bug#81561: 31.0.91; speedbar-window-mode creates stale idle timers

Hello people,

Vincenzo Pupillo <[email protected]> writes:

> Ciao, I wasn’t able to reproduce the problem reported by Daniel. The only way 
> was to kill the 'speedbar-buffer' process. So I added a 'kill-buffer-hook' to 
> handle this event. I also eliminated a few other causes of  stale idle timers.
>
> @Daniel, could you please try this patch?

Perhaps I'm hallucinating, but I seem to recall stale dframe timers
hanging around long before `speedbar-window-mode' came about and it
having something to do with `dframe-timer' becoming buffer-local back in
Emacs 29. If that's correct, shouldn't any functions that rely on that
variable, like `dframe-set-timer' and `dframe-set-timer-internal', be
made to run in whatever BUFFER-VAR is given to `dframe-frame-mode' on
init? If so, instead of wrapping all these `speedbar-timer-fn' call
sites in `with-current-buffer', what about doing so around
`dframe-set-timer' in `speedbar-set-timer'?

Thanks,
J.P.

P.S. If it's determined that this bug is not in fact directly related to
the introduction of `speedbar-window-mode', perhaps it's better to
revert a56e964941f "Speedbar: no more stale idle timers (bug#81561)" on
Emacs 31 and instead apply any future attempted fixes to master?