IRC channel logs

<nexussfan>> unknown or no method inherits keyword specified

<nexussfan>Looks like "iface /dev/wm0 inet6 auto" is not working

<youpi>azeem: I have just uploaded libc0.3 2.42-12~hurd.1 with the gdb-after-fork fix

<nexussfan>even though I fixed the translators, now ifupdown is having issues with ipv6 xD

<rrq>it complains about "auto" (?)

<damo22>shouldnt it be dhcp

<damo22>not auto

<rrq>I think inet6 has an auto method

<damo22>ok tbh ive never used ipv6 but rumpnet supports ipv6 packets

<damo22>it just sends and receives ethernet frames it doesnt care too much what is in them

<damo22>nexussfan: try inet6 dhcp not auto?

<nexussfan>damo22: okay

<nexussfan>Still ifupdown says "unknown or no method inherits keyword specified

<azeem>youpi: thanks! I built it locally on Sunday and I now got proper backtraces

<azeem>youpi: those are the backtraces: https://paste.debian.net/hidden/9c89883d

<azeem>there is a signal handler involved indeed

<azeem>as the Hurd does not have epoll/signalfd, Postgres uses poll and a self-pipe to wake up the other process, see https://github.com/postgres/postgres/blob/master/src/backend/storage/ipc/waiteventset.c

<azeem>need to go now

<youpi>azeem: oh, that backtrace is very instructive

<youpi>the asynccancel support indeed takes a mutex, thus not signal-reentrant

<youpi>I'll have to think about it

<youpi>we'd probably just have to make these asynccancel sections critical to defer signal delivery

<damo22>youpi: i dont understand the current UP xen support, how come it seems to need locore.S as well as xen_locore.S ?

<damo22>is that a mistake?

<damo22>i thought interrupts in xen are events not real irqs

<youpi>damo22: possibly most of it is unused

<youpi>but some of it is needed indeed

<youpi>such as the kdb_kintr to be able to enter the debugger etc.

<youpi>even if interrupts in xen are not real irqs, they follow the same kind of path

<youpi>and there's copyout/etc.

<damo22>ive added some xen smp code that should bring up vcpus

<damo22>but i need to make it compile

<damo22>im not sure how to set up the gdt and idt in xen

<youpi>see fill_descriptor, which makes a hypercall to tell the hypervisor

<youpi>and hyp_set_trap_table

<youpi>called from gdb/idt_init/fill

<damo22>ah ok

<youpi>ACTION afk

<damo22>thanks

<azeem>youpi: ok

<sobkas>Is there any way to get more than 16G of memory in gnumach? Because it doesn't boot above this number and I need apparently 14G to build cmake with ramdisk (storeio)?

<sobkas>rumpdisk isn't useable for me as it causes hangs to the point that shutdown -h now doesn't work

<sobkas>It just sits there doing nothing

<youpi>is that 32b or 64b ?

<sobkas>same for building cmake

<sobkas>64b

<youpi>I haven't tried that much memory with 64b, but probably not much to fix there

<youpi>for 32b it may be more tricky, though PAE should be giving quite a lot already

<youpi>it's however odd that cmake requires that much memory, it didn't need that much before

<sobkas>It needs so much disk space, and when I use ramdisk, you know what happens

<sobkas>And I use ramdisk because it hangs all the time when I use rumpdisk

<sobkas>I made simple script https://paste.debian.net/hidden/007a31bf

<sobkas>Best indicator for now I have is watch sync, it just stops with rumpdisk, while ramdisk doesn't have this problem

<sobkas>Does /hurd/storeio -T typed copy:zero:15G:file:/dev/wd1s1 do something?

<sobkas>It shows /dev/ramdisk0 17G 1.9G 14G 13% /srv/chroot, while it normally would 15G without file

<sobkas>If I get random reboot it will mean that it didn't do anything...

<etno>sneek: later tell sobkas maybe using a network filesystem would work for building cmake ? NFS?

<sneek>Okay.

<Pellescours>I wrote a quick patch to be able to compile hurd with recent libc https://github.com/etienne02/cross-hurd/blob/archhurd/patches/hurd/remove-in_pkginfo-struct.patch

<Pellescours>I don’t know what I should do to have it retro-compatible

<sobkas>I just don't understand:

<sneek>sobkas, you have 1 message!

<sneek>sobkas, etno says: maybe using a network filesystem would work for building cmake ? NFS?

<sobkas>sudo settrans --create --active /dev/ramdisk0 /hurd/storeio -F -T file /dev/wd1s1 sudo /usr/sbin/mkfs.ext2 -F -b4096 -o hurd -I128 /dev/ramdisk0

<sobkas>works

<sobkas>will try this time with tests turned on

<sobkas>etno: nfs have problems with chown

IRC channel logs

2026-02-03.log