Change 254613 had a related patch set uploaded (by DLynch):
Add ve.ce.Surface selectActiveLinkContents

https://gerrit.wikimedia.org/r/254613

Change 254615 had a related patch set uploaded (by DLynch):
When adding a label to a numbered link, force the selection

https://gerrit.wikimedia.org/r/254615

This is related to T114376. I was thinking of a more general fix that allowed us to specify an inner or outer bias when setting a model selection (outer bias meaning go to the 'outside' of a tag that lives within your dm selection offset).

I was initially thinking that might be a bit ambiguous, just because the annotations can pile and we could have <b><i><a>foo. But if we don't mind ignoring the "I want to put the cursor between the <b> and the <i> case, it'd be relatively consistent to just have, essentially a "at the start of offset X" and "at the end of offset X" hint.

Yeah I think inner/outer is crude but enough for the bugs we have now.

When you set a cursor position in Chromium, it will silently get normalized to the "start-most" equivalent position with respect to annotation tags and text nodes.

For example if you tell it to select '<p><b>|<text#>foo</#text>|</b></p>' (i.e. just the content of the bold element) then it will actually select '<p>|<b><text#>foo|</#text></b></p>' (i.e. starting just before the bold element, and ending at the last offset of the text node).

So a function purporting to bias a general model selection would at times fail silently. On the other hand, specific code to select the link content will work on all browsers, because of link nails. Therefore I'd prefer the latter.

Change 254613 merged by jenkins-bot:
Add ve.ce.Surface selectActiveLinkContents

https://gerrit.wikimedia.org/r/254613

What about "nail bias" for links and anything that uses nails in the future. That would fix T114376 too.

Change 254615 merged by jenkins-bot:
When adding a label to a numbered link, force the selection

https://gerrit.wikimedia.org/r/254615