IRC channel logs

<nexussfan>Any Matrix clients for Hurd?

<azeem>youpi: the "One can increase the size of the image" part of the img README is under the "Execution with KVM" section, but I think it merits its own section, it does not seem KVM-specific

<azeem>(also, the amd64 image does not have a swap partition at the beginning I believe, so the "resizepart 2 100%

<azeem>needs to tbe "resizepart 1 100%" there, maybe a comment could be added to that end)

<Alicia>I don't remember if it had a swap, but it has an extended partition at 2, with the filesystem at 5

<azeem>"it" being the amd64 image?

<Alicia>yes

<azeem>hrm, maybe I changed something, but my image only has one (primary) partition

<azeem>my i386 image has two primary partitions, first a 1GB swap partition and the then ext2 one

<Alicia>maybe how the image is built changed, but when I tested it the other day that was the case. 32bit only had two partitions, with the filesystem at 2

<azeem>(so there 'resizepart 2 100%' is correct)

<azeem>well, my image is the Debian 2025 image, so maybe something has indeed changed

<Alicia>ah, I used the latest from https://cdimage.debian.org/cdimage/ports/latest/hurd-amd64/debian-hurd.img.tar.gz

<azeem>right

<azeem>you're right, the current image has the main parition as first extended partition, so I guess it should be 'resizepart 5 100%' there

<Alicia>you also need to resize the extended partition first

<Alicia>I forgot what the error I got otherwise was, but it was slightly confusing :)

<azeem>good point

<youpi>azeem: better yet, I'll harmonize the images into having the root on part2

<youpi>and being extended partition was a mistake, I'll fix that too

<azeem>+1

<Alicia>+1

<etno>[libirqhelp] Is there a special reason for the argument to irqhelp_server_loop() to be a (void *) instead of a (struct irq *) ?

<etno> https://cgit.git.savannah.gnu.org/cgit/hurd/hurd.git/tree/libirqhelp/irqhelp.h#n38

<youpi>etno: so it can be given to pthread_create?

<etno>I knew their had to be sthg I was missing, thanks youpi

<etno>there*

<Pellescours>I’m trying to fix the git test suite. I wasn’t setup to have all tests running (some were skipped), but on the tests that I ran, only one is failing "t5562-http-backend-content-length.sh" step 13 and 14.

<Pellescours>I did some debugging and the read hang instead of returning early due to EOF.

<Pellescours>I need to simplify it but I think there may be a problem in glibc

<youpi>glibc itself does extremely little about read(), it'd probably rather be the translator behind

<Pellescours>ok, I need to simplify the test, but basically, we trigger it only when the GZIP is enforced (and the file is empty or truncated)

<Alicia>youpi: something seems to have gone wrong with grub since skipping the extended partition. I get "error: no such partition." and a "grub rescue> " prompt

<Alicia>on the latest 64bit debian hurd image

<youpi>argl

<youpi>I didn't test booting them indeed...

<Pellescours>for git, I made progress, it’s not the read the problem, it’s the server that do not send the message (I’m at gzip inflate section)

<azeem>hrm, unfortunately, the Perl5 issue/PR regarding send() and getpeername() did not go anywhere it seems: https://github.com/Perl/perl5/issues/24195 / https://github.com/Perl/perl5/pull/24201

<sam_>perl just moves kind of slowly, but you can ping at #p5p on irc.perl.org if you want

<sam_>leonerd is there

<sam_>(he is also in #perl on libera but that's more of a support channel)

<azeem>I think I'll first update the PR with my last proposal on the issue, i.e. setting `${*$sock}{'io_socket_peername'}` during configure() instead of messing with send()

<yelninei>RT signals are not supported right? What would be the "correct" way for something that wants to use SIGRTMIN and SIGRTMIN + 1?

<youpi>it depends what it wants to use it for

<Pellescours>youpi: I think I know what is the problem for git

<Pellescours>The test does "read(fd, buf, len);" with fd = 0 and len = 0

<Pellescours>In linux the read returns, but in hurd it hang

<youpi>what is on stdin?

<Pellescours>the test reads a truncated gzip file, the first iteration of the loop consume the whole input but as inflate returns `CONTINUE`, the loop continue and then tries to read what remains (so 0 byte)

<Pellescours>that’s what I understand of the test. I’m currently trying to reproduce it with a simple test to confirm

<yelninei>control signals for a gc to stop/resume, looking in the history these were SIGUSR1 and SIGUSR2 before, i guess i can just move back to that

<youpi>yelninei: probably

<gnucode>morning people!

<Alicia>good morning gnucode !

<etno>Any Hurd presentation planned for JDLL ? (FR - https://jdll.org/)

<jab>etno: you're probably an accomplished enough hurd developer that you could present.

<etno>Maybe there is already something planned ?

<nexussfan>I'm gonna bring my Hurd t420 to a librelocal :D

<etno>(The submission deadline is tomorrow)

<jab>nexussfan: that sounds pretty cool!

<jab>you should let some of 'em play ballz.

<nexussfan>jab: you going to a librelocal?

<jab>is that a local FSF meetup type deal ?

<nexussfan> https://libreplanet.org/wiki/Group:Librelocal/2026

<jab>I don't actually know the nearest one is to me...

<jab>ehh, it's a little too far away.

<nexussfan>there's one where i live though :D

<jab>well cool!

<jab>oh by the way....I reached out to rms about buying a hurd laptop. apparently he uses icecat. and won't run the hurd unless it runs icecat.

<nexussfan>icecat?

<nexussfan>oh the browser

<jab>yup.

<nexussfan>so first we would need to port firefox i assume

<jab>we have had firefox working in the past.

<nexussfan>what happened?

<jab>maintaining the port is probably a lot of work.

<nexussfan>oh

<nexussfan>i assume rms doesn't care much about if it's an older icecat

<jab>could be...idk.

<Alicia>oh jab, which version of qemu are you using where 64bit doesn't work without -M q35?

<jab>I personally think ladybird is a better web browser to target. if we are trying to target a "modern" web browser.

<nexussfan>maybe

<jab>Alicia: whatever qemu guix packages.

<jab>I can't tell you at the moment, because I'm on OpenBSD at the moment.

<Alicia>ah, I will test that

<jab>Alicia: are you able to run the 64 bit image in qemu without -M q35 ?

<Alicia>yeah, on both of the distros I've tried (Parabola and Debian)

<sam_>ladybird is moving extremely quickly, can't decide which language it wants to use, and isn't functional yet

<nexussfan>exactly

<sam_>it's not even packagable

<nexussfan>first it was the apple one, then rust, then whatever it is now

<nexussfan>gnuzilla seems much more portable though

<jab>I just feel like Firefox's leadership seems to be steering firefox is a bad direction. I personally believe that ladybird will outlive firefox. as much as I like firefox.

<nexussfan>> GNU version of the Firefox browser. Its main advantage is an ethical one

<nexussfan>If mozilla goes bad gnuzilla will still work

<sam_>jab: sure, but that's not the same thing as whether it's suitable for porting or spending effort on now, or relevance to Hurd :)

<sam_>nexussfan: indeed

<nexussfan>Just it won't have the latest html5/js/whatever stuff

<sam_>plus you have the ESR versions of FF which mean you're not rebasing aggressively (which gnuzilla tracks)

<nexussfan>i wonder if i can package gnuzilla for debian

<nexussfan>where is the tar.gzips

<nexussfan>> Guix can also be used to build IceCat source tarballs that are suitable for building IceCat natively for other operating systems.

<nexussfan>So there isn't any

<Alicia>I guess you have to use the git repository. https://cgit.git.savannah.gnu.org/cgit/gnuzilla.git

<nexussfan>okay

<etno>youpi: based on the interactions/interest you received at FOSDEM, do you think that a presentation at JDLL would be make sense ?

<youpi>if you are happy to do it, it'll be welcome yes

<etno>That would be my first show 😅

<youpi>everybody has started with a first show ;)

<youpi>feel free to reuse my slides material :)

<Alicia>hm, with qemu 10.2.0 on guix 64bit hurd boots but then gets "kbd: queue full" when I try to type a username (at first there's just no input showing up), but this happens both with and without -M q35. maybe it's some weirdness caused by running a VM inside a VM?

<Alicia>the most apparent difference is that the guix VM doesn't have kvm for the hurd VM, so I tried it without kvm on a known working host but it did not replicate the issue

<nexussfan>wait how does stallman use icecat

<nexussfan>it's not packaged in trisquel anymore iirc

<nexussfan>now they use Abrowser

<nexussfan>sneek: later tell jab how does stallman use icecat if not packaged in trisquel anymore iirc. now they use Abrowser

<nexussfan>sneek later tell jab how does stallman use icecat if not packaged in trisquel anymore iirc. now they use Abrowser

<nexussfan>whatever

<nexussfan>sneek later tell jab https://logs.guix.gnu.org/hurd/2026-03-14.log#213340

<Alicia>sneek seems to have disappeared

<nexussfan>yeah

<nexussfan>it's down(?)

<nexussfan>never seen sneek down before

<nexussfan>is sneek hosted on darnassus or something?