Add Arch Linux toolbox image

[erazemk]

This PR builds on the previous tries (#328 and #520) of including an Arch Linux image in toolbox.

The PR adds an Arch Linux image, created from the upstream base-devel branch.

It includes most of the packages that the Fedora-based images do, but currently excludes flatpak (required for flatpak-spawn), since it adds ~690 MB of extra packages/dependencies.

The base-devel image (without any extra packages) is 729 MB, when syncing repositories 741 MB and with all the packages (excluding flatpak) 831 MB, which should be the minimum if based on archlinux:base-devel.

If the final size is very important I can try and base it off of archlinux:latest (which includes only the base packages, not base-devel), but that would still require adding some of the missing packages - although the base image size is 416 MB, which is quite an improvement..

Another option could be to separate latest and base-devel versions into separate images, similarly to what is done with Fedora versions.

I can also add the yay AUR helper if there is a need for it.

[softwarefactory-project-zuul]

Build succeeded.

• shellcheck : SUCCESS in 2m 23s
• system-test-fedora-32 : SUCCESS in 6m 52s
• system-test-fedora-33 : SUCCESS in 6m 23s
• system-test-fedora-34 : SUCCESS in 7m 56s
• system-test-fedora-rawhide : SUCCESS in 12m 00s

[xPMo]

Didn't work for me, but I figured it out.
I saw that a mount was failing by manually starting the container: podman container start --interactive --attach archlinux-toolbox

Adding this line to the Dockerfile fixed that.

RUN touch /etc/machine-id

I also added a sudoers entry for the %wheel group for my own purposes, and swapped to plocate since it's faster.

[erazemk]

I just now realized that the proper way to create and test custom images is by building them and using toolbox create --image to use them with toolbox.

I tried running the container that way, but got the error Error: invalid entry point PID of container, which was fixed by adding /etc/machine-id to the image, like you recommended, but I got the following error:

bash: PROMPT_COMMAND: line 0: syntax error near unexpected token `;'
bash: PROMPT_COMMAND: line 0: ` ; printf "\033]0;%s@%s:%s\007" "${USER}" "${HOSTNAME%%.*}" "${PWD/#$HOME/\~}"'

Do you not get this error and if so, which OS are you using?

[xPMo]

I'm not getting that error; I'm not sure where that is coming from (it doesn't match any code in toolbox afaik). Maybe that's coming from your ~/.bashrc? That may still be something that can be fixed in the container config, but more likely is just some bad assumption that your bashrc is making. What do you get from grep -A5 PROMPT_COMMAND ~/.bash*?

which OS are you using?

Latest packaged toolbox on an Arch host.

[erazemk]

Maybe that's coming from your ~/.bashrc?

It seems the problem was indeed with this, I use the fish shell, so I don't have a local bashrc file, just the system one (which I haven't changed). This is weird though, since I based the Dockerfile off of the Fedora files, which work normally with toolbox. Copying the root user's bashrc doesn't help either, I still get the same error (even after removing the part that sources /etc/bashrc).

That may still be something that can be fixed in the container config, but more likely is just some bad assumption that your bashrc is making. What do you get from grep -A5 PROMPT_COMMAND ~/.bash*?

When replacing ~/.bash* with /etc/bashrc I get the following output:

   if [ -z "$PROMPT_COMMAND" ]; then
      case $TERM in
      xterm*|vte*)
        if [ -e /etc/sysconfig/bash-prompt-xterm ]; then
            PROMPT_COMMAND=/etc/sysconfig/bash-prompt-xterm
        elif [ "${VTE_VERSION:-0}" -ge 3405 ]; then
            PROMPT_COMMAND="__vte_prompt_command"
        else
            PROMPT_COMMAND='printf "\033]0;%s@%s:%s\007" "${USER}" "${HOSTNAME%%.*}" "${PWD/#$HOME/\~}"'
        fi
        ;;
      screen*)
        if [ -e /etc/sysconfig/bash-prompt-screen ]; then
            PROMPT_COMMAND=/etc/sysconfig/bash-prompt-screen
        else
            PROMPT_COMMAND='printf "\033k%s@%s:%s\033\\" "${USER}" "${HOSTNAME%%.*}" "${PWD/#$HOME/\~}"'
        fi
        ;;
      *)
        [ -e /etc/sysconfig/bash-prompt-default ] && PROMPT_COMMAND=/etc/sysconfig/bash-prompt-default
        ;;
      esac
    fi
    # Turn on parallel history
    shopt -s histappend

[xPMo]

I use the fish shell, so I don't have a local bashrc file

Yeah, I use Zsh. This tells me that the error is in Arch's bash config. It has this construct in /etc/bash.bashrc:

case ${TERM} in
  xterm*|rxvt*|Eterm|aterm|kterm|gnome*)
    PROMPT_COMMAND=${PROMPT_COMMAND:+$PROMPT_COMMAND; }'printf "\033]0;%s@%s:%s\007" "${USER}" "${HOSTNAME%%.*}" "${PWD/#$HOME/\~}"'

    ;;
  screen*)
    PROMPT_COMMAND=${PROMPT_COMMAND:+$PROMPT_COMMAND; }'printf "\033_%s@%s:%s\033\\" "${USER}" "${HOSTNAME%%.*}" "${PWD/#$HOME/\~}"'
    ;;
esac

Somehow the expansion ${PROMPT_COMMAND:+$PROMPT_COMMAND; } isn't working as intended. It should only happen if PROMPT_COMMAND is already non-empty. Meanwhile toolbox has this:

    if ! [ -f /etc/profile.d/vte.sh ] && [ -z "$PROMPT_COMMAND" ] && [ "${VTE_VERSION:-0}" -ge 3405 ]; then
        case "$TERM" in
            xterm*|vte*)
                [ -n "${BASH_VERSION:-}" ] && PROMPT_COMMAND=" "
                ;;
        esac
    fi

To play nice with Arch, this should be PROMPT_COMMAND=':' instead of a space, OR shouldn't set it at all.

EDIT: The root cause is __vte_prompt_command , defined in /etc/profile.d/vte.sh (in pkg vte-profile) and used in /etc/bashrc (in setup). Conversation and fix here. This fix is recent, so the workaround here is still necessary for Fedora versions <= 33 afaik.

[xPMo]

https://bugs.archlinux.org/task/71002

[xPMo]

@erazemk: I closed #716 about the profile errors, Arch's image should add the vte-common package to support VTE terminals on hosts. I updated my arch-img branch to include it.

Additionally, /etc/machine-id being absent is no longer an issue. 🙂