2026-01-18.log

joschgordon1: yes, cathode and ground are connected00:02
gordon1josch: yeah that looks like a short in the ltc6803 then, doesn't it?00:04
+ chomwitt (~chomwitt@2a02:85f:9a0a:f200:1ac0:4dff:fedb:a3f1)00:04
joschrwa_: to rebuild the dts you can just run "make rockchip/rk3588-mnt-reform2-dsi.dtb" after copying the dtb into your kernel source and adding the line building it to the relevant makefile00:04
joschgordon1: :(00:04
joschrwa_: you probably want to revert 57f899539491da628816353c1e1bb1f37e03fad3?00:04
gordon1josch: but why does the mb work from battery then?00:05
gordon1clearly not understanding something00:05
joschi hope i didn't dream that bit00:05
gordon1oh wait duh00:05
gordon1because diode is blown00:05
gordon1yeah so likely ltc6803 is bust then00:06
joschgreat, yeah i cannot repair that :/00:06
gordon1but can you double check cell voltages, if it is blown some cells might be shorted or out of balance00:06
joschyes, i just checked them they are all at 3.35 V00:06
gordon1good00:07
gordon1josch: not to be pushy, but general gist is the same as with removing the diode - big blob around all the pins of the IC will desolder it, however i strongly recommend low temp chipquik in this case00:08
gordon1but i would like second opinion on my diagnosis first (minute ?)00:09
joscheven if we know that the ltc is bust, i'll not attempt swapping it myself -- the board is too expensive for me to accidentally rip a pad off00:10
gordon1fair enough00:11
joschbut i'm happy that swapping the diode actually worked better than i thought with both your tips :)00:12
joschnever swapped a SMD component before00:13
gordon1maybe soon you'll be more confident to swap TSSOP/TQFPS then :)00:13
joschhehe00:13
gordon1i still would claim it is doable with soldering iron, however hot air will make it much easier00:14
joschi can ask the local hacker space for people who have experience with this00:15
gordon1oh, wait00:16
gordon1nah, nvm00:17
joschuh the chip is not very cheap00:17
* mark_ -> mjw00:17
gordon1i think there is no reason why C9-C12 pins on LTC should be shorted to ground, and there is nothing else connected to those but the cathod of D3000:18
joschthere is no rush with fixing this -- i'm back on my old a311d :)00:18
joschi probably just rushed too much mysel00:19
joschi was experimenting with fitting the wifi/bt board of the pocket in the classic and that must've shorted something unintentionally00:19
gordon1you know what puzzles me the most in this case? why D30 behaved like a fuse before F2 did?00:20
gordon1twice00:21
^alexdevice sacrifices itself to save fuse00:22
joschnoble :)00:22
^alexmr. murphy strikes again00:23
gordon1tbh i wouldn't be surprised that this diode is cheaper than the fuse00:23
josch(but much harder to replace)00:24
joschokay, seems i can get the chip for 40 eur but i'll wait for input from minute 00:25
joschi mean maybe they also have capacity at MNT for me paying them to carry out the repair, would probably be the safest option00:25
gordon1ACTION added some BAT46WJ,115 in his mouser cart just in case00:27
joschhehe00:27
joschi also have more than a dozen leftover which i can mail to anybody in this chat for free if somebody should be in need :)00:27
minutejosch: mnt has repair service00:29
gordon1wow yeah ok LTC6803 is far from being cheap00:29
minuteyeah the chips on classic are incredibly expensive00:29
minuteluxury computer00:29
gordon1it is a cool chip tho00:29
gordon1my oscilloscope has less bits that that00:30
minuteyeah, linear/analog stuff is quite fancy (also quite weird)00:30
minute"turn your reform into a hi res 8-channel ADC"?00:31
joschminute: it does but if you are already stacked full with tasks then i can also live with a board that cannot charge batteries and put it into desktop reform or something00:31
minutejosch: i think wakest can do the repair (with a bit of my guidance). it just depends on how quickly do you need it00:32
minute(i'm not even in berlin atm)00:32
joschokay, i'll write a mail (and no urgency, i'm back at a311d right now :D)00:33
minutejosch: i hope by mail you mean support.mnt.re!00:34
joschyes, i don't think there is real mail anymore, right?00:34
joschfor support i mean00:34
minutenot one that is public knowledge at least ^^ (because of spam etc)00:35
minutejosch: i also need to finally fix your ls1028a, argh!00:35
gordon1so all my fancy trox screws are stainless ones, not really magnetic, so it is just a matter of time when i'll drop one right on the board and short something, better to stock up!00:35
joschminute: no worries -- a debian developer has borrowed me his for the time being00:35
minutegordon1: ugh! no carbon torx avail?00:36
minutejosch: ok cool00:36
gordon1maybe there is, it's just ones i had00:36
gordon1probably need to change those again00:36
gordon1these ones are also very quick to wear the blackening out00:37
joschminute: seeing the magic you do on the daily i feel not bad at all that you priotize that over the ls1028a of mine :)00:39
joschgordon1: oh good you remind me -- i should swap out the torx screws before sending this unit in to not cause extra work XD00:40
josch(but i need to stack up on torx as i really like them and now i need to get both 5 and 6 mm lengths to also replace those in the display)00:42
gordon1yeah, those probably good idea to replace with liberal amount of loctite00:45
gordon1those will work themselves loose otherwise00:45
joschthey absolutely will00:48
joschmy wife found M2 screws in the milk of her cereal one day00:48
joschthey are everywhere by now00:48
gordon1oh i thought you're talking about big m4s00:48
gordon1but yes00:49
joschoh00:49
joschno, the big M4s never worked themselves loose for me00:49
joschand that is after having opened and closed this thing multiple times a day for nearly five years :D00:50
josch(but the hinges wil die)00:50
gordon1probably because they threadlocked from factory00:50
joschthe original ones came with some white gunk, yes. My new rk3588-dsi reform didn't have that though.00:51
josch(i swapped displays two days ago)00:51
- voltaire28_ (QUIT: Ping timeout: 240 seconds) (~jlafon@28.162.2.93.rev.sfr.net)01:00
joschokay, juuuuust to be sure i did a couple of cold boots with u-boot 2026-01-11 on a311d and it seems to find nvme just fine01:13
- cow321 (QUIT: Ping timeout: 265 seconds) (~deflated8@user/meow/deflated8837)01:39
joschrwa_: i reverted 57f899539491da628816353c1e1bb1f37e03fad3 for you in this MR: https://source.mnt.re/reform/reform-debian-packages/-/merge_requests/161 Once it's done building, you could try running the kernel from the repository the gitlab CI pipeline creates.01:47
- johl (QUIT: Quit: Lost terminal) (~johl@wikidata/Jens-Ohlig)01:51
+ cow321 (~deflated8@user/meow/deflated8837)02:09
- AnimaInvicta (PART: !!unknown attribute: msg!!) (~AnimaInvi@88-120-179-216.subs.proxad.net)02:21
- cow321 (QUIT: Read error: Connection reset by peer) (~deflated8@user/meow/deflated8837)02:46
+ cow321 (~deflated8@user/meow/deflated8837)02:49
+ bkeys (~Thunderbi@98.19.128.69)03:11
- bkeys (QUIT: Quit: With every step we take, danger will follow closely) (~Thunderbi@98.19.128.69)03:25
+ bkeys (~Thunderbi@h69.128.19.98.dynamic.ip.windstream.net)03:25
- chomwitt (QUIT: Ping timeout: 246 seconds) (~chomwitt@2a02:85f:9a0a:f200:1ac0:4dff:fedb:a3f1)03:39
- paperManu (QUIT: Ping timeout: 256 seconds) (~paperManu@146.71.9.156)03:44
+ siviq (~siviq@user/siviq)03:49
- siviq (QUIT: Quit: Client closed) (~siviq@user/siviq)03:59
- 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
- lidstah (QUIT: Remote host closed the connection) (~lidstah@gateway/tor-sasl/lidstah)06:30
+ lidstah (~lidstah@gateway/tor-sasl/lidstah)06:33
- ohnoitsnoah_ (QUIT: Quit: Konversation terminated!) (~ohnoitsno@23.112.150.169)07:41
+ pickybear (~pickybear@user/pickybear)07:44
- pickybear (QUIT: Ping timeout: 240 seconds) (~pickybear@user/pickybear)08:15
+ voltaire28_ (~jlafon@28.162.2.93.rev.sfr.net)08:31
- voltaire28_ (QUIT: Ping timeout: 264 seconds) (~jlafon@28.162.2.93.rev.sfr.net)08:36
rwa_josch: i added the repo to my sources, but when i want to install kernel 6.18.5 it wants to remove flash-kernel - that doesn't seem right10:11
joschrwa_: yes it does because you need a more recent flash-kernel version10:15
joschrwa_: but it should upgrade it, not remove it10:15
joschokay, i just verified: linux 6.18 has Breaks: flash-kernel (<< 3.110~)10:23
joschthis is important because flash-kernel before 3.110 does not support linux 6.1810:23
joschbut the MNT debian repository has flash-kernel version 3.110+reform20260112T221244Z+1 which is greater than 3.110~, so that should work10:24
+ voltaire28_ (~jlafon@28.162.2.93.rev.sfr.net)10:25
rwa_ok, that explains why it tries to remove it, it can't find a newer version in the repos that i have in my sources10:39
joschattention, there is something broken with gdm right now:11:01
joschhttps://community.mnt.re/t/attention-when-upgrading-only-black-screen-with-cursor-instead-of-gdm/415011:01
- voltaire28_ (QUIT: Ping timeout: 244 seconds) (~jlafon@28.162.2.93.rev.sfr.net)11:29
- gianmarcogg03 (QUIT: Quit: https://quassel-irc.org - La chat diventa comoda. Ovunque.) (~quassel@user/gianmarcogg03)11:48
+ gianmarcogg03 (~quassel@user/gianmarcogg03)11:48
+ AnimaInvicta (~AnimaInvi@88-120-179-216.subs.proxad.net)12:11
+ siviq (~siviq@user/siviq)12:30
+ johl (~johl@wikidata/Jens-Ohlig)12:50
johlHello, friends. Since I last updated the packages with aptitude, gdm (I guess?) won't start and I cannot log in. I tried to get to a console login with HYPER+ALT+3, but no luck. What can I do? 12:56
johl(mnt pocket reform machine, btw ;)12:58
joschjohl: you might be affected by the post which I just created (link above your post)12:59
siviqIt's a known issue, see josch https://community.mnt.re/t/attention-when-upgrading-only-black-screen-with-cursor-instead-of-gdm/4150. "Update 1: Pressing ctrl+alt+F5 (or F2, F3, F4, …) gets you to a textual login prompt. You can start sway from there or continue working on the text terminal. This is how you’ll be able to fix this situation once12:59
siviqa solution is known without resorting to booting from an sd-card."12:59
joschIf somebody else could look into this i would be extremely grateful. I have to take care of the children today...13:01
johlF5 would be HYPER+5, correct?13:01
joschyes13:01
johljosch: Thank you! That sounds exactly like my problem13:02
johlctrl+alt+hyper+5 won't give me a text console13:03
johlsiviq: Any ideas what I could try?13:04
joschi think you have to press hyper first or something -- there was a bug in the second layer in the kb firmware13:04
josch(or attach another keyboard via usb)13:04
siviqjohl wish I could be of help, but have no unit yet. It's in production. Been reading here and the community, how I learned about the upgrade issue.13:04
+ paperManu (~paperManu@146.71.9.156)13:05
minutejohl: josch: yeah it was something unintuitive like having to press hyper _last_13:06
minutejohl: josch: i think hyper changes alt to altgr or sth13:06
johlI think I tried all the combinations now of what keys to press in which order now :)13:07
minutejohl: ok i will double check how to do it13:10
minuteone moment...13:10
minuteah on the fw on my keyboard this is already fixed "unfortunately"13:12
minutebut i had written the trick down somewhere13:12
minuteyep13:13
minutejohl: josch: https://community.mnt.re/t/alt-f1-doesnt-switch-to-console/2304/413:13
minuteah, sorry, trick is not written down there13:13
minutebut IIRC it was: press and hold ctrl+alt+5 first, then tap hyper while holding those 313:14
joschinteresting, i have stock firmware and i can switch the tty by first pressing hyper, holding it and then ctrl+alt+513:15
joschtapping hyper last also works13:16
josch(but creates a bunch of ^[5 while holding ctrl+alt+5 of course)13:16
minutejosch: weird/funny btw, i also immediately had LLM vibes here https://community.mnt.re/t/keyboard-issue-on-boot/3997/1413:21
minutejosch: i think fixed firmware was officially released, but it's possible johl didn't update to it yet. @ johl what are the 2 dates displayed when you go to "system status" in the OLED menu?13:22
joschminute: i tried to formulate my reply in a way which doesn't accuse anybody but makes a general statement. Humans are of course perfectly capable to write in a style which sounds like what LLMs spit out.13:24
johl2025062313:24
josch20251118 here13:24
minuteah yes, that probably makes the difference13:25
minutejosch: you got an external usb keyboard you could try?13:25
minuteah13:25
minutesorry13:25
minutejohl: you got an external keyboard you could try?13:25
johlNot with me, but I can ask around to get one. 13:26
minutejohl: sorry for these issues!13:27
joschi'm trying to get to the bottom of this by bisecting the system image... lets see...13:27
johlBut also, once I get to text console there isn't much I can do to fix it, right? Sounds like a Sunday filled with non-computer things and keeping an eye on the forum for an update :)13:28
joschjohl: you cannot fix it (yet) but at least you can start sway and start using your computer normally13:28
johlWell, that's something.13:32
johlHolding ctl+alt+5 and then going tappity tap on hyper won't do anything, for what it's worth. Time to organize an external USB keyboard, I guess13:34
joschoh drats :(13:35
johlIt is what it is.13:42
+ pickybear (~pickybear@user/pickybear)14:42
- pickybear (QUIT: Client Quit) (~pickybear@user/pickybear)14:44
+ voltaire28_ (~jlafon@28.162.2.93.rev.sfr.net)14:59
+ chomwitt (~chomwitt@2a02:85f:9a0a:f200:1ac0:4dff:fedb:a3f1)15:01
- robin (QUIT: Quit: Leaving) (~robin@user/terpri)15:13
- voltaire28_ (QUIT: Ping timeout: 240 seconds) (~jlafon@28.162.2.93.rev.sfr.net)17:11
- johl (QUIT: Ping timeout: 265 seconds) (~johl@wikidata/Jens-Ohlig)17:15
+ nsc (~nicolas@i5C74DCE4.versanet.de)17:16
+ voltaire28_ (~jlafon@28.162.2.93.rev.sfr.net)17:26
- pomel0 (QUIT: Ping timeout: 240 seconds) (~pomel0@user/pomel0)17:28
+ pomel0 (~pomel0@user/pomel0)17:29
+ johl (~johl@wikidata/Jens-Ohlig)18:06
johlExternal keyboard acquired and connected. Unfortunately, this doesn't change a thing. :(18:08
- johl (QUIT: Ping timeout: 244 seconds) (~johl@wikidata/Jens-Ohlig)18:14
minuteuff18:33
siviqSome keyboards have by default media keys active instead of function keys. If you haven't returned it and it's this type, try making the function buttons active by either holding fn key and then press f1 - f5 or better, activate function buttons first. Then try switching to other tty. Hope it helps.18:39
minutei wonder if soon we need some "repair" mode in system image that fixes a borked system like that18:51
minutelike microsoft always had :D18:51
minutealso why is gdm, as such a critical piece of software, _so bad_ at failing gracefully18:51
minutei'm annoyed that gnome is bound to gdm3 like this when gdm3 is so badly engineered18:52
minuteor we need to switch from unstable to testing at some point...18:59
siviqHaving restore point prior system update would be very, very good improvement. That is on my list for research next on the MNT community, what people have discussed so far and come up with.19:00
minutesiviq: right, maybe systemd has a solution for this?19:00
minutewe could also introduce something like holding a key at startup to disable the login manager and go straight to normal console login.19:01
minutedisable the display manager i mean19:01
siviqPossibly. I'm coming from different background (networking) and Linux is not my best skill.19:01
siviqminute Or some kind if pin hole button for emergency boot into console.19:02
minutewell, keyboard combo would be easier to do because that already exists :D19:03
siviqWell, you know your hw19:03
siviq:)19:03
- voltaire28_ (QUIT: Ping timeout: 240 seconds) (~jlafon@28.162.2.93.rev.sfr.net)19:05
minuteah, normally on PCs rescue mode and stuff like this is handled by the boot manager, like grub19:06
minuteand we don't have graphical uboot still ;///19:06
minutebut i think i could make a systemd service that would detect a certain key combo and react to it by disabling gdm19:07
minutenaja. maybe not while i'm in surgery recovery in another country19:08
siviqIt is a good point. From what I've gathered so far, problem like this in the past have lead to improvements of the MNT Reform system. Speedy recovery! minute19:12
minuteyes, it is always very painful for me when our users run into big issues like this19:13
minutethinking a bit more, doing any kind of key sniffing in a global userland service could introduce new security issues. but what we could do is use OLED menu and the reform2-lpc sysfs, or another signal/flag in the usb data of the keyboard. like a virtual LED that can be checked in sysfs19:17
minutesysrq + g also looks interesting19:18
minute> When using Kernel Mode Setting, switch to the kernel's framebuffer console19:18
- pomel0 (QUIT: Ping timeout: 246 seconds) (~pomel0@user/pomel0)19:19
+ pomel0 (~pomel0@user/pomel0)19:19
minutealso > The hangwatch[17] daemon will invoke pre-configured SysRq triggers when system load average exceeds a certain threshold19:20
minutelol ok? > Ctrl+Alt+⇧ Shift+Esc: Introduced in systemd v257 as a more reliable SAK for Linux distributions but may require support from the currently running desktop environment.[3] It is implemented as of gdm (GNOME display manager) 4719:21
minutethat indeed switches to the login screen on gnome for me19:22
minutesession continues to run in the background though19:22
minuteso it's a bit like screen lock19:22
siviqThere you go :) Also making use of the OLED is actually awesome idea. It could have option in there to boot into 'recovery' mode.19:23
minuteyeah19:25
joschminute: systems like nixos can do rollback, maybe that could/should be investigated19:33
joschthough i fear that nixos also requires early display support to choose an earlier system state?19:50
theesm1josch: choosing earlier states happens via bootloader entries in guix at least, i think it's the same for nixos (so display support is somewhat required)19:59
cararemixedOh yeah. I was offline earlier and had rebooted. gdm broke for me so I turned autologin on. Then switched to ly and disabled gdm (this isn’t the first time gdm has broken for me from a simple update).20:15
cararemixedOf course now emacs core dumps on me. At least it’s not my primary editor but I think my install is having issues 🫠. Had an audio crash, a volume go read only (no corruption found so far), and a keyboard give me those persistent T errors on the display where resets don’t fix it w/o reboot.20:17
cararemixedI was going to check in here but ended up needing to do a doctors appointment and a late night pharmacy run…20:18
- elektron (QUIT: Ping timeout: 240 seconds) (~elektron@apoc.halo.nu)20:20
+ elektron (~elektron@apoc.halo.nu)20:20
cararemixedIt might be time to double check my backup and start fresh again.20:21
cararemixedapt is so slow to work with on my home wifi and the old wifi card. It loves to hang on bug report downloads for 30-40mins at a time.20:23
- AnimaInvicta (QUIT: Ping timeout: 255 seconds) (~AnimaInvi@88-120-179-216.subs.proxad.net)20:49
+ AnimaInvicta (~AnimaInvi@88-120-179-216.subs.proxad.net)20:51
+ voltaire28_ (~jlafon@28.162.2.93.rev.sfr.net)20:54
joschcararemixed: how did you turn autologin on? Did you boot from an sd-card to rescue your system on nvme/emmc?21:08
cararemixedUsed tty3 to sign in.21:09
cararemixedIn the past I could ssh from another machine if I was on wifi21:09
cararemixedAlso not on sd. Using the nvme drive21:10
cararemixedAutomatic sessions are useful since the drive requires a passcode to boot anyway21:10
joschcararemixed: so in contrast to johl earlier, ctrl+alt+f3 worked for you?21:10
cararemixedBut with ly its back to pam auth21:10
cararemixedYeah. I think I might have fiddled with systemctl to explicitly link the service21:11
cararemixedI forgot but it was a pain in the past when it didn’t work and I believe that’s when I tried to fix it.21:12
cararemixedOf course right now my system is super unstable. I was going to wait till I had the V2 core shipped to rebuild the install but I am no longer sure what might be my issue or a reform configuration issue21:13
joschokay21:14
joschfor what it's worth, switching ttys with ctrl+alt+fX should be enabled by default21:14
cararemixedit definitely didn’t used to work on the system images I downloaded. It could have been a keyboard layer issue.21:15
cararemixedBut I tried many times and failed21:15
cararemixedI have keyboard stuff I want to change so I might bump looking at that firmware up on my list.21:17
joschcararemixed: ah you had a pocket reform not a classic reform, right?21:22
animoneI just happened to run an apt upgrade last night on my pocket reform, and got hit with the latest gdm bug.. no big deal though, managed to ssh in and disable gdm for now21:26
cararemixedjosch: yes21:26
animonealso to file under neat pocket tricks, I recently noticed the UI is set to 2x in sway.. setting to 1.5 or 1.75 gives me way more screen realestate as expense of readability21:27
joschanimone: it used to be like that but we disabled fractional scaling for performance reasons21:29
joschhttps://source.mnt.re/reform/reform-tools/-/merge_requests/13121:29
joschminute will probably have more details21:29
animoneso with the scale set higher, it's not actually driving the display at the higher res?21:29
josch(note to self to write more elaborate commit messages)21:30
animoneI usually just change it with wlr-randr when I need more screen space21:31
animoneand yeah the performance difference is noticeable sometimes, especially since I'm still on the imx8mp21:32
+ vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50)21:36
vagrantcoh wow, u-boot 2026.01 includes mnt-reform2-rk3588 ... nice!21:38
vagrantcsurprised i missed that, but yay.21:38
cararemixedjosch: I'm creating a new boot sd card, is there a more recent image you'd recommend or should I go to the last one listed here: https://source.mnt.re/reform/pocket-reform/-/releases21:44
cararemixedSeems a bit old so it will take a bit of extra time to do the updates.21:44
- voltaire28_ (QUIT: Ping timeout: 240 seconds) (~jlafon@28.162.2.93.rev.sfr.net)21:54
vagrantcanyone know if the upstream u-boot 2026.01 mnt-reform2-rk3588 works well ... is missing anything major?22:10
joschvagrantc: you have bluerise to thank for that :)22:29
joschcararemixed: i maintain a list of archived artifacts in the README here: https://source.mnt.re/reform/reform-system-image/#old-image-downloads22:30
josch2025-11-05 should be the last one working22:30
cararemixedty Unclear since the pocket docs link to the other22:31
joschoh actually january 15 should also still be fine: https://source.mnt.re/reform/reform-system-image/-/jobs/1787822:31
josch(i'm currenty bisecting system images and 2026-01-16T00:00:00 still works)22:32
+ voltaire28_ (~jlafon@28.162.2.93.rev.sfr.net)22:37
vagrantcare the reform images still using an older u-boot, or are they using 2026.01? :)22:54
- bkeys (QUIT: Ping timeout: 260 seconds) (~Thunderbi@h69.128.19.98.dynamic.ip.windstream.net)22:56
vagrantcwait, is bluerise Peter Robinson <pbrobinson@gmail.com> ?22:56
joschvagrantc: if you are interested in using newer u-boot please help fix the current issue with rk3588 u-boot where the latest version makes linux not recognize the nvme at all22:57
vagrantcoh, that sounds quite bad22:58
joschor the current gdm mess22:58
joschi'd love to spend my time on other things but software keeps breaking...22:58
vagrantci don't have high hopes, but might poke at the u-boot stuff22:59
vagrantci will need to refresh my memory about the various options to recover if u-boot loaded off of eMMC turns out to load, but not actually boot ...23:01
vagrantcas the boot order prefers eMMC, if i recall...23:01
joschit's just so many moving parts... just moving from one dinstall to the next upgrades over 100 packages...23:01
vagrantcjosch: there's this cool site that focuses on reform with debian stable :)23:02
joschvagrantc: with rk3588 u-boot will always be loaded from emmc unless you zero that area, then it tries to load it from sd-card which is why i have emmc zeroed and do experiments on the sd-card23:02
+ ohnoitsnoah_ (~ohnoitsno@72.181.111.171)23:03
joschcorrection, over 230 packages for one dinstall...23:03
joschvagrantc: i know but debian stable will not be stable if nobody runs unstable and finds all the bugs...23:03
vagrantcjosch: i know, i know :)23:03
vagrantcjosch: but if the churn is getting to you, sometimes good to poke at some other bits for a spell :)23:04
joschi fear nobody else is going to fix this though XD23:04
vagrantcfears are sometimes unfounded, and if not, it will be waiting for you another day23:04
vagrantcbut do as you care to :)23:05
vagrantcthanks for all the work you put into it :)23:05
joschthank you, i still got a little bit of free time to spend on this today :)23:05
joschthere are just tons of binnmus... which transition is this?23:06
vagrantchmmm... i'll need multiple SD cards to play with this, as the way i select between Debian and Guix is popping in a microSD with /boot with Debian (as guix makes it tricky to use a separate /boot)23:06
josch"Rebuild to enable GCS on arm64"23:06
joschoh23:06
vagrantcit is revealing quite some bugs where GCS is not supported23:07
bremnerGCS != GCC?23:08
joschit's hundreds of binNMUs because of this23:08
vagrantcbremner: i think GCC has been supported on arm64 for quite some time ...23:08
vagrantcthough i will admit to not having a clue what GCS is about23:09
vagrantcbut at least it is producing a lot of bugs23:09
vagrantchttps://wiki.debian.org/ToolChain/GCS23:09
bremneroic23:11
bremnerto be fair, new versions of GCC are also fun23:12
joschbremner: but usually we do not binNMU the world because a new compiler version got uploaded23:12
bremnerunderstood, but there are a surprising number of bugs with every new version. It's quite a record to keep up.23:13
sigridbugs with every new version of what, gcc?23:14
bremneryes23:14
joschyes, but we have mass-rebuilds for that, no? The results just do not get uploaded but FTBFS gets reported (by santiago etc)23:14
joschokay, i have the diff between last working and first failing23:15
bremneranyway, we digress...23:15
joschindeed but here i have some on-topic stuff for your eyes:23:15
joschhttps://mister-muffin.de/p/ZLrW.diff23:16
joschthat's the diff between working gdm (before) and broken gdm (after)23:16
minutejosch: gdm is not in there?23:17
joschminute: exactly23:18
joschand here is a diff of journalctl between working and broken:23:18
joschhttps://mister-muffin.de/p/bN7S.diff23:18
joschand i see this: /usr/bin/gnome-shell: symbol lookup error: /usr/lib/aarch64-linux-gnu/libgjs.so.0: undefined symbol: __aarch64_ldadd4_acq_rel, version mozjs_14023:18
joschm(23:18
joschunless this is a red herring and somebody sees something more obviously wrong?23:19
minuteuff that does not look good23:20
cararemixedAlmost done backing up. Got my new media ready for quicker recovery then I'll do an update and see how it breaks.23:20
minutewhat is that weird spam > WARNING: Unknown key type EIGHT_LEVEL_WITH_LEVEL5_LOCK23:21
joschminute: that's my keyboard layout, sorry23:22
minuteand what's > +dbus[5545]: Could not get password database information for UID of current process: Looking up user ID 60583: not found23:22
minute+dbus[5545]: Failed to start message bus: Memory allocation failure in message bus23:22
+ spew (~spew@user/spew)23:23
+ bkeys (~Thunderbi@134.22.115.162)23:23
minutejosch: did something change about arm64 build flags / compilers?23:25
minutejosch: maybe -moutline-atomics was enabled?23:26
joschminute: yes, there is a mass bug filing and apparently a mass rebuild to build everything in the archive with the default build flags, see https://lists.debian.org/msgid-search/aQEfqGt9PBDJZDuS@ariel.home23:27
joschfound it23:27
joschdowngrading libmozjs-140-0 to 140.6.0-1 fixes it23:27
minutejosch: uff uff23:28
minutejosch: great work! but yeah, probably this compiler option change broke it?23:28
joschmaybe that or maybe just the rebuilt against new versions23:28
minuteno bugs yet? https://bugs.debian.org/cgi-bin/pkgreport.cgi?pkg=libmozjs-140-0;dist=unstable23:29
joschyes, i also didn't find any23:29
joschi'm filing one now against gnome-shell and tag the affected packages like gdm3 and libmozjs23:30
minutejosch: many thanks!23:31
minutejosch: if i understand correctly, the compiler runtime has to provide these symbols23:40
minuteit might be libgcc-s123:43
joschhttps://bugs.debian.org/112592723:47
minutejosch: ok this might be more complicated btw: you are testing on cortex-a72, which doesn't have atomics23:49
minutejosch: so it's possible that mozjs would behave differently on cortex-a7823:49
joschnice23:50
joschminute: feel free to send a mail to the bug :)23:50
minutejosch: but it could be that this error shows up on rk3588 because of the cortex-a53s?23:51
minutejosch: i'm just wondering because on a78 (rk3588 big cores) this emulation function wouldn't be used normally as the cpu has native "LSE" atomics23:52
joschthis is tested on ls1028a23:52
joschi used the rk3588 to build the image23:52
minutejosch: yeah i think it's decided at runtime23:52
minutejosch: anyway. it's just interesting that it also doesn't work on rk358823:53
joschmaybe somebody can confirm the fix on rk3588?23:56
minutei can try to upgrade gdm on my pocket and then downgrading mozjs? i hope i have enough wifi bandwidth here23:59

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!