IRC channel logs

<jack-sparrow>hi

<nexussfan>Hello

<jack-sparrow>i try to create #gentoo-hurd

<jack-sparrow>if some want follow me

<nexussfan>gentoo/hurd is kinda dead

<jack-sparrow>i'd like restart

<jack-sparrow>kinda ?

<nexussfan> https://hurdos.com/wiki/hurd/running/distrib.html

<jack-sparrow>nexussfan: you like help me to this project ?

<damo22>(gdb) bt

<damo22>#0 task_terminate (task=0xffffffffdc0dec18) at ../kern/task.c:291

<damo22>#1 0xffffffff81088cf8 in exception_no_server () at ../kern/exception.c:260

<damo22>#2 0xffffffff81089c46 in exception_try_task (_exception=_exception@entry=0x1, code=<optimized out>, subcode=<optimized out>) at ../kern/exception.c:177

<damo22>#3 0xffffffff81089d48 in exception (_exception=_exception@entry=0x1, code=code@entry=0x1, subcode=subcode@entry=0x7ffffffdfff8) at ../kern/exception.c:107

<damo22>#4 0xffffffff81072bae in i386_exception (exc=exc@entry=0x1, code=0x1, subcode=0x7ffffffdfff8) at ../i386/i386/trap.c:649

<damo22>#5 0xffffffff81073065 in user_page_fault_continue (kr=kr@entry=0x1) at ../i386/i386/trap.c:122

<damo22>#6 0xffffffff81069eee in vm_fault (map=<optimized out>, vaddr=vaddr@entry=0x7ffffffdf000, fault_type=0x3, change_wiring=change_wiring@entry=0x0, resume=resume@entry=0x0, continuation=continuation@entry=0xffffffff81072fc0 <user_page_fault_continue>) at ../vm/vm_fault.c:1486

<damo22>#7 0xffffffff81072def in user_trap (regs=0xffffffffdc0d0c18) at ../i386/i386/trap.c:525

<damo22>#8 0xffffffff8103dcea in _take_trap () at ../x86_64/locore.S:691

<damo22>page fault in userspace with address 0x7ffffffdfff8

<damo22>(gdb) p *regs

<damo22>$2 = {r15 = 0x0, r14 = 0x0, r13 = 0x8, r12 = 0x30, r11 = 0x246, r10 = 0xffffffff, r9 = 0x0, r8 = 0x8, edi = 0x7ffffffe0040,

<damo22> esi = 0x3, ebp = 0x7ffffffe0020, cr2 = 0x7ffffffdfff8, ebx = 0x7, edx = 0x50, ecx = 0x30, eax = 0xc8800000000, trapno = 0xe,

<damo22> err = 0x6, eip = 0x4344a5, cs = 0x1f, efl = 0x10246, uesp = 0x7ffffffe0000, ss = 0x17}

<damo22>(gdb) x 0x7ffffffe0000

<damo22>0x7ffffffe0000: 0x20002002

<damo22>(gdb) x 0x7ffffffdfff8

<damo22>0x7ffffffdfff8: Cannot access memory at address 0x7ffffffdfff8

<damo22>is the stack in the wrong place in userspace?

<damo22>maybe the problem is ebp is set past the stack?

<damo22>no no

<damo22>stack goes from 0x7ffffffe0000 -> 0x7fffffffffff

<damo22>but it grows down

<damo22>i think pci-arbiter ran out of userspace stack?

<Pellescours>like it malloc too much? or like a stack overflow?

<damo22>uesp = 0x7ffffffe0000 is bad

<damo22>that is exactly the last address in the stack so you cannot push more

<damo22>looks like userspace stack overflow?

<Pellescours>is there a recursive call in pci-arbiter? (I don’t think userspace is the problem as it worked perfectly before, but just in case)

<Pellescours>damo22: when pci-arbiter start what is the stack value? (breakpoint in main if you can)

<damo22>how do i break into a user thread from bootstrap?

<damo22>im running gdb on gnumach

<damo22>wrapped around the entire qemu

<Pellescours>break in the kernel when the task spawn and check it’s values?

<damo22>help welcome

<damo22>im on * 2665a507 (HEAD, origin/smp64-upstream) x86_64: Remove PERCPU_DS selector settings

<damo22> https://git.zammit.org/gnumach-sv.git/log/?h=smp64-upstream

<Pellescours>I’ll try that after I finished my night, going back to bed. But nice progresses btw

<Pellescours>(your last commit is very similar to gnumach master)

<damo22>ok will rebase later

<jack-sparrow>nexussfan: hi here

<jack-sparrow>what miss for do working pipewire and gnome ?

<damo22>there is no audio driver yet

<damo22>i am trying to see if i can port one pci ALSA driver as a JACK backend

<damo22>if that works okay, it may be worth investing time to port the rest

<damo22>i chose ALSA because i contributed there a fair bit

<damo22>it has the largest range of sound card support in the free world

<nexussfan>I like ArneBab's idea for sound on hurd

<nexussfan>settrans play /hurd/mediaplayer_play

<nexussfan>cp song1.ogg song2.ogg play

<damo22>well that is ok for playing a queue of songs

<damo22>but im more interested in pro audio recording

<jack-sparrow>what is settrans ?

<nexussfan>Yeah I guess

<nexussfan>But for playing audio it's a nice hurdish way

<damo22>jack-sparrow: hurd has a concept called translators you should read about

<damo22>when you open() read() write() a filesystem node, you can have installed a translator on that node that runs to provide the translation of those calls

<jack-sparrow>i'd like, have you a link

<damo22> https://www.debian.org/ports/hurd/hurd-doc-translator

<jack-sparrow>i will be try to ported hurd on gentoo follow my way on #gentoo-hurd

<jack-sparrow>damo22: thank i will see and thank to mistral.ai for translation it's pratice

<damo22>any user program can be written and executed to provide the implementation of accessing a filesystem node

<damo22>thats what makes hurd infinitely extensible and interesting

<jack-sparrow>great !

<damo22>you know how Linux kernel usually provides nodes such as /dev/* ?

<damo22>in hurd these are just normal filesystem nodes, and translators are used in userspace to provide the hardware access

<jack-sparrow>Can two or more Hurd kernels communicate with each other over the network with a translator?

<damo22>yes

<nexussfan>really? wow

<jack-sparrow>serious ?

<damo22>well nfs for example

<nexussfan>Oh

<damo22>but i dont see why more cant be written

<nexussfan>What happened to a distributed hurd?

<nexussfan>Where different parts of the Hurd are ran on different machines; on the network for example

<jack-sparrow>can interconnect hurd system like a mesh

<jack-sparrow>perpas for share ALU ressouces

<damo22>you could probably write a network pager to allow memory to be shared over the network but it would be slow

<nexussfan>Then you could have a super-hurd computer like boinc

<jack-sparrow>Like a team leader who gets a team to work.

<jack-sparrow>what is boinc ?

<damo22>mpi could be used

<damo22>over ssh

<nexussfan>Or like a distcc

<jack-sparrow>great

<jack-sparrow>what is mpi ?

<jack-sparrow>hurd need more interested people

<nexussfan>True

<damo22> https://www.open-mpi.org/ i think this is the site

<nexussfan>> A filesystem-based package manager

<nexussfan>This would be awesome

<nexussfan>The first hurd-only package manager(?)

<damo22>you could probably write a filesytem based package browser for .deb

<nexussfan>That would be nice

<damo22>so you just ls debian/testing/blah

<nexussfan>It's funny how there's a perl library for HURD translators but no python one

<nexussfan>You can tell hurdextras is pretty old

<damo22>but you dont want to fetch the package list live every time

<damo22>it overloads the traffic of the server

<nexussfan>True

<damo22>so you want to make your filesystem deb browser use your local cache

<jack-sparrow> https://www.open-mpi.org/faq/?category=buildcuda

<nexussfan>> CUDA-aware

<jack-sparrow>need hurd use for AI system

<nexussfan>None of that on the HURD

<nexussfan>Ollama doesn't work on the HURD afaik

<nexussfan>maybe llama.cpp

<jack-sparrow>need port

<jack-sparrow>seems interst industrial

<jack-sparrow>interest*

<nexussfan>How would I make python bindings for trivfs

<jack-sparrow>imagine computer use hurd can share ressouce for AI system no need big DC

<jack-sparrow>what is trivfs ?

<gfleury>For distributed computing, take a look the work from brent https://www.freesoft.org/blogs/soapbox/hurd

<nexussfan>> It looks like nothing was found at this location. Maybe try a search?

<gfleury> https://www.freesoft.org/blogs/soapbox/

<gfleury>And go to hurd

<gfleury>That should work

<jack-sparrow>i pass my night on irc well well

<jack-sparrow>just debian/hurd is active now ?

<jack-sparrow>gentoo/hurd is tring with a poor men and arch/hurd no maj since 2019

<nexussfan>debian/hurd and guix/hurd

<jack-sparrow>guix seems only package manager

<aculnaig>Good morning

<damo22>uesp = 0x7ffffffe0000 is bad

<damo22>should i just try increasing the user stack size?

<azert>damo22: check if anything uses recursion

<azert>stack exhaustion is often a result of an infinite recursive loop

<azert>perhaps some error handling path?

<damo22>its a bit difficult to know if there is any recursion in a translator just by reading the source, there are many functions, i need a backtrace

<damo22>maybe i can break on task_terminate, and then dig into the user stack?

<jab>jack-sparrow: guix/hurd is still active. Most of their developers target GNU/Linux, but they have a few interested in the Hurd: https://guix.gnu.org/en/blog/2024/hurd-on-thinkpad/

<jab>If you're interested in alpine Linux, then you might talk to sergey. He wanted to start an alpine Hurd distro at some point.

<jack-sparrow>jab: i will try to port on gentoo

<sobkas>So

<sobkas>#define _IOT_sync_merge_data _IOT(_IOTS(struct sync_merge_data), 1, 0, 0, 0, 0)

<sobkas>is ok?

<jab>jack-sparrow: best of luck!

<sobkas>I have made definitions of _IOT_ and after an install it started to behave strangely, ie. glxgears window become full screen, but after a reboot it works ok, even a bit faster... ???

<sobkas>~120fps

<jab>sobkas: sounds pretty cool!

<sobkas>so I have problem with nfs translator

<sobkas>settrans: /hurd/nfs: Translator died

<sobkas>\/hurd/nfs: rpc

<sobkas>settrans: /hurd/nfs: Translator died

<sobkas>q++

<sobkas>++

<sobkas>sorry problems with keyboard

<azert>aculnaig: what do you think about the httpfs translator?

<aculnaig>azert: it has some potential but needs a deep rewrite.

<azert>is it true that in order to open a directory, it assumes that none of the up directories in the full path can be 404?

<azert58>does it support WebDAV?

<aculnaig>azert: to answer to your first question, it does do a traversal up to the file, checking every parent, and if it encounters a 404, it returns an error

<azert58>that’s the bug I remember

<aculnaig>azert: to answer to your second question, no, it does not support WebDAV.

<azert58>not an easy fix by the way

<azert58>I think it needs changing to libnetfs

<aculnaig>if you want to look at it together we can pair this friday

<aculnaig>at the moment i am writing a fix to let stat works correctly

<aculnaig>the last patch i am writing from the mailing list of today

<azert58>I think it’s better that you just keep doing what you feel like

<aculnaig>ok

<azert58>by the way, this might interest you, it’s a well maintained LGPL WebDAV c++ library https://davix.web.cern.ch/davix/docs/devel/lib-examples.html

<aculnaig>azert58: thank you

<jab>aculnaig: it would be cool if httpfs supported sitemaps. I believe WordPress supports that...probably a fair number of other sites.

<jab> https://www.sitemaps.org/

<azert>I think that there are too many ways one could write an httpfs translator

<azert>depending on how one want to use it

<jab>azert: my biggest issue with network translators (https is probably one of the worst), is that if

<jab>I have httpfs set up on ~/gnucode.org/, then cd ~/gnucode.org; ls;

<jab>hangs my system for 30+ seconds.

<jab>hangs my shell rather.

<azert>yes that is very annoying

<azert>but nfs has that issue on Linux too, you just need a good server and a decent connection

<azert>just figured out that davix is based on this library https://notroj.github.io/neon/

<azert>that is in C

<azert>jab: I am convinced that the “right” way to do networking filesystems is that Dropbox way

<azert>and both MacOS as well as Windows partially moved in that direction

<azert>that is that you want a translator that uses the parent filesystem as a cache, and has a distributed network backhand that store chunk of data by their hash

<azert>I think Cephfs in Linux is the professional version of this concept

<jab>hmm. networking is something I want to learn more about. :)

<youpi>Cleartext-Password

<youpi>EWIN :)

<jab>hahaha!

<Gooberpatrol66>does ceph have client caching? never heard of that

IRC channel logs

2026-01-13.log