bug#1458: 23.0.60; mouse-posn-property: Args out of range

[Stefan Monnier <monnier@IRO.UMontreal.CA>]

>> You patch to timer_check doesn't seem right: it may cause the
>> current-buffer to be changed even if the timer doesn't do anything, just
>> because the current-buffer before running the timer wasn't the
>> selected-window's buffer.  It causes no end of trouble in Gnus
>> (e.g. when connecting to an IMAP server, it calls read-passwd during
>> which calls to read-event can run timers, so we start read-passwd in
>> the IMAP process's buffer and we finish it in the *Group* buffer or
>> which ever buffer is currently displayed, after which
>> send-process-string fails because it can't find the process).

> I see.  The change is reverted.

> Now I'm not certain how to deal with the problem described in bug#1458.
> Perhaps we should make the selected window's buffer current only if the
> selected window's buffer was current before running the timer.  WDYT?

I don't think so.  I think we need to track this bug further down.
Maybe the problem is that some code uses current_buffer (maybe to build
the mouse events?) rather than selected_window->buffer.


        Stefan