IRC channel logs

<diegonc>a patch for the linux kernel for libdrm support on Hurd, funny :D

<jitsukai>i was see FOSDEM 2026 about Hurd. beautifull works

<jitsukai>i have some question about, like mach and multi-cpu. one pers cpu or on for all cpu ?

<jitsukai>i don't familar with NetBSD rump

<jitsukai>may be and i speak about on hurdfr, one mach pers cpu, with one orchestrator kernel and client kernel for more cpu chips...

<jitsukai>as speak there under a sentrans

<jitsukai>sorry... juste is noctambul speaker

<hwpplayer1>jitsukai: hello

<Gooberpatrol66>one mach kernel for all cpus

<damo22>jitsukai: gnumach can run all the cpus with symmetric multiprocessor SMP

<damo22>then you can run multiple threads in various processes or the same process simultaneously

<damo22>it isnt 1 mach process per cpu

<damo22>rump is currently configured to be single threaded

<damo22>per unikernel

<damo22>although it has multiple threads for different parts of it, there is only cpu emulation for one cpu

<etno>When including hurd/irqhelp.h, GCC complains that error_t is not defined. What does the correct usage look like ?

<etno>(Hi ! Btw 😅)

<etno>Context: I am having irq registration issues in a driver program that I build using Hurd makefiles. No error_t issues there. Then making a standalone little reproducer for the irq issue, but hit the aforementioned error.

<azert>etno: looks like a bug

<azert>a missing #include

<azert>can you fix it?

<etno>I should compare with master, as my version of irqhelp.h comes from debian and may be stale

<etno>I have been running the build of irqhelp.c with 'gcc -E' and I can see the type being defined. But I can't make sense of the comments there that should help trace the include tree

<azert>I would use grep

<etno>That would work to find where error_t is defined; but I am also wondering what should be the proper way to fix it.

<etno>For instance, is 'error_t' a notion in Hurd or inherited somehow from mach or libc ?

<Darelelve>etno: The error_t type is declared in glibc/bits/types/error_t.h. It is also used when parsing input in argp, for example.

<etno>Ok, thanks Darelelve . According to you, what is the proper way to get the error_t type from a header in Hurd source ?

<etno>I see hurd.h

<etno>errno.h has it as well

<etno>That fixes my problem, thanks

<Darelelve>etno: I would just include errno.h.

<etno>Yes

<etno>So the remaining question is whether irqhelp.h should have this include ?

<Darelelve>etno: Hmm, the build works for me both with and without the file.

<etno>So we could add it, and it would make the life of people using libirqhelp from outside the Hurd easier

<Darelelve>youpi: Hi, I'm figuring out file_get_storage_info so that ext2fs can get the store from storeio. So, I managed to make storeio be between rumpdisk and ext2fs in grub.cfg, that is, if I don't remove -T typed '${root}' from ext2fs, the boot works fine, but I don't understand how to get the store itself via file_get_storage_info.

<Darelelve>I was thinking of changing diskfs_init_main so that during the bootstrap stage it opens the store not from store_parsed_open, but via store_create, where file is parent_task (this is storeio, right?). But when I run it, I get the error ext2fs: storeio: (ipc/mig) bad request message ID, and file_get_storage_info returns error -303... I think I'm doing something wrong.

<azert>etno I would agree to that, I think your discovery deserves a patch

IRC channel logs

2026-04-21.log