IRC channel logs

<GladFan>Hello, I have installed Debian Hurd on my machine. Wahoo!

<GladFan>ZhaoM: Thanks, I can access the machine through ssh. I am still trying to set up my keyboard which was partially working in the debian hurd livecd.

<ZhaoM>happy hacking

<azert>I’m reading about pagers and tmpfs. Would it be possible to make tmpfs use a different pager then the default one?

<azert>The reason I ask this is that swap seems to be limited to 2gb on Hurd and 64gb on Linux

<azert>While I think that one may want a tmpfs that is bigger than ram

<gnucode>Well I updated my 32-bit hurd today.

<sneek>gnucode, you have 1 message!

<sneek>gnucode, ZhaoM says: thanks for the suggestion, it seems vim has a similar plugin called any-jump. Maybe they should all be covered in the wiki, because I think it can make the hurd development environment more friendly to new contributors

<gnucode>I seem to have some more bugs. :)

<gnucode>Alt + Right Arrow in the console used to switch consoles

<gnucode>tty1, tty2 etc.

<gnucode>That does not seem to be working at the moment.

<gnucode>I believe that I had configured the dvorak layout to work on the console, and now it is back to qwerty...

<gnucode>I also can't seem to start X

<gnucode>When I login to the console as my regular user the prompt string is weird.

<gnucode>It used to say

<gnucode>joshua@dobby $

<gnucode>Now it says something like

<gnucode>2354? $

<gnucode>also I tried starting emacs in the console....

<gnucode>Emacs starts, but was unusable very quickly. I got garbage on the screen. I was able to successfully quit emacs via C-x C-c

<gnucode>So basically at the moment, my T43 is stuck on console mode with no X. I bet vim probably works, but I have not tested that.

<gnucode>hmmmm tty used to say "dev/tty1"

<gnucode>no it say /dev/console

<gnucode>maybe I am using the Mach console and not the Hurd console?

<gnucode> hmmm /etc/default/hurd-console has ENABLE=true

<gnucode>sneek later tell paulz The Hurd works well on an T60. https://guix.gnu.org/en/blog/2024/hurd-on-thinkpad/ Of course debian GNU Hurd will work better than guix will on the Hurd

<sneek>Okay.

<youpi>gnucode: all these symptoms are because of not using the hurd console, yes

<youpi>sneek: later tell ZhaoM I don't really understand your discussions. vim has integrated function-jumping support: :ta

<sneek>Okay.

<gnucode>youpi: I will try to start it manually then. Thanks!

<gnucode>hmmm I could not manually start the Hurd console

<gnucode> # console -d vga -d pc_mouse --repeat=mouse -d pc_kbd --repeat=kbd -d generic_speaker -c /dev/vcs

<gnucode>It said something about vga not working.

<youpi>"something" is not precise enough as a bug report

<gnucode>give me a second...I will give you a second.

<gnucode>console: Starting driver vga failed: invalid argument

<gnucode>sorry. No idea what I meant by "I will give you a second."

<gnucode>I'm working, and answering phone calls at the same time as chatting.

<gnucode>first I tried running:

<gnucode> $ sudo console -d vga -d pc_mouse --repeat=mouse -d pc_kbd --repeat=kbd -d generic_speaker -c /dev/vcs

<gnucode>I got this error:

<gnucode>console: Starting driver vga failed: invalid argument

<gnucode>then I ran

<gnucode>$ sudo console -d pc_mouse --repeat=mouse -d pc_kbd --repeat=kbd -d generic_speaker -c /dev/vcs

<gnucode>and my Hurd machines seems to be stuck? The Hurd console does not seem to be loading

<youpi>you do want the vga driver, otherwise you won't be able to see anything :)

<youpi>it is loaded, just not displaying anything

<gnucode>well that stinks!

<youpi>you asked for it by dropping the vga driver :)

<gnucode>hahaha! C-alt-del brought me back to the mach console

<youpi>I'd say run this through gdb:

<youpi>gdb --args console -d vga -d pc_mouse --repeat=mouse -d pc_kbd --repeat=kbd -d generic_speaker -c /dev/vcs

<youpi>put a breakpoint on vga_display_start, and go through it step by step to see where it stops with EINVAL

<gnucode>I will give that a try in a few thanks!

<gnu_srs>Good news: I managed to build libdrm again followed by a successful build of mesa _with all the same options as on GNU/Linux_. Only 2/104 tests fail :)

<gnucode>gnu_srs: that's awesome news!

<gnu_srs>gnucode: Tks!

<gnucode>sudo gdb --args console -d vga ...

<gnucode>b vga_display_start

<gnucode>vga_display_start is not defined as a funtion.

<gnucode>I probably need to download the debugging files for console...

<youpi>that's the hurd-dbgsym package

<gnucode>thanks

<gnucode>apparently I already had hurd-dbgsym already installed.

<youpi>ah but it's a symbol in a dynamically-loaded module, so it's expected that it's not available initially, no problem

<youpi>gdb will catch the module loading and find it there

<gnucode>gotcha

<gnu_srs>And youpi doesn't care about mesa/libdrm :(

<youpi>I don't have time to care for everything

<gnucode>what makes you think he doesn't care?

<gnucode>I don't know how useful this is....

<gnucode>sudo gdb --args console -d vga ...

<gnucode>b vga_display_start

<gnucode>with a lot of step and next

<gnucode>I see this a lot:

<gnucode>../../console-client/driver.c: NUMBER

<youpi>you can "apt source hurd" to get the source

<gnucode>warning: NUMBER ../../console-client/console.c: no such file for directory

<gnucode>I see numbers 55, 661, 654, and 725 is the last

<gnucode>I do have the source code locally...oh

<gnucode>you are suggesting that I run

<gnucode>sudo gdb --args ./path/to/console -d vga ...

<gnucode>let me try that

<youpi>no

<youpi>I mean you want to first run apt source hurd

<youpi>to get the source code

<youpi>so that gdb can show you the source code lines

<gnucode>kk

<gnucode>I bet the main problem here is that I do not really know how to use gdb well. I am still not seeing the source code lines. I will take lunch and try to watch a youtube tutorial. See what happens.

<youpi>probably you just need to "cd" to a directory from which the ".../../console-client/console.c" makes sense

<gnucode>gotcha

<gnucode>OK....

<gnucode>I believe that I "cd"ed to the right directory.

<gnucode>if I type in "next" a lot, then I get to the error pretty quickly.

<gnucode>the last line that I can get to is:

<gnucode>../../console-client/console.c 723

<gnucode>fs_help_touch (&dir->nn_stat, TOUCH_ATIME, console_maptime);

<gnucode>I can step into that function, and I get some weird errors

<gnucode>or seemingly errors

<gnucode>../sysdeps/mach/hurd/htl/pt-mutex-lock.c 30 : no such file or directory

<youpi>that's inside libc

<youpi>you don't need to step into that

<youpi>don't use next from main

<youpi>put the breakpoint on vga_display_start

<youpi>and it'll end up in there when it does

<gnucode>anyway my gdb seems to be stuck now. C-c doesn't need to be unlocking it.

<gnucode>and I did type in "b vga_display_start"

<youpi>if you step into a critical section, the application won't answer to gdb debugging requests

<gnucode>then I think I did "run" or "start"

<gnucode>gotcha. I think I may have to hard shutoff...

<youpi>run or start?

<youpi>that makes a lot of difference

<gnucode>I know...I can't actually see what I typed in, because the screen doesn't show it anymore.

<gnucode>I used both "run" then "start"

<gnucode>or maybe the other way around

<youpi>start puts a breakpoint main

<gnucode>one showed C code and showed the program execution

<youpi>you don't want that, or else you want to use continue to get to the breakponit inside the vga driver

<gnucode>man. I really need to learn some gdb stuff.

<gnucode>It does not appear the gdb is tripping up on that breakpoint.

<gnucode>sudo gdb --args console -d vga ...

<gnucode>b vga_display_start

<gnucode>run

<gnucode> " /usr/bin/console: Starting driver vga failed: Invalid argument

<gnucode>warning: illegal rpc: do_mach_notify_send_once

<gnucode>( New bogus thread id 0 )

<gnucode>Cat fetch registers from thread bogus thead id 0

<gnucode>next

<gnucode>can't fetch registers from thread bogus thread id 0

<youpi>ah, it doesn't manage to catch vga_display_start apparently

<gnucode>It seems to be the case. :(

<youpi>you can probably break on driver_start instead

<youpi>just make sure not to step inside the mutex_lock function

<youpi>just step inside the ->start call

<gnucode>i will try

<gnucode>ok. ../../console-client/driver.c

<gnucode>I stepped in line 254 if (driver_list[i].ops->start)

<gnucode>then I got to line 263

<gnucode>return err

<gnucode>return err;

<gnucode>p err

<gnucode>$2 = EINVAL;

<gnucode>I don't know how helpful that is...

<youpi>it's not on the if line that you want to step

<youpi>but on the ->start call

<gnucode>let me try again. I thought I had it, because I saw a variable that had EINVAL. I appreciate you holding my hand through all of this, but if you have better stuff to do, it's totally find to ignore me. :)

<gnucode>well I have to run soon. I will try to investigate some other time.

<gnucode>thanks for all the help

<gnu_srs>youpi: You could have said congrats. Not: (17:11:04) youpi: I don't have time to care for everything

<gnu_srs>It is called social behaviour. Maybe you re not interested in the patches. Then I'll keep them for myself :(

<youpi>I don't have time to backlog everything, yes

<gnu_srs>Go back to your cave and stay there :)

<gnu_srs>I was not a backlog, and you know it!

<youpi>?

<youpi>never mind

<youpi>really, I don't have time

<youpi>my inbox is *full* of stuff to deal with

<gnu_srs>youpi: https://darnassus.sceen.net/~hurd-web/contributing/: Fix building mesa. Contributing or not?

<youpi>that's useful to do, sure

<youpi>and thanks for doing it

<youpi>just do not expect me to always notice stuff happening on irc, I just cannot backlog everything

<gnu_srs>Its as easy as congrats, nothing else :)

<gnu_srs>In the middle of a conversation.

<azert>gnu_srs: congrats!

<azert>I wonder if you plan to submit

<azert>upstream?

<gnu_srs>Howdy. I've managed to upgrade my Hurd image to current sid. And now I would like to create some images for amd64. Any good ideas?

<azert>Which kind of ideas?

<gnu_srs>azert: Yes, these changes will be supplied both upstream and to Debian.

<azert>Can you get the mesa software renderer to run?

<gnu_srs>azert: I could start with an image already containing usrmerge, and would like to know how to get rid of it.

<azert>Ah, Im not a Debian expert

<azert>Isn’t usrmerge just copying some files around?

<gnu_srs>azert: I've not tried mesa software rendering yet, working on patching xserver-xorg-video-vmware

<gnu_srs>azert: No, usrmerge create symlinks of /bin, /sbin, /lib, etc to /usr :(

<azert>gnu_srs: ok but effectively that results in moving the file in /usr to / merging them together

<azert>I known someone got pissed off by that since it interferes with their way of doing things

<youpi>gnu_srs: debootstrap or mmdebstrap should be able to produce an image

IRC channel logs

2025-01-26.log