fix(bash): fix preexec of child Bash session started by enter_accept

[akinomyoga]

I noticed that if a new child Bash session is started by enter_accept, the preexec hook provided by bash-preexec is not triggered at all. This is because the child Bash session inherits the environment variable READLINE_POINT of the parent shell session, and bash-preexec considers we are still processing the Atuin menu even when a new user command is processed inside the child Bash session.

How to reproduce the issue

Consider the following bashrc:

# bashrc
source ~/.mwg/git/rcaloras/bash-preexec/bash-preexec.sh
preexec() { echo "[$FUNCNAME]"; }
precmd() { echo "[$FUNCNAME]"; }
eval "$(atuin init bash)"

With the current main branch, we first start a new child session by manually typing the command bash and running it:

$ bash        # <-- This starts a new child Bash session
[preexec]    # <-- preexec by parent session
[precmd]
$ echo 1
[preexec]
1
[precmd]
$ exit
[preexec]
exit
[precmd]     # <-- precmd by parent sesssion
$

This registers the history entry "bash" in Atuin's database. Any problems don't happen so far. Then, we next press the keys upupupenter to select the entry "bash" and execute it through the enter_accept feature.

$ bash        # <-- This is executed through Atuin's enter_accept
[preexec]     # <-- preexec by parent session
[precmd]
$ echo 1
1      # <-- preexec is inactive
[precmd]
$ echo 2
2      # <-- preexec is inactive
[precmd]
$ exit
exit      # <-- preexec is inactive
[precmd]     # <-- precmd by parent session
$

Solution

In this PR, the problem is solved by removing the export attribute of READLINE_POINT and READLINE_LINE before running the user command with enter_accept. See the code comments added in this PR for details.

Checks

• I am happy for maintainers to push small adjustments to this PR, to speed up the review cycle
• I have checked that there are no existing pull requests for the same thing

[ellie]

nice one - thank you for the detailed explanation too!

[akinomyoga]

Thanks!