It would really help if we knew what text unit resolution was used when this was reported. Anyway, this is probably quite hard to reproduce for blind users

cc @feerrenrut @Qchristensen

There are a couple of issues here, and it also varies between browsers. To test, I threw together a test page and just uploaded it on another domain I had easy access to: http://22point.com.au/random/test.html

I wrote three lines of text, one plain text, then a paragraph break, then another long line with a long link in the middle, then a BR linebreak, and finally another long line.

Exact results depend on your zoom, so I set my zoom such that the two plain text lines fill two lines each, and the link takes up one full line and a part line either side.

In Firefox, with text unit resolution left at the default "paragraph" OR changed to "line", the results are the same:

• If hovering over plain text on a line, the plain text on that line is read.
• Hovering over any part of the link, NVDA reads the part of the link on the first line.
• Hovering over the plain text at the end of the link, NVDA reads all the plain text from that paragraph (but not the link).

With Text unit resolution set to Word, hovering over any plain text word, NVDA reads that word (as expected). However, hovering over any part of the link on either line, NVDA reads "This" - the first word in the link. Set to "character" is essentially the same behaviour - NVDA reads that current character on any plain text, but "T" on any part of the link.

In Chrome, with text unit resolution at Paragraph:

• Initially in an earlier, simpler version of the text document, I found that hovering over any plain text reads that whole block of plain text, regardless of where in the paragraph the mouse is hovering. The same is observed for links (ie, it reads the whole link).

WIth text unit at Line:

• Hovering over any plain text reads the first line (or part line) of that block of plain text, regardless where in the paragraph the mouse is. The same behaviour is observed for links (ie, it reads the first line or part line of the link).

With text unit at word or character: the first word (or character) of the current block of text (plain or link) is read.

Once I updated the document, to the current version (with the paragraph and line break) I can't get mouse tracking to read reliably at all? It will often read the first or last lines, but then just reads the document name and goes silent.

Microsoft Edge seems to ignore text unit resolution and just reads the whole paragraph of text of the current type regardless.

Internet explorer also seems to ignore text unit resolution:

• Hovering over the first line, the start of the second line, or the link, reads from the start of that block of text to the very end of the document.
• Hovering over any of the plain text after the link (on that line or the next), reads from the start of that paragraph (I separated this second line...), to the end of the document, including the link.

So a bit inconsistent and not ideal.

This inconsistency would be good to fix, but I think this would be a fairly time consuming bug to fix. I am going to label this as P3.

@Qchristensen, @feerrenrut I guess this is still reproducile? I am fully blind so unfortunately I cannot test it but with #9064, #8572 and other recent PRs I think the priority for this issue could be higher now.

Yes this is still reproducible with the latest alpha and also NVDA 2018.4rc1

This is still reproducible in 2019.2rc1 and being reported by users.

I've had another user report this against 2020.4 still, who also added that the same happens with formatted text such as bold or italic text.

This is still an issue in 2023.2. It seems to occur in Chromium-based browsers and apps. Chrome and Edge exhibit this behavior, as does an app like Discord. Firefox does not have this issue.
This seems to occur whenever another tag is met in rich text, such as non-editable text in an HTML document. Take this example:

This is a paragraph, and this is a span element inside of the paragraph. The spand text will break the mouse tracking reading, regardless of text resolution setting.

<p>This is a paragraph, and <span>this is a span element inside of the paragraph.</span> The spand text will break the mouse tracking reading, regardless of text resolution setting.</p>

Narrator and JAWS do not have this issue on Chromium-based applications and browsers.