bad display with image and U+2060 WORD JOINER

[vinc17fr]

I've just reported Debian bug 1100487.

There is a display issue with w3m 0.5.3+git20230121-2.1 when one has an image (valid or not) and text with the U+2060 WORD JOINER character.

Here is a testcase (from my Debian bug report), which contains something like:

<p>Text.</p>
[...]
<p>Text.</p>
<p><img src="data:image/png;base64," alt=""/></p>
<p>foo&#x2060;bar abcdef.</p>

Search for ab, then move the cursor to the right up to the end.

One successively gets:

foo⁠bar aacdef.
foo⁠bar aabdef.
foo⁠bar aabcef.
foo⁠bar aabcdf.
foo⁠bar aabcde.
foo⁠bar aabcdef
foo⁠bar aabcdef.

Alternatively, put the cursor at the end ([End] key), then at the beginning of the foo⁠bar abcdef. line, then move the cursor to the right: same display issue starting with bar.

It is important that the test part is not displayed on the first page, hence the sequence of paragraphs. Note that this issue disappears if either the image or the U+2060 WORD JOINER character is removed.

I first noticed this issue on https://bugzilla.mozilla.org/show_bug.cgi?id=1876874 when I was searching for "C\+\+" (specifically "of C\+\+").

[rkta]

Can reproduce using ISEARCH. I also created a ticket for my forked version, see [0]. [0]: https://todo.sr.ht/~rkta/w3m/75