IRC channel logs

2026-02-28.log

back to list of logs

<gnucode>sneek later tell gfleury It might be a good idea to email in your bug that you cannot compile rocksdb with smb = 6. That way there is a record of it.
<sneek>Got it.
<gnucode>I really feel like the Hurd needs pledge and unveil. That would be soo cool!
<nexussfan>?
<gnucode>nexussfan: are you not familiar with OpenBSD's pledge and unveil ?
<gnucode>(pledge "stdio");
<nexussfan>Not sure what that is. I don't use openbsd much
<nexussfan>only really free/netbsd
<gnucode>that means that the current program after that point will out output text or take input. If it tries to do anything else, then OpenBSD kills the program immediately.
<nexussfan>Interesting
<gnucode>pledge is probably in my non-experienced opinion, one of easiest and greatest security improvements in software in a long time.
<gnucode>You can actually use those calls on Linux now a days too via cosmopolitan libc.
<nexussfan>We might want to do it a hurd-like way if possible
<gnucode>when I mentioned it as a potential Google summer of code idea, Samuel made it sound like it was possible to do should be possible to do something like that on the Hurd.
<nexussfan>we still do gsoc?
<gnucode>unveil is also pretty cool. It allows a program to restrict its view of the filesystem.
<nexussfan>I mean you can already use subhurds to sandbox apps
<gnucode>nexussfan: we do. That's how we got rust ported. :)
<nexussfan>Oh wow
<gnucode>And we should definitely lean into using subhurds to sandbox apps if it works.
<gnucode>BUT I was under the impression that one subhurd cannot access the parent hurd's X11 session.
<nexussfan>x11's network transparancy could help
<nexussfan>"subdo" would be nice
<gnucode>with 2 lines of code "pledge" and "unveil", developers could easily state this program is allowed these syscalls and only can access these bits of the filesystem. That's pretty awesome.
<gnucode>what's "subdo" ?
<nexussfan>> https://hurdos.com/wiki/community/weblogs/ArneBab/niches_for_the_hurd.html
<nexussfan>> $ subdo --no-lasting-changes ./virus
<nexussfan>> Example: Let a virus run free, but any effect vanishes once the subhurd closes.
<gnucode>doesn't our current subhurd have some filesystem size limits ? I thought that you cannot have a filesystem larger than 4GB on a subhurd...I could be completely wrong about that.
<nexussfan>IIRC that was the old limit for ext2fs
<gnucode>I also need to get searching working on hurdos.com/wiki
<nexussfan>Yeah that would be nice
<gnucode>it shouldn't be that hard to get set up. I just have not done it yet.
<nexussfan>> https://darnassus.sceen.net/~hurd-web/faq/2_gib_partition_limit/
<nexussfan>Not sure where 4GB is from
<gnucode>I thought I read that in the mailing list a while ago. that the current subhurd has a pretty low maximum filesystem size. I've never actually ran a subhurd, so I don't know.
<gnucode>But with my T420 and 12GB of ram. I can definitely play around with it.
<nexussfan>12GB of RAM? Wow, nice
<nexussfan>My T420 only has ~8GB
<nexussfan>Which most hurd machines have 8GB iirc
<nexussfan>But IIRC someone said there was a 30GB RAM hurd system but only for a very short amount of time
<gnucode>yup! also which SMP GNU Mach image did you install ? gnumach-image-1-amd64-smp or gnumach-image-1.8-amd64-smp ?
<nexussfan>Those are the same
<gnucode>nexussfan: I believe you can upgrade the amount of ram that you 420 has. I think the max is 16GB.
<nexussfan>gnumach-image-1 is short for the latest 1.x version, gnumach-image-1.8
<gnucode>gotcha.
<nexussfan>gnucode: yeah, I know
<gnucode>and I believe that damien got the Hurd running on a thinkpad that didn't have a CD slot.
<nexussfan>My T420 has CD but it's broken on Hurd for me
<gnucode>I actually have a Dell optiplex 7010 I think...I can run the Hurd on that. Probably need to try rumpnet. That supports 32 GB I think.
<gnucode>I think that some of my next wiki edits should be "here's how to run lwip".
<gnucode>Here's how to run eth-filter
<nexussfan>Yeah
<nexussfan>You know you can use darnassus to edit wiki articles?
<nexussfan>Nobody does it anymore they all do it on the mailing list
<gnucode>I actually use darnassus to make tiny wiki edits.
<gnucode>when I need to add or delete pages or large edits, I email patches.
<nexussfan>Oh okay
<nexussfan>Wow they didn't put hurd support in the mesa changelog(?) https://docs.mesa3d.org/relnotes/26.0.1.html
<gnucode>sometimes darnassus is a bit slow to accept the edits...
<nexussfan>Is "Hurd NG" ever even gonna be done?
<nexussfan>Or at least worked on?
<nexussfan>Doubt so
<gnucode>how would you define Hurd NG ?
<ArneBab>nexussfan: as far as I know, Hurd NG isn’t pursued anymore, because Mach proved to actually be fast enough.
<ArneBab>Hurd NG was Hurd on L4
<nexussfan>Alright
<ArneBab>(Well: Apple choosing a Mach kernel and getting it fast proved that Mach works in production)
<nexussfan>Yeah
<gnucode>I find it interesting that one of the Hurd developers created x15 to try to pursue Hurd NG. He seemed to dislike the Hurd's async design and he preferred sync stuff. idk
<ArneBab>But with Rump, drivers are being moved out of Mach, so a shrinking of Mach *is* happening.
<nexussfan>> Extended GNU distribution http://i-hug.sarovar.org/downloads/GNU/extended/
<ArneBab>(and IPC proved to actually be fast on modern systems)
<nexussfan>Link is dead
<nexussfan>lost media, not on IA
<gnucode>nexussfan: so what can we do now that mesa works ?
<nexussfan>gnucode: play frozen-bubble!
<nexussfan>also plib should work now
<nexussfan>and lots of apps which for some reason need gl should also now work
<nexussfan>Like kitty
<gnucode>hahaha!
<nexussfan>But I think the most important part is that now tuxkart and frozen-bubble work
<gnucode>that is pretty cool!
<nexussfan>I'm sure you remember those 2 games
<gnucode>I do. I've played frozen bubble before. I was not really a big fan of tuxkart.
<gnucode>actually I've never really been a big gamer.
<nexussfan>Yeah, frozen-bubble was fun. Too bad all of the multiplayer servers are dead
<nexussfan>I host my own frozen-bubble multiplayer server but the central server lists are all dead
<gnucode>I don't see how you would play that game on multiplayer mode...
<nexussfan>> mesa build-depends on missing: - libdrm-dev:hurd-amd64
<nexussfan>We still need libdrm-dev?
<nexussfan>gnucode: you can talk about it in #frozen-bubble ;)
<gnucode>hahaha! I didn't know that they had a irc channel!
<nexussfan>There used to be a freenode one
<gnucode>claws-mail works on the hurd...
<gnucode>installing now.
<nexussfan>Right now mesa still doesn't work on buildd
<nexussfan>So do you have to compile it yourself instead?
<gnucode>hmmm
<nexussfan>can the dependency be removed for it?
<gnucode>hmmm, claws-mail is not starting.
<nexussfan>I'll ask about mesa libdrm-dev dependency on the mailing list if it doesn't get fixed
<dfceaef>hi all, I'm testing libite on hurd. Can you confirm stat("/dev/zero", &st) == -1 on hurd-i386?
<gnucode>dfceaef: I am trying to check for you now...
<gnucode>oh wait...I'm on hurd-amad64
<gnucode>hurd-amd64*
<mvkbed>hi
<gnucode>hello mvkbed !
<mvkbed>i spun up guix/hurd in qemu and why doesn't shutdown reboot halt work reliably?
<gnucode>I know that damien did some work on getting reboot to work in Debian GNU/Hurd. Perhaps Guix/hurd is just behind and hasn't caught up with some of the newer features ?
<gnucode>I don't even know if guix/hurd supports amd64 yet.
<janneke>gnucode: guix/hurd amd64 does not have smp yet
<nexussfan>amd64 does have smp, not sure about guix though
<gnucode>I believe that janneke is saying that guix has guix/hurd amd64 just not SMP.
<nexussfan>oh
<nexussfan>> Hurd LiveCD
<nexussfan>There used to be one?
<gnucode>mvkbed: I will say that my Hurd machine running on real hardware doesn't quite shutdown as well as I'd like. I do have to hold the power off button to force the machine to shutoff after "sudo poweroff"
<gnucode>I didn't know there was a livecd. I knew about the installer cd.
<nexussfan> https://commons.wikimedia.org/wiki/File:HURD_Live_CD.png
<nexussfan>GNU 0.3 though
<gnucode>still cool.
<nexussfan>I wonder why midori was deleted from debian
<mvkbed>the problem is that when forced shut down the ext2 partition will politely shit itself and never recover then I have to reinstall it. how are you guys even "developing" it like java compatibility etc if it can't shutdown. surely you want a working system to make some progress. do you know some things that I don't?
<nexussfan>???
<nexussfan>sudo halt
<nexussfan>then once it says "HAlting Hurd"
<nexussfan>You can force shut down
<gnucode>I don't force shut it down. I do what nexussfan said. "sudo poweroff" I see "halting hurd" I wait a few seconds, then I force shut it off. At that point ext2fs is unmounted.
<gnucode>mvkbed: try Debian GNU/Hurd. You can actually run X on it. I'm typing this to you on a T420 running i3 and Emacs.
<mvkbed>it doesn't say anything. just sits there. sometimes it doesn't work with shepherd shutting his services and servers, other times it just stays like that for very long time. maybe it should be a bit verbose.
<gnucode> https://hurdos.com/wiki/hurd/running/flash_qemu_img.html
<gnucode>mvkbed: are you using qemu ?
<mvkbed>gnucode: i want to work on this shutdown thing now. i want to fix this.
<gnucode>do you have kvm enabled ?
<mvkbed>yes I'm using qemu
<mvkbed>I have kvm enabled
<nexussfan>acpi halt should work then
<gnucode>is your user a member of the kvm group ?
<gnucode>I made that mistake on guix...thinking that I had kvm enabled, but then realized that my user could not access /dev/kvm
<mvkbed>oh same here. how does it affect that?
<gnucode>your user has to be able to access /dev/kvm
<gnucode>do a ls -lha /dev/kvm
<gnucode>whatever group /dev/kvm is on, your user has to be a member of that group.
<gnucode>qemu should give you a warning message if you do not have kvm enabled instead of silently continuing.
<gnucode>also debian GNU/Hurd may have ext3fs or ext4fs soon. We've got a cool developer that added a journal.
<nexussfan>the journal for ext2fs is not that stable yet, right?
<gnucode>I haven't tried it. no idea.
<gnucode>it's not merged yet. That's for sure. and in some conditions it apparently has a 10% slowdown.
<mvkbed>good thing I had the snapshot of vm. but I still don't get the shutdown problem.
<gnucode>it's a leftover from a witch that cursed a cursed the project in the 1800s. We're still trying to dispell her dark magic.
<gnucode>*sounds like it could be true*
<nexussfan>xD
<mvkbed>:D
<nexussfan>I wonder how the CHUG <https://www.gnu.org/software/hurd/community/chug.html> is doing. how do I mail that i'm interested? ;)
<gnucode>haha
<nexussfan>darnassus - KVM guest on dalaran
<nexussfan>Aw man
<nexussfan>I thought it was actually running Hurd on a real machine
<gnucode>nexussfan: I don't see how it could. My hurd that I'm running on...I think I'm getting 10 - 60 KB/s
<gnucode>that's rumpnet
<nexussfan>Still faster than gnu.org though
<gnucode>gnu.org is fast for me state side.
<nexussfan>For me I get ~135 kB/s, according to apt
<nexussfan>download
<gnucode>that's not too bad.
<nexussfan>Yeah
<mvkbed>hey. a heads up on my issue.when you type shutdown, it does nothing and there is a blinking cursor. if you try to type anything nothing appears, but if you constantly spam enough characters, it will then say xkb: something about queue being long or overflowing, then suddenly shepherd will stop log service and then everything will follow and it will shutdown.
<mvkbed>now what is the actual issue? what service isn't stopping? strace isn't installed dmesg can't be installed because internet never works on qemu don't know why. no matter which nic model
<gnucode>mvkbed: are you typing shutdown whilst in X ?
<mvkbed>in tty. there is no X
<nexussfan>I have no problem shutting down in X
<mvkbed>nexussfan: are you on guix/hurd?
<gnucode>we're all on debian GNU/Hurd.
<mvkbed>oh. atleast tell me how to see what's the problem I'll fix it. I don't have any problems with C coding
<gnucode>try poweroff. See if that works better for you.
<gnucode>also are you using a child hurd ?
<mvkbed>did that. no I'm not using childhurd. I'm using pure guix hurd
<gnucode>i would ask janneke ; He's one of the guix/hurd developers.
<gnucode>also try debian hurd too. :)
<gnucode>would you happen to know how many packages guix/hurd has ?
<janneke>there will be a blog post with updates on guix/hurd tomorrow
<nexussfan>wow
<janneke>guix/hurd 32bit has about 1.7% substitutes available (389 out of 23,223), 64bit: 0.9% substitutes available (204 out of 22,610)
<nexussfan>Surprising
<janneke>that is not to say that the others don't build, but most probably many of them are blocked by test suites of basic packages failing
<janneke>guix runs all test suites by default, also on hurd
<janneke>you can run `guix weather --system=...'
<janneke>so, guix/hurd is still nowhere near where debian is
<nexussfan>Understandable
<mvkbed>janneke: but I think guix hurd should be also worked on by someone (I think I'll do it) atleast I want to make it usable system enough for reliable coreutils as well as emacs ecosystem support.
<gnucode>mvkbed: there might be a hurd team that you could join.
<mvkbed>I'm trying to find couldn't find. any idea?
<gnucode>janneke you might mention that journal that the Hurd is working on. Last hurd blog post I think you mentioned ongoing SMP / 64 bit support. The journal, rumpnet, rumpdisk are both very exciting improvements.
<gnucode>mvkbed there's a page for it on the guix manual.
<gnucode>info guix RET i teams
<gnucode>if you have git cloned guix.... $guix-src/etc/teams.scm
<mvkbed>oki. thanks a lot fellow gnu
<janneke>gnucode: of course, and certainly!
<janneke>mvkbed: yes, we can always use some help
<rrq>mvkbed: late to the party; I turn off hurd with "shutdown -n -h -P now"
<mvkbed>rrq: it gives error: unknown option -n
<rrq>hmm different program versions I guess? min is from sysvinit-core=3.14-4 .. I'm using Debian GNU Hurd amd64
<mvkbed>it only works good when I spam letters non stop until the xkb fills out or something then it stops the logging service and shepherd stops all services and shutsdown
<mvkbed>rrq: unfortunately I've decided to take myself to work on guix/hurd
<rrq>fair enough; sorry for noise
<youpi>gnucode: subhurds don't have more limits than normal hurds