IRC channel logs
2026-03-17.log
back to list of logs
<azert>need some help, what is the correct way to build mach after adding a new type to mach_types.defs ? <azert>updating the system headers? <gnucode>I'm reading the irc logs from yesterday...about sockets. <sneek>Welcome back gnucode, you have 1 message! <sneek>gnucode, Alicia says: I think I found the -M q35 thing, it only happens for 4+GB of RAM. But I imagine it's not specific to rumpdisk because when I tried for 20GB it quickly bootloops back to grub, with or without -M q35 <gnucode>it makes me wish sergey submitted his plan 9 networking thing. It think it's called ip. or ipfs. <gnucode>sergey had an implementation for it. as far as I know he did not submit the WIP code. I've pleaded for him to do so. but I don't know where he put said code. <nexussfan>hopefully soon btrfs works so i can use it :D <gnucode>nexussfan: if by gnupfs you mean gnunet ... have you ever tried gnunet ? <gnucode>the last time I tried it, it was super slow! <nexussfan>gnucode: I've tried gnunet but it's very broken as of now for me <gnucode>but it's probably a decent idea anyway. <gnucode>nexussfan: in my opinion, bcachefs may be easier to port to the Hurd. <gnucode>something like 90% of bcachefs already runs in userspace. <nexussfan>i've seen that hurd is interested in it anyway <gnucode>did you know that Kent overstreet talked to me and sergey by the way ? <nexussfan>maybe we can try to do something like freebsd and have linux kernel translator <gnucode>but I'm not really a developer. We'd need someone with serious developmental chops to be able to port it. And even then, updating it would probably be really hard. <gnucode>nexussfan: well DDE is linux...the problem is that no one maintains it. Linux changes its internal API rapidly. It's very hard to maintain that amount of code churn. <nexussfan>maybe we could get some help from the freebsd devs <gnucode>filesystems are hard...so I've heard. <gnucode>yeah. Samuel's interested in that idea. rumpdisk is the filesystem drivers, but then you use netbsd's ffs via rumpfs. <gnucode>ffs is not the most recent filesystem design, but it is externally maintained. <gnucode>as long as rump is maintained...apparently that is becoming a slight concern... <gnucode>I don't actually know. I think we may only be interested in ffs...what other filesystem does netbsd support ? <gnucode>if they successfully port hammerfs...that would be cool. <gnucode>are any of you guys "libre software only" ? or to ask another way, would ya'll use zfs on the Hurd ? Is that zfs licensing issue a real problem or an imaginary one ? <damo22>hurd separates licensing concerns from the OS <damo22>you can even have a proprietary driver in userspace <gnucode>damo22: would you run rumpzfs on / on the Hurd if it ever got ported ? <nexussfan>one day... in the vast future... once hurd nvidia support finally happens... you can just download it and settrans /dev/gpu0 or whatever(?) <gnucode>well we may have ext3fs or ext4fs soonish... <nexussfan>gnucode: i like btrfs. I use btrfs for my systems <Gooberpatrol66>it took the ai bubble for nvidia to start supporting linux, so here's the plan, we get the stock market to pump $1 trillion into hurd <gnucode>nexussfan: I guess I've drunk the Kent Overstreet kool-aid. He makes it sound like btfs is a bit of a jumble of a code base. <nexussfan>gnucode: I always hear about issues with RAID, data loss, etc. but btrfs actually has less data loss for me than with zfs <gnucode>Gooberpatrol66: I didn't realize that the AI boom pushed nvidia to support Linux. interesting. <damo22>i thought nvidia had a massive source leak and so they decided to "support" linux <damo22>so if one of my disks fail i can still mount it <gnucode>can you check and correct for accidental bit flips on ext4 ? <nexussfan>once ollama and some other stuff works on hurd maybe we can have ai datacenters running on hurd? <gnucode>don't we seem to have a maximum GB ram limit at the moment on the Hurd ? <damo22>we dont need AI, its a waste of electricity <gnucode>I've heard some people mention that 16+ GB in qemu causes some problems. <damo22>if there are bugs causing problems with limits they can be fixed <nexussfan>Gooberpatrol66: settrans /dev/gpu0 /hurd/nvidia --gpu 5090 <gnucode>am I correct in thinking that all open source OSes just port the Linux graphic stack ? <Alicia>16GB works. over that I get a bootloop <etno>grpc just finished building, the directory takes 22.6GB 😅 <etno>bear is built, and seems to work. I ran it on the hurd project, and with geany+lsp_plugin+clangd, it gives all the fancy navigation and hover doc magic stuff. Will try on gnumach next. <azert>could I get some help on adding an rpc to gnumach? <azert>I've seen on the wiki here and there that you need to recompile glibc after adding an rpc <azert>this looks involved, any howto or step missing? <azert>I have a patch to propose, but I cannot test it if I don't go through this <p4r4D0xum>yes, sam_ added some preliminary stuff to official gentoo repo and ported crossdev to build i686-gnu toolchain, building it as ee speak