| josch | minute: currently reform-system-image:mkimage.sh just downloads the latest reform-setup-wizard artifacts from https://source.mnt.re/reform/mnt-reform-setup-wizard/-/jobs/artifacts/main/raw/target/aarch64-unknown-linux-gnu/release/reform-setup?job=build-cross | 00:03 |
|---|---|---|
| josch | minute: this is not intended, no? this means that the system image also will contain the latest experiments: https://community.mnt.re/t/installer-crashing-instead-on-ssd-step/3893 | 00:03 |
| josch | minute: would you like me to hot-fix this with the reform-setup-wizard version from August, before the SSD installation page was added? | 00:06 |
| minute | josch: oh yes, thanks | 00:06 |
| ch | i think lvfs-testing was updated, so people could try fwupdmgr enable-remote lvfs-testing && fwupdmgr refresh --force and then there should be updates | 00:07 |
| ch | looking at josch (: | 00:07 |
| josch | in about half an hour i can do that! :) | 00:08 |
| josch | ch: Devices with the latest available firmware version: Pocket Reform System Controller 1.0 | 00:14 |
| ch | boo | 00:14 |
| josch | (output of sudo fwupdtool get-updates) | 00:15 |
| ch | hmm, it is in the testing metadata though | 00:16 |
| josch | let me paste what i see here, sec... | 00:16 |
| ch | maybe also paste fwupdmgr refresh --force --verbose | 00:18 |
| josch | ch: https://paste.debian.net/1398424 | 00:21 |
| ch | yeah strange | 00:23 |
| ch | maybe tomorrow then | 00:23 |
| josch | it's like debian, just wait 6 hours :) | 00:24 |
| ch | tbh i dont know how i actually manage to do that | 00:25 |
| ch | but i guess for the important packages i just dput && go to bed | 00:25 |
| josch | it's a good strat | 00:25 |
| josch | minute: could you please enable the gitlab CI runner for https://source.mnt.re/josch/mnt-reform-setup-wizard/ | 00:39 |
| josch | unfortunately the old artifacts in the original mnt-reform-setup-wizard repo all expired | 00:51 |
| josch | there are some old wizard binaries in existing system images but sourcing those would not be ideal | 00:52 |
| + mjw (~mjw@195.23.216.83) | 01:35 | |
| minute | ch: solved the hang | 02:26 |
| minute | ch: it was something very simple | 02:26 |
| minute | ch: it's the sleep_ms(1) | 02:26 |
| minute | ch: obvs one cannot sleep in a timer interrupt m))) | 02:26 |
| minute | this was the main problem https://source.mnt.re/reform/pocket-reform/-/merge_requests/56/diffs#1ca29a0ee9ffa6b8a763636f1f51dc66e4193f7a_559_567 | 02:59 |
| minute | i've also disabled all printfs in spi handler by default... | 02:59 |
| minute | sadly the assert never made it i guess https://github.com/raspberrypi/pico-sdk/issues/108 | 03:06 |
| minute | (but this should have really occured to me) | 03:06 |
| - mjw (QUIT: Ping timeout: 244 seconds) (~mjw@195.23.216.83) | 03:32 | |
| - synnfynn (QUIT: Quit: Lenore..) (~synnfynn@user/synnfynn) | 03:35 | |
| - paperManu (QUIT: Ping timeout: 245 seconds) (~paperManu@64.58.44.160) | 03:48 | |
| - nsc (QUIT: Ping timeout: 250 seconds) (~nicolas@i5C74DF5A.versanet.de) | 03:51 | |
| + nsc (~nicolas@247-96-142-46.pool.kielnet.net) | 03:52 | |
| + paperManu (~paperManu@64.58.44.160) | 03:55 | |
| - op_4 (QUIT: Remote host closed the connection) (~tslil@user/op-4/x-9116473) | 04:05 | |
| + op_4 (~tslil@user/op-4/x-9116473) | 04:05 | |
| - paperManu (QUIT: Ping timeout: 245 seconds) (~paperManu@64.58.44.160) | 04:25 | |
| - sigrid (QUIT: Ping timeout: 260 seconds) (~sigrid@ftrv.se) | 05:04 | |
| - robin_ (QUIT: Quit: Leaving) (~robin@user/terpri) | 05:48 | |
| ch | minute: that... makes sense | 08:33 |
| josch | minute: oh no, when you yesterday said "really gaming now" it was actually playing the firmware game on a saturday night?? XD | 09:26 |
| + reform14647 (~m@ip2504e528.dynamic.kabel-deutschland.de) | 09:45 | |
| reform14647 | minute: i have two ways forward to temporarily patch the reform-setup-wizard situation in the latest system images: | 09:46 |
| + sigrid (~sigrid@ftrv.se) | 09:47 | |
| reform14647 | a) use reform-setup-wizard from unstable -- this works (i'm writing from a setup installed with that right now) and i can merge this right now | 09:47 |
| reform14647 | b) use an older version of reform-setup-wizard from git (before the SSD page was added) but newer than what is in Debian -- for that i need the gitlab CI runner to be enabled for https://source.mnt.re/josch/reform-setup-wizard and/or privileges to create new branches in the reform/reform-setup-wizard repo | 09:48 |
| reform14647 | this is just a stop-gap-measure. I think in the medium term, there should be a new git tag for reform-setup-wizard and mkimage.sh should get the CI artifacts from that or you do reform-setup-wizard development in a branch that is not the master branch. | 09:50 |
| reform14647 | the reason i need the CI runner is that gitlab removed the old artifacts weeks ago but that will not happen if there is either a new tag or a separate branch that mkimage.sh can get the artifacts from | 09:51 |
| + gustav2 (~gustav@c-78-82-55-162.bbcust.telenor.se) | 10:02 | |
| - reform14647 (QUIT: Quit: Leaving) (~m@ip2504e528.dynamic.kabel-deutschland.de) | 11:39 | |
| josch | i now have sysctl and keyboard firmware 20250927 on my pocket reform thanks to ch -- lets see if there are any issues when daily driving it :) | 12:14 |
| ch | \o/ | 12:16 |
| ch | minute: i just discovered, if you have the embargo remote, you can do sth like `fwupdmgr install --allow-older a5a1c402d8e97c20fe267b0050b89dac9edca9b4` and it will prompt for the version you want; from there you can also pick any of the g* versions | 12:16 |
| ch | josch: i think your quectel reset on sysctl update is not normal. at least the wifi card doesnt reset for me. i assume your quectel is in the slot i have the wifi in? | 12:23 |
| ch | josch: ie the other slot is nvme? | 12:23 |
| josch | ch: correct | 12:23 |
| ch | and it happens every time, i guess? | 12:24 |
| josch | yes | 12:24 |
| + plomtest (~plom@user/plomtest) | 12:24 | |
| josch | every time i update the sysctl firmware, yes | 12:24 |
| - plomtest (QUIT: Remote host closed the connection) (~plom@user/plomtest) | 12:25 | |
| ch | ah i see why :( | 12:26 |
| ch | MODEM_PWROFFn doesn't go via the latch | 12:26 |
| josch | --verbose plz | 12:27 |
| ch | so, sysctl has a few pins to control power | 12:30 |
| ch | when the sysctl resets, they reset to something (probably digital inputs), so they are not driven anymore | 12:30 |
| + plomtest (~plom@user/plomtest) | 12:31 | |
| - plomtest (QUIT: Remote host closed the connection) (~plom@user/plomtest) | 12:31 | |
| ch | the signals going to display+som go there via a latch chip (74HC573), so the timing is controlled | 12:31 |
| ch | and also that avoids an accidental reset of these signals | 12:31 |
| ch | but the two signals going to the modem M.2 go direct | 12:31 |
| josch | ooooh | 12:32 |
| ch | so the quectel will see the reset and indeed reset | 12:32 |
| ch | and i assume the wifi cards just ignore these lines | 12:32 |
| josch | ah well, i mean how often do i update the sysctl firmware :) | 12:32 |
| ch | tbh i'd be quite annoyed ;) | 12:33 |
| ch | good thing i didnt put the quectel in yet | 12:33 |
| josch | compared to the other issues like sysctl reset on battery, brownouts, keyboard resets etc, this is really not an issue XD | 12:33 |
| josch | (imho) | 12:33 |
| josch | i mean it comes right back up | 12:33 |
| ch | yeah if you don't flash the firmware all day long | 12:33 |
| josch | i also have the wifi board on order so i'll not rely on LTE in the future :) | 12:34 |
| josch | (and i always could have ethernet plugged in if i wanted) | 12:34 |
| ch | the ethernet might also need a good look one day (: | 12:35 |
| josch | what's wrong with ethernet? | 12:35 |
| ch | i had some trouble with it the other day, but didnt spend time on it | 12:35 |
| josch | oh, i didn't have any so far | 12:35 |
| ch | https://paste.debian.net/hidden/12f106ef/ 'now which version is better' | 12:36 |
| ch | (probably neither) | 12:37 |
| ch | i can confirm the sysctl doesnt crash anymore on shutdown, at least not in a way i can notice | 12:47 |
| ch | and now i just got a linux stacktrace which i think means the mmc failed to init :O | 12:48 |
| ch | also did booting to some visible display output get slower at some point? | 12:48 |
| josch | good question and there are other things where i feel there are regressions but i cannot test because we no longer have the artifacts to test this around and rebuilding stuff is time-consuming | 13:00 |
| josch | i am contemplating to press the button which preserves artifacts more regularly for reform-system-image and/or reform-debian packages | 13:01 |
| josch | like, maybe every month or every two months? | 13:01 |
| josch | for example, 1080p playback is maxing out all my cpu cores on a311d with mpv and lots of frames are dropped. Was this always like that? I remember differently but i cannot easily test the state from one year ago... :/ | 13:02 |
| - casparvitch (QUIT: Remote host closed the connection) (~casparvit@user/casparvitch) | 13:13 | |
| + casparvitch (~casparvit@user/casparvitch) | 13:13 | |
| + plomtest (~plom@user/plomtest) | 13:29 | |
| ch | i asked hughsie on fedi about the 502 | 13:32 |
| minute | ch: i think yours is the better way (the one that reads only if the write worked), but afaik you can't check for errors like that, because > Number of bytes written, or PICO_ERROR_GENERIC if address not acknowledged, no device present, or PICO_ERROR_TIMEOUT if a timeout occurred | 13:51 |
| ch | yeah | 13:52 |
| ch | in my branch i have >= 0 now | 13:52 |
| minute | ah good | 13:52 |
| minute | shouldn't it be >= 1? :D | 13:52 |
| ch | uh | 13:52 |
| ch | i'll make a note | 13:52 |
| minute | or == 1 | 13:52 |
| minute | because you want to write 1 byte there | 13:52 |
| + paperManu (~paperManu@64.58.44.160) | 14:02 | |
| minute | fascinating https://blogs.gnome.org/gtk/2025/09/15/casilda-1-0-released/ | 14:34 |
| josch | minute: any thoughts on how i should proceed with fixing the reform-setup-wizard? see my messages up in the backlog. I'm currently blocked by gitlab CI runner not being enabled on https://source.mnt.re/josch/mnt-reform-setup-wizard/ | 14:34 |
| minute | josch: ah yes, sorry. i can enable the runner in 5 minutes. | 14:35 |
| josch | minute: i don't want to be pushy, it's no problem if this doesn't get fixed today (it's sunday after all) but just giving you the option of acting if you would like :) | 14:35 |
| minute | josch: no problem, it's good to chase me sometimes :3 | 14:40 |
| josch | okay, thank you! <3 | 14:41 |
| josch | (same here, please ping me as often as needed) | 14:41 |
| josch | (my work is very much interrupt driven, unfortunately...) | 14:41 |
| minute | josch: runner assigned! | 14:43 |
| josch | okay, great! | 14:43 |
| josch | then i'll temporarily use the build artifacts from reform-setup-wizard/josch for the system image | 14:43 |
| josch | (until the main repo is fixed) | 14:44 |
| minute | that's fine! | 14:46 |
| josch | minute: do you have time/headspace to talk about a more sustainable solution for the future? | 14:46 |
| josch | maybe git tag more often and then retrieve the CI artifacts from the tag? | 14:47 |
| josch | or work with branches instead? | 14:47 |
| minute | josch: i'm in a hurry r/n but system image should rather pull a known good tag from setup wizard yes. | 14:48 |
| minute | ideally possible to override via CI variable | 14:48 |
| minute | (for test builds) | 14:48 |
| josch | oh no, building reform-setup-wizard fails because rust crates like serde fail to build right now in unstable... :( | 14:49 |
| josch | i'll investigate | 14:49 |
| josch | problem fixed (will prepare a MR for reform-setup-wizard in a bit) and i'm now flashing the image created by https://source.mnt.re/reform/reform-system-image/-/merge_requests/140 | 15:33 |
| minute | nice | 15:40 |
| josch | is this loading for anybody? https://source.mnt.re/reform/reform-system-image/-/jobs/13401/raw | 15:55 |
| minute | no not really | 15:55 |
| josch | oh no :( | 15:55 |
| josch | was there a gitlab upgrade or is the css just not loading in? the interface looks different | 16:11 |
| josch | ah it works now | 16:12 |
| josch | could this be connected with me having deleted 20 GiB of old artifacts? | 16:12 |
| ch | probably | 16:12 |
| ch | that happens on a background worker | 16:12 |
| ch | maybe you have now io-starved something relevant :D | 16:13 |
| ch | or the job log was part of the 20gb | 16:13 |
| ch | seems to load for 13402 | 16:13 |
| josch | minute: to help with such problems in the future, what do you think of (by default) only produce *all* system images when reform-system-image builds for the main branch in the "reform" namespace and otherwise default to only building the reform-system-any image? | 16:17 |
| josch | for example, before i deleted one artifact worth 20 GiB there were five times 20 GiB artifacts for each commit I did to the pipewire-pulse fixup branch from the day before yesterday | 16:18 |
| ch | at least the MR build should probably do a full build, otherwise you dont know ahead of merging if it will build? | 16:27 |
| ch | (maybe i'm missing something obvious) | 16:28 |
| ch | # [battery] [ERROR] gauge did not respond | 16:28 |
| ch | ^ random observation | 16:28 |
| josch | ch: very rarely are the bits touched which would make the non-any images not build | 16:29 |
| josch | but committing five times to a feature branch is "normal" i'd say and it shouldn't happen that doing so kills future CI builds without a good error message | 16:30 |
| ch | +1 | 16:30 |
| ch | maybe its also fine if the merged build breaks, after all it can be fixed | 16:31 |
| ch | (sometimes i forget we're not in an enterpriseland) | 16:31 |
| josch | it's all not optimal but 20 GiB is a lot of data, so i'm also not too upset about this not being all smooth. I think it broke now because we had the unusual combination of the pipewire-pulse fixes 2 days ago, the reform-setup-wizard fixes yesterday and today and then today the weekly rebuild was scheduled on top of all that in a relatively short timeframe | 16:33 |
| ch | somewhat tempted to cut the ixeth cable short, but i'll suck at putting a new rj45 on | 16:44 |
| ch | josch: does the pocket eth link at 1gb/s for you? | 16:46 |
| ch | maybe my cable is broken | 16:46 |
| josch | ch: i think it did a few weeks ago but i cannot try right now because i do not have a gigabit ethernet switch at home | 16:48 |
| ch | hm, replugging gives 1gbit/s | 16:58 |
| ch | lets ignore this | 16:58 |
| ch | also this advice from 2020 is still true: https://www.linuxquestions.org/questions/linux-embedded-and-single-board-computer-78/networkmanager-and-ipv4ll-link-local-4175672529/#post6108347 | 16:59 |
| * gachikuku_ -> gachikuku | 17:24 | |
| ch | sigh, making the sysctl crash on startup is not great | 17:57 |
| + jogu (~jogu@user/jogu) | 17:58 | |
| ch | one of these moments where i'm guessing "wouldnt have happened with rust" | 18:17 |
| + erle (~erle@user/erle) | 18:23 | |
| ch | maybe i solder something to TP5,6 | 18:40 |
| minute | ch: how did you do it? | 18:42 |
| minute | or... what's cooking? :D | 18:42 |
| ch | i tried adding a second usb class driver | 18:46 |
| ch | but tinyusb crashes on me | 18:46 |
| ch | what i really want is to have persisted feature flags on the sysctl, so i can stop having a custom firmware build | 18:47 |
| ch | but then i need some way of setting those, thus usb interface | 18:47 |
| ch | i'm hoping if i get swd to work i can step through tinyusb init | 18:47 |
| ch | GND on green wire #electricalcrimes | 19:20 |
| ch | [rp2040.core0] halted due to breakpoint, current mode: Handler HardFault | 19:29 |
| ch | promising | 19:29 |
| + MartiniMoe (~moe@p200300eecf1ede3ccabdf6241f8d413a.dip0.t-ipconnect.de) | 19:31 | |
| - GNUmoon2 (QUIT: Remote host closed the connection) (~GNUmoon@gateway/tor-sasl/gnumoon) | 19:33 | |
| + GNUmoon2 (~GNUmoon@gateway/tor-sasl/gnumoon) | 19:33 | |
| + AnimaInvicta (~AnimaInvi@88-120-179-216.subs.proxad.net) | 19:51 | |
| - jogu (QUIT: Remote host closed the connection) (~jogu@user/jogu) | 19:55 | |
| ^alex | ch, oh we love "current mode: Handler HardFault" | 20:08 |
| ^alex | also IMO instead of adding another USB interface to the syscon, you could extend the serial parser | 20:09 |
| ch | ^alex: do you happen to know what the actual problem is here: https://source.mnt.re/reform/pocket-reform/-/merge_requests/60 | 20:23 |
| ch | on second thought, i guess inline without static is really dangerous there | 20:25 |
| + synnfynn (~synnfynn@user/synnfynn) | 20:55 | |
| + mjw (~mjw@195.23.216.83) | 21:42 | |
| - synnfynn (QUIT: Quit: ..) (~synnfynn@user/synnfynn) | 22:01 | |
| - gustav2 (QUIT: Quit: Quit) (~gustav@c-78-82-55-162.bbcust.telenor.se) | 22:15 | |
| * nyana -> jana | 22:23 | |
| * jana -> janaa | 22:23 | |
| - MartiniMoe (QUIT: Remote host closed the connection) (~moe@p200300eecf1ede3ccabdf6241f8d413a.dip0.t-ipconnect.de) | 22:30 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!