rlwrap doesn't work in vterm

[HyunggyuJang]

Ref: https://www.reddit.com/r/emacs/comments/tp7el9/getting_rlwrap_to_work_in_vterm/:

I am using vterm as a terminal emulator in emacs. While running sml/nj repl in vterm rlwrap doesn't seem to work. (It works in other terminal emulators that I have: yakuake, kitty).

Any suggestions?

I'm experiencing the same issue.

[Sbozzolo]

This is a little bit too little information to help. It would be useful to have "steps to reproduce" and define what "doesn't seem to work" exactly means.

[HyunggyuJang]

This is a little bit too little information to help. It would be useful to have "steps to reproduce" and define what "doesn't seem to work" exactly means.

Thanks for your elaboration.

Minimum steps to reproduce

1. Launch vterm in emacs
2. Type rlwrap read
3. Type Ctrl-e
   • Expected: move cursor to the end of the line
   • Actual: character '^E' appears

From terminal emulator like kitty, it behaves as expected.

[dominicfreeston]

Can confirm I've faced the same issue. e.g. using babashka on MacOS

rlwrap bb repl

Pressing the up arrow inputs <[[A instead of cycling through the history.

[Uthar]

Expected: move cursor to the end of the line Actual: character '^E' appears

I have exactly the same problem. Does anyone know what could be causing this?

[papadakis-k]

Minimum steps to reproduce

1. Install/compile rlwrap 0.45 or later.
2. Install ed editor
3. Launch vterm in emacs
4. Type rlwrap ed
5. Type Ctrl-e

Expected: move cursor to the end of the line Actual: character '^E' appears

Investigation

There is this commit here, in rlwrap's upstream, that explicitly disabled rlwrap working inside Emacs, not sure what was the reasoning behind though.

In the NEWS file you can now read:

 when run inside an emacs shell buffer, rlwrap will execute
 the client in its place (just as when stdin is not a terminal)  

This commit was added on 0.45 version, meaning rlwrap 0.44 or earlier might work inside Emacs.

It seems that you can skip the EMACS check by unsetting the environment variable _INSIDEEMACS, and after some very basic testing I have done, this seems to work.

Update

After further testing I realize that when using the shell mode (M-x shell) it all starts to make sense, the line editing capabilities of the shell mode would get in the way of rlwrap and vice versa, and that's probably the reasoning behind this specific change in rlwrap. If we unset _INSIDEEMACS in a shell mode and then run rlwrap ed , we get a warning rlwrap: warning: Your terminal 'dumb' is not fully functional, expect some problems and the terminal is scrambled. eshell mode also acts similarly, but not that bad.
Fortunatelly, both shell and eshell mode set terminal to "dumb", so maybe the check in rlwrap should be enchanted in a way that it checks INSIDE_EMACS variable but also check that the TERM variable is set to "dumb", otherwise act as normal.

Update 2

I have created an issue on the upstream, also created a PR.