Transient menu annoyingly introduces extra window when Helm mode enabled

[daedsidog]

• I have updated gptel to the latest commit and tested that the issue still exists (I have this bug with the latest commit, but I recreate the bug with the latest release)

Bug Description

Exiting out of the gptel-menu transient menu after invoking something that relies on completing read that was rerouted to Helm mode creates a buffer that stays after the transient menu has been closed.

Backend

None

Steps to Reproduce

1. Install helm-mode and gptel
2. Enable Helm globally with (helm-mode 1)
3. M-x gptel-menu -i C-g C-g to introduce the aforementioned buffer (sometimes, just a single C-g is enough to exit out of gptel-menu, but other times you need two: one to exit the -i menu and another to exit gptel-menu

M-x gptel-menu:

-i:

C-g (or C-g C-g):

Additional Context

Emacs version: GNU Emacs 29.4 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.24.43, cairo version 1.18.2)
OS: WSL2 Arch Linux

Helm versions:

helm-20250219.1650
helm-ag-20241028.144
helm-core-20250217.1544 
helm-project-20230608.1212

Transient version:

transient-20250216.1730

Bug can be recreated using both the latest gptel master and v0.9.7.

[daedsidog]

Also occurs with model selection. I truly don't remember this happening before.

This could be a Transient-specific bug, but Magit works without issue for me.

[karthink]

What is your value of transient-display-buffer-action?

…

On Wed, Feb 19, 2025, 11:49 PM daedsidog ***@***.***> wrote: Also occurs with model selection. I truly don't remember this happening before. — Reply to this email directly, view it on GitHub <#656 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACBVOLGJJYET56J4PHGWSPL2QWCI5AVCNFSM6AAAAABXQBNIF2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNZQG4YDSNJSGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> [image: daedsidog]*daedsidog* left a comment (karthink/gptel#656) <#656 (comment)> Also occurs with model selection. I truly don't remember this happening before. — Reply to this email directly, view it on GitHub <#656 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACBVOLGJJYET56J4PHGWSPL2QWCI5AVCNFSM6AAAAABXQBNIF2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNZQG4YDSNJSGA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[daedsidog]

What is your value of transient-display-buffer-action?

(display-buffer-in-side-window (side . bottom) (dedicated . t) (inhibit-same-window . t))

[karthink]

> What is your value of transient-display-buffer-action? `(display-buffer-in-side-window (side . bottom) (dedicated . t) (inhibit-same-window . t))`

That looks fine. My guess is that Helm doesn't respect display-buffer settings then? The best way to test if this is a gptel issue would be to use a magit command that invokes (Helm's) completing-read and see if the problem occurs.

[daedsidog]

The best way to test if this is a gptel issue would be to use a magit command that invokes (Helm's) completing-read and see if the problem occurs.

I've tried lots of Magit commands and couldn't recreate it (e.g. C-x g b b, branch checkout, which I assume uses completing-read).

[karthink]

Can you try (and suggest) a command that isn't a suffix, i.e. it doesn't end the transient session?

[karthink]

Also please try this transient:

(transient-define-prefix test-pre ()
  "TODO"
  ["Test"
   ("r" "read" "r"
    :class transient-option
    :prompt "Pick: "
    :reader (lambda (p i h)
	      (completing-read p '("one" "two" "three")
			       nil nil i h)))])

Use the r option to pick a value, then quit the transient. Does the bug occur?

This is almost definitely a Helm + Transient issue, not gptel. You can track the status of this report at magit/transient#358

[daedsidog]

Use the r option to pick a value, then quit the transient. Does the bug occur?

Yes

This is almost definitely a Helm + Transient issue, not gptel. You can track the status of this report at magit/transient#358

I suspected as much. Closing as it is not gptel-related.