bug#12600: 24.2.50; linum-mode: line numbers in fringe do not refresh when resizing frame

[martin rudalics]

> Quite a few struct members related to the
> display engine (and I suppose other core structures as well) are
> insufficiently, in accurately, or even downright incorrectly
> documented.  I try to fix every such problem I bump into, FWIW.

If these concern the display engine exclusively, there's no problem.
Problems arise when the window handling code is supposed to change them.

>> For example, what does the "displayed buffer's text modification events
>> counter as of last time display completed" mean?
>
> This one sounds quite clear, please tell which part needs further
> clarifications.  It would be better to say "counter of the displayed
> buffer's modification events as of last time display completed", but
> somehow I'm guessing it's not what you had in mind.

That's exactly what I had in mind: Either something that corresponds
textually to what is used in buffer.h or just mention the name of the
corresponding field used in buffer.h.  As it stands currently, I have to
either guess what is meant or go through the comparisons in the code to
find out.

>> Suppose redisplay has set this to 37.  Apparently, setting it to 36
>> means that the next redisplay will notice that for this window
>> display is not accurate and has to be redone because 36 < 37.
>
> Not necessarily.  It depends on what is recorded in the buffer's
> BUF_MODIFF slot.

IIUC this can only increase monotonously (modulo wraparounds, but in
that case ...).

>> But why is a flag insufficient to accomplish that?  IIUC it's set only
>> once by mark_window_display_accurate_1 and everywhere else reset to 0.
>> Why can't we set it to "accurate" in mark_window_display_accurate_1 and
>> "inaccurate" everwhere else?
>
> How would this work when the displayed buffer is modified behind
> redisplay's back?  The last_modified slot belongs to the window, which
> redisplay controls, but it doesn't control the buffer, which could
> well be displayed in other windows as well and modified through there.
> The display engine treats each window separately; when it displays
> some window, it doesn't use information from other windows.

When last_modifed_flag is set, the window must be redisplayed.  Now if
someone else (including the display engine) decremented last_modified,
things were different indeed.  But as it stands I don't see any such
assignment.  OTOH when the buffer iself is modified it has to be
redisplayed anyway because we hardly know where the change happened.  So
I don't see why such a comparison of counters is useful.  But maybe I'm
missing something.

>> And why do we have additional fields for last_overlay_modified and
>> window_end_valid?  What sense does it make to handle these separately?
>
> Because changes in overlays do not constitute changes in buffer
> contents, and are done via yet another set of primitives, yet they do
> require redisplay.  Also because each one of these can allow and
> disallow certain distinct redisplay optimizations, at least in
> principle.

I doubt whether such optimizations exist.  Suppose we could exclude a
change to happen within the bounds of a window as for example a buffer
displayed with a window starting at M and ending at N and an overlay
modified at some position < M or > N.  Wouldn't this require to
associate each single overlay with a modification counter?

> As for window_end_valid flag, it's there to allow yet another set of
> redisplay optimizations.

I think that a correct implementation would have for each window w

(w->last_modified < MODIFF) => (NOT (w->window_end_valid))

but strongly doubt that this implication holds currently.

>> For example, wherever last_modified is reset to 0 last_overlay_modified
>> is always reset to 0 too.  Isn't that plain overkill?
>
> The potential for separate optimizations is not realized, that's true.
> But it exists; I'm not sure we should remove it, and I surely won't
> spend any of my personal time trying.  Fixing real bugs in the display
> engine is enough to fill my free time already.  Of course, if you feel
> like it, and if no one objects, feel free to make these cleanups.

I don't intend to make any cleanups.  I'd like to have a simple routine
we can use to reset any such structur members within window.c and have
that DTRT.  Currently, `window-end' is clearly broken and we should fix
it.

martin