Eli Zaretskii <eliz@gnu.org> writes:

> Unfortunately, I cannot match either of your two images with the
> corresponding dump of the glyph matrix: there are small
> inconsistencies in both the "OK" and the "not OK" cases.  For example,
> the first image shows "Top L1" in the mode line of the window that
> displays the article, but the matrix dump says:
>
>  68: (1) ' 1:--- . news:g.c.w.arnoldzwicky Austenian humbleb...  Top
> L19 (Article) '
>
> which is line 19.  This is a minor discrepancy, but since there are
> many similar discrepancies, some of which are probably parts of the
> bug, I feel I have no firm ground from which I can begin studying the
> data and making hypotheses about what went wrong.

I may have matched up the wrong bits of the log file or something.  I've
now redone it, but made sure that each log file bit really belongs to
the image.

> Could you please produce 2 sets of screenshot+matrix dump which are
> coherent with each other?  I must be sure that when I see a
> discrepancy, it is part of the bug, not some artifact that needs to be
> ignored.
>
> Also, I don't understand the second screenshot: which parts of it are
> correct and which are bogus?  The text before the image looks
> unfinished, but its continuation, starting with screen line 20, which
> is clearly present in the glyph matrix dump, is not shown.  Are you
> saying that Emacs was supposed to display that continuation, but never
> did, and instead left the old parts on display?

I've produced four screenshots instead, which should make it clearer
what it's supposed to look like, and what's the bug.

First, I've selected the article, and it's displaying the topmost part
of the article, and everything is fine.