IRC channel logs

2026-05-15.log

back to list of logs

<jab>morning hurd friends!
<janneke>hey jab!
<janneke>jab: i managed to boot guix/hurd (also debian/hurd) on an X200; the X equivalent of the T420
<janneke>it was a bit of a struggle/puzzle because i needed to add the acpi server, which had some dependencies
<janneke>and one of them only lives in debian/salsa; so yeah ;)
<jab>janneke: awesome! Let me know when you are ready to try out "ext3/4fs". Milos gave me some specific instructions for how to get it working.
<janneke>ooh, journaling would be grand! I thought journaliing was being added to ext2, i haven't heard of ext3/ext4?
<janneke>bjc: ^^
<janneke>yelninei: &&
<bjc>i am excited for progress in this area
<bjc>is it actually ext4?
<bjc>i'm not sure why ext3 and 4 get lumped together so much, tbh. i guess they're largely identical
<janneke>2,3,4 share a lot afaik
<bjc>right, but presumably there's a reason one is 3 and the other is 4
<bjc>i know 3 added journaling, but i'm not sure what 4 did
<janneke>ACTION has to run, ttyl
<jab>oh yeah, Milos coded essentially "ext3/4fs".
<jab>It's in patch form at the moment. It is still in review, but in my non-technical opinion, it seems like it'll get merged eventually. The main hold up is no one has really reviewed it yet.
<jab> https://hurd.ion.nu/news/2026-q1.html -> has the info about "ext3/4fs".
<azert>jab: maybe we should make a Debian/guix package with Milos’ filesystem, without having it merged with the hurd until it is reviewed
<azert>I’d like to apt install hurd-ext3fs and get it in my path
<azert>that’s how fuse filesystems are distributes
<azeem>makes sense, it'd be good if it could be released as a separate tarball or somethign then
<jab>azert and azeem, that's not a bad idea. I'll ask Milos if I can publically post how one can test out "ext3/4fs".
<jab>he emailed me the instructions about how to use it on real hardware privately.
<azert>I honestly think it would be a good policy for new drivers or servers to sit around for some time out of tree before getting included. This also to show some committment
<azert>of their authors to maintain it
<jab>that's not a bad idea. I believe that's why Samuel is fairly interested in using one of rump's filesystems. It's externally maintained.
<jab>oh speaking of rump...
<jab>I guess one of the problems of rump is that it is using an odd buildsystem. I think something about the newer bits of NetBSD are using a "template" build system...
<jab>anyway, maybe we could trying to build rump with BLUE, which is the guile based build system that guix has been working on making for some time.
<jab> https://fosdem.org/2026/schedule/event/3A7VGM-blue/
<azert>jab: in regards to rump, it is up to them to use any build system they want
<azert>problem is that I expect netbsd to ditch rump any day
<azeem>why is that?
<azert>because it’s not very useful for them
<azert>except rumpfs, which they are using themself
<azert>they don’t have userspace drivers
<azert>ideally, netbsd and the hurd would share the driver framework in userspace
<jab>damo22: has mentioned that NetBSD does use rump to test some things, but yes its future is questionable...
<sam_>it does get changes, like in say https://mail-index.netbsd.org/tech-kern/2026/03/13/msg030892.html where people mentioned its use and duplication in the initial patch, but yes, it doesn't get a huge amount of love either
<sam_>that said netbsd doesn't move at the speed of light (which is a feature) and i don't know if it means it's on its way out
<sam_>getting the hurd patches upstream would be nice but of course it takes effort