Re: Suggestion: A fringe indicator that shows the last/first line before scrolling

[Stefan Monnier <monnier@iro.umontreal.ca>]

>> It's neither.  It's just the result of how the code is written: there's
>> only one command (the mouse-drag) which happens to read several
>> events  waiting for the event that marks the end of the drag.

> What events actually terminate the drag ?  Martin said it was in

That is yet-another-problem due to the way the code is written.
Actually this part of the mouse-drag is really ugly.  When the drag
ends, the command doesn't end: it waits for the next event (i.e. the
next command).  This is because it want to know if the next event is one
of mouse-region-delete-keys.

I hope we can get rid of this madness at some point.  E.g. by (re)using
the delete-selection-code for this feature.

>> There are other cases where scrolling can happen without running
>> post-command-hook, e.g. scrolling triggered by process filters
>> (e.g. scrolling in the *compilation* buffer or in tail-mode buffers).

> Fair enough - means post-command-hook is a nonstarter like it was for
> point movements.

Not really.  It just means it will only work 99%.
For a more reliable result you'll probably want/need to hack the C code
to introduce some kind of pre-redisplay-hook, since what you're looking
for is linked to the display changes rather than to execution of commands.


        Stefan