IRC channel logs
2026-05-05.log
back to list of logs
<janneke>hmm, using --enable-ncpus=8 and using --smp 4 with qemu for the 32-bit hurd makes booting after running ...system/rc *very* slow for me (experimental guix/hurd) <janneke>i'm using 4GB, not pae yet, could this be a memory issue? <janneke>the 64-bit hurd seems to run fine with --smp 4 (using 8GB) <rrq>in uefi boot, I'd like gnumach to direct console to tty rather than serial; to do that, do I give it "console=tty" rather than "console=com0" ? <rrq>i.e., what should "console=...." say? <rrq>yes, and I'm not sure if it's a herring really... I need a way to confirm that gnmach is entered from grub <youpi>rrq: gnumach doesn't have a driver for the ife console <rrq>ok. but can it run without console? <janneke>i tried --enable-pae (together with --enable-ncpus=8) for the 32-bit hurd <janneke>with gnumach 1.8+git20260224, that doesn't boot for me in qemu (drops into the debugger) <janneke>ext2fs: part:1:device:wd0: Input/output error <rrq>youpi: ok. I believe I get UEFI boot happening, i.e. modules loaded, but I can't confirm the handover from grub to gnumach <janneke>so we'll probably create a gnumach-smp package, as --smp 4 with 4GB on the 32-bit hurd is very, very slow to boot (run guix's init) <rrq>ACTION finds .. panic("can't find a console device"); <rrq>appears my first option is to makes such an efi console driver that ties to uefi runtime services <azert>rrq: maybe more useful to make gnumach bootable without any console <azert>the Mach console is useful only for debugging. If one wants to debug I think a serial console would be useful in any case <azert>if one doesn’t want to do development, he doesn’t need the Mach console. The Hurd console already supports efi <azert>Im saying this because I expect it to be tricky to implement the framebuffer console both on gnumach as well as the Hurd. <rrq>I'm confused. Isn't gnumach the kernel module for Hurd? <azert>rrq : the Hurd has another console driver <azert>it can sit on top of the kernel one <azert>On a bios based system it is something like the following Apps -> Hurd console -> Mach console. But it can also be Apps -> Hurd console -> VGA <rrq>ah ok. though the efi runtime services offer the character api that would slot in nicely in (how I understand) the gnumach console driver setup; just a matter of glueing <azert>You would need to link in the efi libraries to gnumach <azert>And I’m not even sure this works after grub took over the os <azert>but happy to hear from you if you can make it work <azert>I am not aware of Linux doing that! <rrq>mmm maybe a "null" console would be easier of course <rrq>thanks. I alaways appreciate getting less to do :) <rrq>ACTION can spend the time on improving spelling <azert>I hope you spend time explaining us if you made it to work with the Hurd console <rrq>is that a translator? <rrq>should be doable, as another load module, unless it depends on filesystem (would be more chicken/egg) <azert>it is two translators in fact <azert>the relevant code is in hurd/console-client/fb.c <azert>there is a console and a console-client <rrq>(I expected "wiki" in the url) <rrq>ok. I'll go for "null console" to begin with. <janneke>ok, i believe that i booted a 32-bit hurd on a x220 yesterday, while the 64-bit failed <janneke>couldn't reproduce it today, the 32-bit hurd fails to find part:1:*wd0 in a similar way <janneke>i'm so confused -- i've ordered new ssd's because getting weird errors, possibly the disk is just broken <jab>janneke: are you using rumpdisk ? <sneek>Welcome back jab, you have 1 message! <sneek>jab, nexussfan says: My email is nexussfan [at] d u c k {DOT} com