jaeger changed the topic of #crux-devel to: CRUX (https://crux.nu/) development channel | Logs: https://libera.irclog.whitequark.org/crux-devel/
SiFuh has quit [Ping timeout: 255 seconds]
SiFuh has joined #crux-devel
<jaeger> ok, clean install tests and upgrade tests (3.6->3.7 and 3.7->3.7) went fine here
SiFuh has quit [Ping timeout: 260 seconds]
SiFuh has joined #crux-devel
<SiFuh> jaeger: Anyone made sure this time that all the 3.6's have been changed to 3.7?
SiFuh has quit [Ping timeout: 252 seconds]
SiFuh has joined #crux-devel
<farkuhar> Upgrade using -rc4 went fine on a 2015 HP laptop
<farkuhar> I noticed however that the ISO still provides xorg-libxfont, which has recently been removed from the repo. It would save a trivial amount of space to delete this package from the ISO for the actual 3.7 release.
<SiFuh> Good catch
<ivandi> just noticed that upgrading from 3.6 to 3.7 doesn't install dumb_runtime_dir package. this package is in 3.7 core
<farkuhar> ivandi: Good catch. The machine I just upgraded already had pam_xdg (from contrib), so I wouldn't have missed any of the functionality provided by dumb_runtime_dir. Even so, I don't think the setup program tests for the presence of pam_xdg when deciding which packages to install (and I didn't have "pam_xdg: dumb_runtime_dir" in /etc/prt-get.aliases), so you're correct that it should have been injected.
<ivandi> the setup-helper script does not check for pam_xdg and does not inject dumb_runtime_dir
<ivandi> i do not understand why dumb_runtime_dir is in core
<ivandi> imho it belongs to opt, is there anything in core that needs XDG
<stenur> Well ideally XDG stuff is set during login
<stenur> Sorry!!! False channel!!!
<stenur> The /run/USER directory should be available by then, too
<stenur> I mean i _could_ create a CRUX-only pam_xdg_crux that is a bit smaller by reaping out creation of /run as such, as that is created by /etc/rc. But then again
<stenur> -rw-rw---- 1 ports ports 7399 Aug 2 16:46 /x/balls/pam_xdg/pam_xdg-0.8.1.tar.gz
<stenur> -rw-rw---- 1 ports ports 6320 Aug 2 16:48 /usr/ports/built/pam_xdg/pam_xdg#0.8.1-1.pkg.tar.xz
<stenur> -rwxr-xr-x 1 root root 14216 Aug 2 16:48 /lib/security/pam_xdg.so*
<stenur> This is not the world. At the moment it is hard to believe there will be any future changes, it could also be integrated as a C file like start-stop-daemon.
<stenur> -rw-r----- 1 steffen steffen 16294 Aug 2 15:26 ../toolbox.git/pam_xdg.c
<stenur> -rw-r----- 1 steffen steffen 3181 Aug 2 00:22 ../toolbox.git/pam_xdg.8
<stenur> But if the situation prevents this then so it may be.
<beerman> i think i got something to test rc4 on today
<beerman> is everybody downloading it now? :D
<stenur> 1.1 GB????
<stenur> coincidentally, i just got message my data volume is out, thirty seconds ago.
<jaeger> ok, the dumb_runtime_dir thing is an easy fix, at least
<jaeger> SiFuh: we try to every time, of course... but nobody's perfect, heh
<SiFuh> jaeger: :-) I already had checked on the rc3 version
<jaeger> I'll fix setup-helper and remove xorg-libxfont
<jaeger> argh... I wish inkscape didn't break with every single poppler update
<beerman> me too..
<ivandi> beerman: egl-wayland signature mismatch in 3.7 ports
<beerman> ivandi: noted
<ivandi> are md5sums still used. pkgmk creates them by default. probably PKGMK_IGNORE_MD5SUM should default to 'yes'
<beerman> jaeger: during setup while selecting bootloaders, it looks like its just throwing some bash syntax errors inbetween menues just after hitting enter. it goes on though, so not fatal
<beerman> it doesn't do that with selecting depending packages for the bootloader selection
<jaeger> hrmm, ok
<beerman> maybe its me on that hardware, we will see
<jaeger> I see it, though it disappears quickly... will look into that as well
<farkuhar> Speaking of bash syntax errors, the 2022-07-26 commit to core/bash/profile introduced the "bashism" [[ ]]. But /etc/profile is also sourced by dash, which doesn't recognize this syntax for tests. Previous versions of /etc/profile were more careful to avoid bash-specific syntax.
<jaeger> grep: stray \ before :
<jaeger> I wondered about that change to /etc/profile but forgot to check it more
<jaeger> It's this line: pdeps=`grep "^$f\:" $depsfile|sed "s|^$f: ||g"`
<jaeger> that \ is superfluous and now causes the warning
<jaeger> I'm guessing that changed with the latest grep update
<jaeger> yeah, grep 3.8 release email mentions stray backslashes
<jaeger> so far trivial fixes, not sure it would be worth an rc5 for these
<jaeger> We should revert that bash-specific change in profile as well, can reproduce the error with dash
<jaeger> unrelated, here's an inkscape port with a patch and without the poppler build: https://crux.ninja/tmp/inkscape.tar.xz
<beerman> jaeger: i have been hitting weird kernel issues while installing rc4. but it might be the hardware i am using so i am currently trying alpine. so far, this seems to run
<jaeger> What kind of issues?
<beerman> is there anything special needed to boot on an older hp 19" server?
<jaeger> Hard to answer that, I don't know anything about that particular server
<beerman> uhm.. kernel message read something like "kernel dazed and confused, [..] trying to continue" - and some ext4 errors though drives seem to work ok
<jaeger> My tests have been on virtualbox and libvirt/qemu virtual hardware as well as a ryzen physical system
<beerman> old, but i saved it from the press for this purpose and it fails
<jaeger> I don't see anything special in that list, 5000 series xeons and ddr3 are old and established... I haven't used HP's smart array cards much but they've been around a long time so I'd expect them to work
<beerman> weird, since alpine doesn't seem to have the same issue with it
<jaeger> Is there any indication if there's a specific pci device or kernel module causing the message?
<jaeger> Or more specifically, what's generating the NMI
<beerman> i'd bet on the raid controler, but it has been in use till recently but the usecase moved somewhere else, unless i managed to damage something durin the transport i'd like to think this one works
<jaeger> Because the 'dazed and confused' message comes from an NMI
<jaeger> Can it be disconnected temporarily to see if the message goes away?
<beerman> doesn't look like it can, at least i didn't find anything
<jaeger> I have some older Dell servers I can test on that are of the same generation, though of course won't have the exact same hardware
<beerman> alpine just runs and i am on my way to setup lxc on it to at least virtualize crux
<beerman> when i get there, i'd try to take whatever alpines kernel config is and build a iso with it
<beerman> or the error returns again :D it also seemed to flake out when i started writing data onto the harddrives
<jaeger> another option would be to install crux from the alpine media and also use the alpine kernel config, I guess
<beerman> heh, nifty idea, i'll maybe try that
<beerman> first time i installed alpine outside docker, the setup is pretty straight forwards
<jaeger> I have a single alpine VM here to run an NTP server, was just curious about it a while back and it's stuck around
<beerman> from my experience within docker containers i can say that i like it
<jaeger> I have a coworker who was a huge fan of alpine containers and then was very confused when he couldn't run dynamic executables he wrote there in glibc-based containers
<beerman> lol
<beerman> yeah, thats a thing to keep in mind for sure ;)
<jaeger> We still do use mostly alpine containers for all our golang stuff, I think
<jaeger> any objection to me reverting that test change in /etc/profile before the release?
<jaeger> (the bash-specific [[ ]] I mean)
<beerman> oh sure, if its problematic with dash its easy to rewrite that
<jaeger> yeah, simple fix, just asking since I think you made that change
<beerman> yep true
<beerman> at least i think so, can't remember each things without looking through git history :D
<jaeger> :)
<beerman> ok, so ilo reports error 0x12, for which they say update the controller firmware or replace the controller
<jaeger> charming
<beerman> but since alpine just seems to work i am tending to believe something is missing. not that it matters, its probably pretty obscure thing to do anyway (installing CRUX on the proliant that is)
<beerman> heh alpine comes with doas instead of sudo, neat
<jaeger> ok, looking at git, *I* made the change but it was copy/paste from your suggestions :D
<jaeger> I would guess it's one of 3 things: 1) kernel option difference, 2) firmware available, 3) microcode update
<beerman> yep, i agree
chrcav has quit [Quit: Lost terminal]
girafe has joined #crux-devel
<jaeger> that was fun, power outage for a couple hours
chrcav has joined #crux-devel
<beerman> really sounds like it
SiFuh has quit [Ping timeout: 248 seconds]
SiFuh has joined #crux-devel