On Tue, Oct 8, 2019 at 10:50 AM Eli Zaretskii <eliz@gnu.org> wrote:

> Sorry, I don't think I understand what the images show me.  They seem
> identical.  Which face bleeds and where?  Please point out what should
> I be looking at to understand the difference.

In both images, the line numbered 19 is the last line in the buffer.

In the it_eob image, the "line" past the end of the buffer has an empty
number,
drawn with the major-tick face (but obviously without number). The rest of
non-lines,
until the end of the buffer, have their empty number drawn in the
line-number face.

In the beyond_zv image, the empty line after the end of the buffer is in the
line-number face, as all the other past that point. I think that's the
right behavior.

There's no reason for the line after the end of the buffer to be drawn with
the
major-tick face, even if it would be a major-tick line if it existed. It's
ugly.

> Did you try to arrange for the last line to be a multiple of one of
> the ticks as well?

In my examples? Yes, that's the whole point of the test: knowing what
happens when
the line after EOB would match a tick line number.

>  Also, what happens if you use the beyond_zv test
> in all the conditions

That's what I've done in the second patch I sent (applied after the first
one, not
alone). In my simple tests, everything works as expected.

> or use the it->what test in all the conditions?

I didn't try that, but as the first check (that uses it->what) is trying to
decide
whether to draw with the current-line-number, I don't think it is relevant
to
the problem I was trying to fix. It is relevant for consistency, of course.

> IOW, I don't understand why we need two different conditions regarding
> EOB for displaying a number with different faces.  What am I missing?

I don't know why (or if) the it->what check is necessary, instead of
beyond_zv.
I *know* that the other conditions (the ones affecting the choosing of tick
faces)
need beyond_zv, at least to get what I think is the right behavior (the one
in the
beyond_zv image). As said, I think that in fact it the first check can be
safely
replaced by !beyond_zv.

> You could simply start with
>
>   tem_it.face_id = lnum_face_id;
>
> and then have a series of tests for replacing that with another face
> ID; it would at least save you the 'else' clause.

Yes, good idea. Thanks.

> What does "between digits" mean?  While producing the glyphs for the
> digits of a single line number, no faces can change, because no Lisp
> is invoked.  Is that what you meant?

Yes, thanks. I expected as much, but the way the original code was written,
and
redisplay being notoriously tricky and an arcane art best left to wizards, I
though I was missing something (I'm not joking).