To generalize, it seems the question in this issue is to implement a report texct under the mouse command. This is similar to setting the navigator object to the current object under the mouse pointer and speaking it, but obviously, not exactly the same.

may be we should redo script_moveNavigatorObjectToMouse to set the review position instead. @jcsteh: do you think that'd be possible?

Actually, I think this issue is related to "selected" text (i.e. holding down the mouse button and highlighting text), rather than just text under the mouse cursor; i.e. the current position of the mouse is irrelevant. We can't currently set nor retrieve the "real" selection in browsers.

@jcsteh, your #3773 (comment) suggests a cantfix. Is this right, or am I misinterpreting it.

It's not a can't fix, but it certainly requires a major refactor of our web support for most browsers:

1. For IE, this is not likely to happen. We need to continue to support IE for now, but we're not adding significant new features for it, as Microsoft is no longer developing it except for serious bug fixes. So, this would be a won't fix for IE.
2. For Edge, it should theoretically be possible to use the real browser selection instead of our virtual selection, though we don't do this already. However, there may be bugs/deficiencies in Edge which present obstacles. @michaelDCurran vaguely recalls that this didn't work at some point in the past, but this would need to be re-evaluated.
3. For Firefox/Chrome, we're hoping to eventually refactor browse mode to allow for things like this. However, this is a major piece of work and we don't have an ETA at this stage.

considering the last PRs #9048 and #9064, Maybe this issue is almost fixed.

I maintain a web application which is based on selecting text, however I found that it is impossible to use it with NVDA. My problem is very similar to this, but the other way: I would like NVDA's selection to be visible to the app as a normal text selection, so that the app can record it (from JavaScript). There are probably a lot of web apps which need this, to apply formatting, Medium's "featured highlight', etc.

I had assumed #1464 was related to that problem, but it was closed in favor of this issue which doesn't mention this at all. I'm happy to open a separate issue if that is the right way to proceed.

This is definitely still an issue which comes up quite regularly for users. Often even moreso simply copying the text rather than reading the selection - Generally when the text is selected with the mouse, it is done by users with enough sight to do that (who can then also see the visual selection they have made) so the main thing they want to do at that point is copy it to go to another window and paste it.

The workaround is to open the context menu and choose "copy" (or NVDA+f2 for passthrough then control+c to copy), but having NVDA able to recognise the text selected by mouse (and thus copy it), would be very beneficial for those users.

I would love it if NVDA could read text I select using my mouse... without work around... I currently use additional chrome add ons to do this... maybe this could be an additional service NVDA could offer?

Now that speech on demand is part of NVDA's speech modes, I think the remaining thing to be added here is to include the mouse cursor selection into the nvda+shift+up or nvda+shift+s command in desktop and laptop keyboard layout.
cc: @CyrilleB79

@Adriani90, I do not get the link between this PR and speech on-demand. Can you clarify.

I wonder if the current issue can be solved by #15830 or at least if #15830 can help its future resolution in some way.

Mostly, sighted people don't want to hear the NVDA reportings all the time, but only in specific cases, such as in this case where selected text with the mouse should be reported. So speech on demand will play a role in this particular case but yeah it was of course not a must have for this issue to be solved. I just mentioned it.