IRC channel logs

2026-02-18.log

back to list of logs

<jab>anybody have a guess how large a qemu hurd image should be if you want to install i3 on it?
<jab>now that I have the hurd installed via qemu, I need to get it to boot via rumpdisk.
<jab>Then I can try flashing that to a image to a laptop.
<nexussfan>not sure how qemu works with rumpdisk, but thinkpads work well with it
<damo22>i dont think any hurd rootfs needs to be bigger than 10G
<damo22>if you put other data on a separate partition
<jab>sounds good.
<jab>nexussfan: really? why wouldn't rumpdisk work with qemu ?
<damo22>rumpdisk was developed on qemu
<jab>sorry, you said "not sure how qemu work with rumpdisk" not that it didn't work.
<jab>damo22: is there a particular qemu invocation that you need to run rumpdisk ?
<jab>I guess I'm going to try to follow the guide that I wrote to try to use rumpdisk:
<jab> https://darnassus.sceen.net/~hurd-web/hurd/rump/rumpdisk/
<damo22>it should all be configured already in debian
<nexussfan>yeah
<jab>ok.
<damo22>just pass noide
<jab>well in that case I should just be able to dd the qemu image to the disk.
<nexussfan>but the installer still allows you to use the IDE drivers if you need to
<damo22>and the right disk device
<nexussfan>however for some reason my t420 hurd machine doesn't like cd0, so i can't use the installer
<jab>nexussfan: I haven't been able to get the installer to work for me lately. Right now, I'm trying to install the hurd via flashing the debian image.
<damo22>i havent reinstalled since the beginning
<damo22>once you have a working disk you can upgrade
<nexussfan>Yeah
<jab>that's impressive.
<damo22>i did lose my partition once due to corruption, so i might have had to blow it all away once
<damo22>but things are more stable now
<jab>damo22: are you using the 64 bit version more often than the 32 bit version yet ?
<damo22>yes
<nexussfan>I only use 64bit hurd.
<damo22>i only use i386 CI to test
<damo22>and occasionally i boot i386 smp to test something
<nexussfan>I used to use i386 Hurd on my T43 but not anymore
<damo22>the gnumach test suite fails on x86_64 for some tests, we need to fix all those
<nexussfan>Waiting for x86_64 smp
<damo22>well im kind of stumped on this page fault
<nexussfan>does PAE work on i386?
<damo22>yes
<nexussfan>nice
<jab>I forget...are ya'll using the 64 bit version on real hardware ?
<nexussfan>Yes
<damo22>no, i develop in qemu
<jab>gotcha.
<damo22>rebooting a real machine is painful
<nexussfan>My tests/coding are in qemu x86_64, but I run ready software on my real hurd machine
<jab>that still is pretty cool.
<damo22>the APs try to execute code at 0x4000000036
<damo22>not a real address
<nexussfan>The only real thing I am working on for hurd right now is xlibre; i'm trying to debug why the kbd driver isn't working
<damo22>nexussfan: if youre interested in gui, why not work on wayland
<rrq>damo22: is that calling vm_map_msync ?
<jab>nexussfan: did you see that Xorg just deleted 2 years worth of commits?
<damo22>rrq: i dont know
<rrq>yeah I just checked some ...0036 address options in the gnumach symbol table
<damo22>it might be a made up rubbish address
<damo22>intended for some other purpose
<damo22>but somehow gets called
<nexussfan>damo22: that's lots of work with drm, compositor support, etc; xlibre support is much closer anyway, only couple more things to fix. maybe after finishing xlibre support i could work on fixing a wayland compositor for hurd
<nexussfan>jab: yes you already told me
<damo22>nexussfan: i thought the mesa patches sobkas worked on might be enough to get wayland working with software rasterisation
<nexussfan>oh! interesting
<nexussfan>there are a lot of different compositors however, which one should be focused on first?
<damo22>probably the gnome one?
<nexussfan>Mutter okay
<jab>I personally vote on sway. :)
<nexussfan>I was thinking of doing weston but that has a buildd time of 100d :O
<nexussfan>I think we should first do an easier compositor
<damo22>gnome integrates well with desktop
<nexussfan>true
<damo22>its not the lightest one but it would be shiny
<nexussfan>once mesa/hurd gets upstreamed i'll check it out
<jab>as far as I can tell, qemu does not use rumpdisk by default by the way. /etc/fstab shows /dev/hd0 not /dev/wd0
<jab>not super hard to fix. I just thought I'd mention it.
<damo22>its not a bug, its on purpose i think because its not quite ready apparently
<damo22>it still chews too much ram
<damo22>that is probably the last remaining issue
<jab>well I'm not sure what I did wrong, but I could not get rumpdisk to run on qemu
<jab>I was following this guide: https://darnassus.sceen.net/~hurd-web/hurd/rump/rumpdisk/
<damo22>-M q35 ?
<jab>I should probably add that in and see.
<damo22>it needs that so it has AHCI controller
<jab>let me try again.
<jab>damo22: I got rumpdisk working on qemu. Thanks!
<jab>morning friends!
<jab>oh that's cool. The 64 bit hurd is using rumpdisk by default.
<jab>It's only the 32 bit image that is not using it by default.
<jab>and it looks like the 64 bit image by default doesn't make you configure your apt sources.list file. That's nice.
<jab>I thought I saw it say that it was upgrading the icewm. Hmm. I don't recall installing that, but that's cool.
<jab>huh...is this already set up to run X ?
<jab>grrr, guix shell for the hurd website is not working for me again...
<jab>sneek later tell youpi, hey samuel the amd64 README is slightly in correct in the losetup line.
<sneek>Got it.
<jab>sneek later tell youpi this is what I used that worked: " sudo losetup -o $((512*1955840)) /dev/loop0 debian-hurd*.img "
<sneek>Got it.
<jab>sneek later tell youpi, actually I couldn't figure it out. I do know that the 1953792 value that the README mentions here is incorrect. https://cdimage.debian.org/cdimage/ports/latest/hurd-amd64/YES_REALLY_README.txt
<sneek>Got it.
<jab>actually I think I did figure out how to resize my hurd qemu image.
<jab>I saw an input output error, so I assumed that I ran the commands wrong. But it was an io error for cd0
<jab``>well I guess I am going to copy the 64 bit hurd qemu image to a disk and try to boot from that disk.
<sneek>Welcome back jab``, you have 1 message!
<sneek>jab``, old says: for example, you can change where something will be installed, without having to mutating the object itself
<jab``>fingers crossed.
<jab``>well I officially have the hurd running on a T420 with 12 GB of RAM!
<jab``>that's awesome!
<jab``>i just need to grow the size of the filesystem
<azert>jab``: that’s great, does x11 work?
<jab``>I got x11 working in qemu. once I get the filesystem resized, I'll try it in real hardware.
<jab``>then I'll need to see if networking works.
<jab``>X works. i3 is pretty snappy. But networking is not working.
<jab``>Time to try rumpnet
<jab``>it is possibly that my filesystem corruption caused the networking issue...
<jab````>hmmm, I think these commands need to be as
<jab````>root
<jab````> https://darnassus.sceen.net/~hurd-web/hurd/rump/rumpnet/
<jab````>I don't know what's happening to my irc nick by the way.
<azert>jab: what is your nic network interface?
<azert>if it is the Intel PRO/1000, then I think that rumpnet is the right driver
<jab>pretty off topic, but did any of ya'll see the guile based build tool talk ? It's called BLUE, and it's fairly impressive what they have got going for it already.
<sneek>jab, you have 2 messages!
<sneek>jab, old says: that the BLUE build system tries to be purely functionnal as much as possible
<sneek>jab, old says: that BLUE wents a to great extent to be able to represent stateful data in a purely manner with delayed expressions. This allows changing the states in the REPL for example without doing any mutation
<Gooberpatrol66>is blue integrated with guix? like can it put build objects in the store?
<jab>Gooberpatrol66: that I don't know, but I'm sure that guix will create a BLUE build system at some point.