stick to SD card for that. | 21:46 |
| vkoskiv | Since oopsies with the internal EMMC can soft-brick the SoM | 21:46 |
| vkoskiv | (requires an adapter board to fix, which I don't have) | 21:46 |
| gustav28 | aha | 22:03 |
| gustav28 | thanks for the info | 22:03 |
| gustav28 | regarding OpenGL, it seems like the drivers are both OpenGL and ES, as the website says "OpenGL/ES" | 22:07 |
| vkoskiv | Yeah, just quite old versions. | 22:08 |
| minute | on a311d with a flag you can get up to OpenGL 3.3 (but no geometry shaders) | 22:10 |
| gustav28 | nice | 22:12 |
| gustav28 | I gtg, but thanks for the insights! | 22:12 |
| vkoskiv | cya | 22:12 |
| - gustav28 (QUIT: Quit: Quit) (~gustav@c-fd68524e.019-141-67626730.bbcust.telenor.se) | 22:15 | |
| minute | josch: ok, this is where it's at so far https://source.mnt.re/reform/qcacld2/-/merge_requests/1/diffs | 22:15 |
| minute | what is SOURCE_DATE_EPOCH? i vaguely remember it is about reproducible builds | 22:16 |
| josch | yes it is | 22:16 |
| josch | when you build the kernel multiple times with the same SOURCE_DATE_EPOCH, you get bit-by-bit identical binaries | 22:16 |
| minute | josch: how could we integrate my qcacld2 binary deb building into reform-debian-packages? (i think it should be there because that's where the kernel is built + reprepro is done?) | 22:17 |
| josch | minute: instead of installing to /etc/modprobe.d, consider installing into /usr/lib/modprobe.d/ | 22:18 |
| minute | josch: ah ok | 22:18 |
| minute | josch: ah damn :D i always make the same mistake again | 22:18 |
| josch | minute: i see two ways: a) let the script which updates the repo on mntre.com download from reform-debian-packages as well as from qcacld2 ci jobs and combine them | 22:18 |
| minute | josch: would this mean running reprepro on the webserver instead of the CI? | 22:19 |
| josch | b) add another few lines to build_custom.sh in reform-debian-packages which either builds the package or downloads it from the other CI job | 22:19 |
| josch | minute: it would mean running it again on the webserver, yes | 22:19 |
| minute | ok. the thing is that the drivers have to be built exactly after reform-debian-packages is done with a new kernel package | 22:20 |
| josch | then maybe you want to move the build process into build_custom.sh in reform-debian-packages | 22:21 |
| minute | yeah, that would make the most sense, also i can probably access the kernel debs then from inside the pipeline? | 22:21 |
| josch | and then in .gitlam-ci.yml make sure that the build_custom job depends on the kernel build job as it requires its artifacts | 22:21 |
| josch | yes | 22:22 |
| josch | as artifacts | 22:22 |
| minute | ok, thanks | 22:22 |
| minute | josch: i have just rebuilt reform-firedecor on my pc against wayfire-dev 0.8.1-1 from debian, and it works fine with debian's wayfire, no "plugin abi mismatch" error | 22:52 |
| minute | so what is going wrong in the reform-debian-packages build | 22:53 |
| minute | josch: ah wait, we are shipping reform-firedecor that was upstreamed to debian by you, right | 22:54 |
| minute | so let me try that on my pc... | 22:55 |
| minute | perhaps the package needs to be rebuilt @ debian somehow? | 22:55 |
| josch | minute: can you show me how to reproduce the issue? can i do that with an sd-card image downloaded from ci? | 22:56 |
| minute | josch: ah, you're running stable on your reform, right? | 22:57 |
| josch | yes | 22:57 |
| minute | josch: you can then download a recent reform-system-image for a311d and try wayfire, the windows have no decorations. if you log the wayfire stderr+stdout manually (i.e. start it from tty1 or so instead of tuigreet), you'll see an error about plugin abi mismatch | 22:58 |
| minute | josch: my guess is that debian updated wayfire and wayfire-dev but that does not trigger a rebuild of reform-firedecor | 22:58 |
| josch | if reform-firedecor has to be rebuilt, then this means that the dependencies of reform-firedecor are too loose | 22:59 |
| josch | it also means that more automated tests have to be added to reform-firedecor to detect this in the future before it happens | 23:00 |
| minute | ok | 23:00 |
| josch | is there something like xephyr for wayland? | 23:00 |
| josch | i.e. automating wayland applications headless | 23:00 |
| minute | i think so, but don't know from the top of my head | 23:00 |
| josch | okay, i'll see what i can find tomorrow | 23:01 |
| minute | > EE 04-04-24 23:02:51.717 - [src/core/plugin-loader.cpp:105] /usr/lib/x86_64-linux-gnu/wayfire/libfiredecor.so: API/ABI version mismatch: Wayfire is 20240311, plugin built with 20231123 | 23:03 |
| minute | i have reproduced it on my pc | 23:03 |
| minute | by just installing reform-firedecor from debian | 23:03 |
| minute | i would like to report this bug but i think reportbug isn't set up correctly on my system (no email from commandline configured) | 23:04 |
| minute | josch: as i'm under pressure to fix this, do you think it's realistic to have this package rebuilt (there are no code changes necessary) at debian relatively quickly? otherwise i have to reintegrate it in reform-debian-packages tomorrow | 23:05 |
| josch | okay, lets try | 23:06 |
| josch | minute: reform-firedecor_2023-10-23-4.dsc uploaded and will be in the archive in ~5 hour | 23:09 |
| minute | josch: awesome, thank you very much! | 23:10 |
| josch | minute: well, lets hope that this was the problem and not something else :) | 23:12 |
| minute | josch: i have high expectations because i cloned our firedecor repo, built against wayfire-dev from debian, installed it and it just worked. | 23:36 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!