No resize-on-refresh

[kryzar]

Hi Leo,

First of all thanks a lot for clifm, it's truly a beautiful program, and it's fast. I'm having an issue with refresh-on-resize.

Describe the bug

I'm running Hyprland (so, Wayland). Basically, refresh-on-resize doesn't seem to work for me. If I open clifm on a directory with a lot of files, and on a large screen, resizing the window to a smaller size doesn't seem to refresh the interface, making it rather hard to read.

To Reproduce

Launch clifm in a big window, and then resize the window to a smaller size.
This is when I launch clifm on a full-width window:

And this is after opening a new window on the right of the screen, making the original window smaller. This should have triggered a resize, but didn't:

My understanding from the doc is that refresh-on-resize should be automatic.

Expected behavior

Interface should refresh automatically when the size of the window is modified, and if clifm is opened in the foreground.

Screenshots

See above.

Desktop (please complete the following information):

• OS: Arch Linux x86_64, Linux 6.16.7-arch1-1
• Terminal: kitty 0.42.2
• Clifm version: 1.26
• Installation source: AUR

Additional context

Note that refresh with rf or Ctrl-L works otherwise very well. Also, since you mention i3 in this issue, I wonder is this might be a Wayland issue, as I've tried launching clifm from different terms, as well as different Wayland compositors. By default, I use Hyprland, launched through uwsm, but I've also had the exact same issue Sway-through-uwsm and Hyprland directly from the Arch Linux tty.

[leo-arch]

Hi @kryzar, and thanks.

TL;DR: We can, but we shouldn't.

The docs refer to clifm's behavior after closing a window spawned by a command launched by clifm itself.

When you spawn a new Hyprland window using a keybinding (usually Super+Return), clifm only gets a SIGWINCH signal. Though we can respond to this signal automatically (via a SIGWINCH handler) and update the screen, we should not do it, because refreshing the list of files involves calling several non-reentrant functions, which should never be used this way (i.e. on SIGWINCH, or more generally, by a signal handler itself).

However, your request asks for a desirable feature. I'll keep this in mind.

Also, calling a feature refresh-on-resize, as we do, is misleading, as it seems to imply exactly what you're asking for. Agreed. Maybe it would be clearer to simply remove those lines from the docs.