Can I conditionally exit the parent fzf process from an execute command?

[bschlenk]

I've been using fzf to build a git branch script. One feature is that while viewing a list of branches in fzf, you can press the right arrow to jump into a commit view for that branch. This works by using --bind right:execute(...) and calling my script again with different args. By using execute, control returns to the branch view when the commit view exits.

Now I want to add additional things to the commit view, like binding ctrl-c to copy the sha. In this case I'd like to exit the script completely, bypassing the original fzf process. I can't figure out a way to do this from fzf directly. It seems to ignore exit codes from the execute process. I tried a hacky idea of passing the original process' pid to the child and then sending a SIGINT from the child, but that causes fzf not to clean things up and the terminal is busted.

Is there a way to do this without completely restructuring my script? I can switch to using --expect and starting the sub fzf window myself, but I like that execute uses a separate window and doesn't otherwise affect my terminal layout.

[junegunn]

You can start fzf with --listen and send abort action to the server listening on $FZF_PORT via POST request.

[bschlenk]

Thanks! I got something working using --listen but it feels pretty hacky. I'm sending become(echo "{message}"; exit 7) and then handling an exit code of 7 specially from the python script. The other issue is that anything I print from the second fzf window is lost when that window closes, so I need to send the message back up to the caller to print it out there instead.