Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.bugs.dist > #1189291 > unrolled thread

Bug#1065570: Interface border is replaced by ascii chars

Back to article view | Back to linux.debian.bugs.dist

Contents

  Bug#1065570: Interface border is replaced by ascii chars x11r <x11r@proton.me> - 2024-03-06 20:10 +0100
    Bug#1065570: Interface border is replaced by ascii chars Samuel Thibault <sthibault@debian.org> - 2024-03-06 23:40 +0100
      Bug#1065570: Interface border is replaced by ascii chars x11r <x11r@proton.me> - 2024-03-06 23:50 +0100
        Bug#1065570: ***UNCHECKED*** Re: Bug#1065570: Interface border is replaced by ascii chars Samuel Thibault <sthibault@debian.org> - 2024-03-07 00:00 +0100
        Bug#1065570: Interface border is replaced by ascii chars Samuel Thibault <sthibault@debian.org> - 2024-03-07 00:10 +0100
          Bug#1065570: Interface border is replaced by ascii chars x11r <x11r@proton.me> - 2024-03-07 00:40 +0100
            Bug#1065570: Interface border is replaced by ascii chars Samuel Thibault <sthibault@debian.org> - 2024-03-07 01:10 +0100
              Bug#1065570: Interface border is replaced by ascii chars x11r <x11r@proton.me> - 2024-03-07 03:20 +0100
                Bug#1065570: Interface border is replaced by ascii chars Samuel Thibault <sthibault@debian.org> - 2024-03-09 22:10 +0100
                  Bug#1065570: Interface border is replaced by ascii chars Samuel Thibault <sthibault@debian.org> - 2024-03-09 22:50 +0100
                    Bug#1065570: Interface border is replaced by ascii chars Samuel Thibault <sthibault@debian.org> - 2024-03-09 22:50 +0100

#1189291 — Bug#1065570: Interface border is replaced by ascii chars

Package: debian-installer
Version: 20230607+deb12u5

I've been working with a small team at my college trying to develop a tool to automate PXE booting and installation. We have targeted Debian as the first OS we want to get working over PXE boot. On every test we've ran of the Debian installation process, there's a visual bug that occurs late in the installation process. Effectively, the border of the progress interface turns into Çâö repeatedly. You can see an image of it in the issue here https://github.com/xeluior/genisys/issues/82

This happens every install. We have tested it on a small selection of hardware (I don't have the details on hand right now) and in virtual machines (VirtualBox). The debian-installer version is the one bundled with https://deb.debian.org/debian/dists/bookworm/main/installer-amd64/current/images/netboot/netboot.tar.gz

I'm not sure where to start looking for this so let me know what other information would be useful.

Thank you,
Robert Gingras

[toc] | [next] | [standalone]

#1189313

x11r, le mer. 06 mars 2024 18:59:50 +0000, a ecrit:
> I've been working with a small team at my college trying to develop a tool to automate PXE booting and installation. We have targeted Debian as the first OS we want to get working over PXE boot. On every test we've ran of the Debian installation process, there's a visual bug that occurs late in the installation process. Effectively, the border of the progress interface turns into Çâö repeatedly. You can see an image of it in the issue here https://github.com/xeluior/genisys/issues/82

That looks like an utf-8 encoding issue. Please tell us exactly how you
boot it (kernel parameters and how you access the console).

Samuel

[toc] | [prev] | [next] | [standalone]

#1189315

The machine is PXE booted with the kernel parameters: initrd=debian-installer/amd64/initrd.gz ip=dhcp auto=enable language=en country=US locale=en_US.UTF-8 keymap=ansi hostname=debian domain="" url=tftp://10.0.0.1/preseed.cfg

Console is accessed over VGA. VMSVGA for the VirtualBox tests we've ran and a VGA KVM for the tests on bare metal.

On Wednesday, March 6th, 2024 at 5:28 PM, Samuel Thibault <sthibault@debian.org> wrote:

> x11r, le mer. 06 mars 2024 18:59:50 +0000, a ecrit:
> 
> > I've been working with a small team at my college trying to develop a tool to automate PXE booting and installation. We have targeted Debian as the first OS we want to get working over PXE boot. On every test we've ran of the Debian installation process, there's a visual bug that occurs late in the installation process. Effectively, the border of the progress interface turns into Çâö repeatedly. You can see an image of it in the issue here https://github.com/xeluior/genisys/issues/82
> 
> 
> That looks like an utf-8 encoding issue. Please tell us exactly how you
> boot it (kernel parameters and how you access the console).
> 
> Samuel

[toc] | [prev] | [next] | [standalone]

#1189316 — Bug#1065570: ***UNCHECKED*** Re: Bug#1065570: Interface border is replaced by ascii chars

Hello,

Please re-send your mail unencrypted, so everybody can read the
information.

Samuel

[toc] | [prev] | [next] | [standalone]

#1189317

Hello,

x11r, le mer. 06 mars 2024 22:39:23 +0000, a ecrit:
> The machine is PXE booted with the kernel parameters: initrd=debian-installer/amd64/initrd.gz ip=dhcp auto=enable language=en country=US locale=en_US.UTF-8 keymap=ansi hostname=debian domain="" url=tftp://10.0.0.1/preseed.cfg
> 
> Console is accessed over VGA. VMSVGA for the VirtualBox tests we've ran and a VGA KVM for the tests on bare metal.

I still cannot reproduce with unpacking the tarball and running

kvm -net nic -net user,tftp=$PWD,bootfile=pxelinux.0 -drive -m 1G

and passing the kernel parameters.

Does the mangling happen right from the first screen, or after some
steps?

It could be useful to share your preseed.cfg (take care of removing any
hardcoded password).

Samuel

[toc] | [prev] | [next] | [standalone]

#1189321

The mangling does not happen immediately. It happens after the "Installing the base system" step (not sure what the step after that is supposed to be). Here's a pastebin of the preseed.cfg: https://pastebin.com/K7vwkpMu

On Wednesday, March 6th, 2024 at 6:06 PM, Samuel Thibault <sthibault@debian.org> wrote:

> Hello,
>
> x11r, le mer. 06 mars 2024 22:39:23 +0000, a ecrit:
>
> > The machine is PXE booted with the kernel parameters: initrd=debian-installer/amd64/initrd.gz ip=dhcp auto=enable language=en country=US locale=en_US.UTF-8 keymap=ansi hostname=debian domain="" url=tftp://10.0.0.1/preseed.cfg
> >
> > Console is accessed over VGA. VMSVGA for the VirtualBox tests we've ran and a VGA KVM for the tests on bare metal.
>
>
> I still cannot reproduce with unpacking the tarball and running
>
> kvm -net nic -net user,tftp=$PWD,bootfile=pxelinux.0 -drive -m 1G
>
> and passing the kernel parameters.
>
> Does the mangling happen right from the first screen, or after some
> steps?
>
> It could be useful to share your preseed.cfg (take care of removing any
> hardcoded password).
>
> Samuel

[toc] | [prev] | [next] | [standalone]

#1189323

Hello,

x11r, le mer. 06 mars 2024 23:29:24 +0000, a ecrit:
> The mangling does not happen immediately. It happens after the "Installing the base system" step (not sure what the step after that is supposed to be). Here's a pastebin of the preseed.cfg: https://pastebin.com/K7vwkpMu

I still cannot reproduce with:

mkdir tmp
cd tmp
wget  https://deb.debian.org/debian/dists/bookworm/main/installer-amd64/current/images/netboot/netboot.tar.gz
tar xf netboot.tar.gz
wget -O preseed.cfg https://pastebin.com/raw/K7vwkpMu
dd < /dev/zero > disk.img bs=1M count=1 seek=10000
kvm -net nic -net user,tftp=$PWD,bootfile=pxelinux.0 -drive file=disk.img,cache=unsafe -m 1G

and appending

ip=dhcp auto=enable language=en country=US locale=en_US.UTF-8 keymap=ansi hostname=debian domain="" url=tftp://10.0.2.2/preseed.cfg

on the boot kernel command line. It does show up fine up to the reboot
step.

Any idea what is different with your case? (except the virtualization
software)

Samuel

[toc] | [prev] | [next] | [standalone]

#1189327

Sorry, I don't mean to make you chase information. The only other difference is our pxelinux.cfg. We are using a minimal pxelinux in order to bypass the interactive menus with the following content.

default linux
label linux
  kernel debian-installer/amd64/linux
  append initrd=debian-installer/amd64/initrd.gz ip=dhcp auto=enable language=en country=US locale=en_US.UTF-8 keymap=ansi hostname=debian domain="" url=tftp://10.0.0.1/preseed.cfg

I attempted a minimal repro using a command similar to yours but using "qemu-system-x86_64 -enable-kvm" (because I can't seem to find the package that has the kvm bin) and did not get the bug. I know it's not isolated to VirtualBox because I have video of the bug happening on a bare metal install (https://youtu.be/-OiJVJp8SxQ bug occurs at about 1:05).

After some digging, I saw that the original netboot.tar.gz had pxelinux.cfg symlinked to debian-installer/amd64/pxelinux.cfg which has a default file symlinked to ../boot-screens/syslinux.cfg (realpath debian-installer/amd64/boot-screens/syslinux.cfg). The QEMU reproduction I tried I edited that file directly, whereas the other tests I've done a script has replaced it pxelinux.cfg/default with a new file (removing the symlink entirely). I modified our script-generated tftp directory (for VirtualBox) to fix the symlink (putting the above contents into the debian-installer/amd64/boot-screens/syslinux.cfg file directly) and that didn't fix the issue (I thought I was onto something).

I reset the TFTP dir to just the preseed.cfg (so not our pxelinux.cfg) and input the kernel parameters in the default menu. One thing I noticed was the default parameters include VGA=788 (which our menu above was missing). This appears to have fixed the issue. I ran the QEMU test mentioned earlier using the pxelinux.cfg above, but qualitatively it looked like it was running in a different graphical mode. With this information, I think this may be related to the absence of the "VGA=" parameter in our boot parameters. I see there's an existing bug https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=969944 regarding the same parameter, but I don't see enough information there to say it's the same issue.

That's all the relevant information I can think of for now. Maybe see if your KVM is able to reproduce using the pxelinux.cfg above or removing the "VGA=788" parameter from the kernel command line? 

On Wednesday, March 6th, 2024 at 6:57 PM, Samuel Thibault <sthibault@debian.org> wrote:

> Hello,
> 
> x11r, le mer. 06 mars 2024 23:29:24 +0000, a ecrit:
> 
> > The mangling does not happen immediately. It happens after the "Installing the base system" step (not sure what the step after that is supposed to be). Here's a pastebin of the preseed.cfg: https://pastebin.com/K7vwkpMu
> 
> 
> I still cannot reproduce with:
> 
> mkdir tmp
> cd tmp
> wget https://deb.debian.org/debian/dists/bookworm/main/installer-amd64/current/images/netboot/netboot.tar.gz
> tar xf netboot.tar.gz
> wget -O preseed.cfg https://pastebin.com/raw/K7vwkpMu
> dd < /dev/zero > disk.img bs=1M count=1 seek=10000
> 
> kvm -net nic -net user,tftp=$PWD,bootfile=pxelinux.0 -drive file=disk.img,cache=unsafe -m 1G
> 
> and appending
> 
> ip=dhcp auto=enable language=en country=US locale=en_US.UTF-8 keymap=ansi hostname=debian domain="" url=tftp://10.0.2.2/preseed.cfg
> 
> on the boot kernel command line. It does show up fine up to the reboot
> step.
> 
> Any idea what is different with your case? (except the virtualization
> software)
> 
> Samuel

[toc] | [prev] | [next] | [standalone]

#1189611

Control: reassign -1 console-setup
Control: tags -1 + d-i

Hello,

x11r, le jeu. 07 mars 2024 02:13:22 +0000, a ecrit:
> That's all the relevant information I can think of for now. Maybe see if your KVM is able to reproduce using the pxelinux.cfg above or removing the "VGA=788" parameter from the kernel command line? 

Ahah! That's indeed the trigger. I also have to pass -vga vmware to
qemu, so the linux console stays in pure text mode, no fbcon.

As I was guessing, it's the console-setup configuration that mangles
everything, we'll be able to have a look.

I couldn't reproduce it with all other parameters being the default,
I'll dig more.

Samuel

[toc] | [prev] | [next] | [standalone]

#1189622

Samuel Thibault, le sam. 09 mars 2024 22:02:46 +0100, a ecrit:
> x11r, le jeu. 07 mars 2024 02:13:22 +0000, a ecrit:
> > That's all the relevant information I can think of for now. Maybe see if your KVM is able to reproduce using the pxelinux.cfg above or removing the "VGA=788" parameter from the kernel command line? 
> 
> Ahah! That's indeed the trigger. I also have to pass -vga vmware to
> qemu, so the linux console stays in pure text mode, no fbcon.
> 
> As I was guessing, it's the console-setup configuration that mangles
> everything, we'll be able to have a look.
> 
> I couldn't reproduce it with all other parameters being the default,
> I'll dig more.

d-i debian-installer/locale string en_US

This is the eventual culprit, it should be

d-i debian-installer/locale string en_US.UTF-8

Do we actually support non-UTF-8 locales, actually?

Samuel

[toc] | [prev] | [next] | [standalone]

#1189623

Samuel Thibault, le sam. 09 mars 2024 22:42:36 +0100, a ecrit:
> Samuel Thibault, le sam. 09 mars 2024 22:02:46 +0100, a ecrit:
> > x11r, le jeu. 07 mars 2024 02:13:22 +0000, a ecrit:
> > > That's all the relevant information I can think of for now. Maybe see if your KVM is able to reproduce using the pxelinux.cfg above or removing the "VGA=788" parameter from the kernel command line? 
> > 
> > Ahah! That's indeed the trigger. I also have to pass -vga vmware to
> > qemu, so the linux console stays in pure text mode, no fbcon.
> > 
> > As I was guessing, it's the console-setup configuration that mangles
> > everything, we'll be able to have a look.
> > 
> > I couldn't reproduce it with all other parameters being the default,
> > I'll dig more.
> 
> d-i debian-installer/locale string en_US
> 
> This is the eventual culprit, it should be
> 
> d-i debian-installer/locale string en_US.UTF-8

I'm fixing the d-i manual.

Samuel

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.bugs.dist

csiph-web