Last modified: 2013-04-22 16:14:25 UTC
Maybe related to bug 41217. Try to round-trip <span title="6">a</span>
Verified, still the case. This is a bit lower priority for us as it 'only' produces a syntactic diff- the content does not really change. We'll see how far we get with selective serialization, and might add extra support for this if that is not sufficient. Do you see any situations where this could be an issue beyond producing a dirty diff?
(In reply to comment #1) > Do you see any situations where this could be an issue beyond producing a dirty > diff? Yes. <span title="&amp;">a</span> round trips to <span title="&">a</span> which is a different attribute.
Argh- good catch!
May as well take this one.
Patch pushed for review in https://gerrit.wikimedia.org/r/#/c/43859/. In attributes, this patch does not preserve unnecessary entities like a bare '&' escaped to &. It can thus still cause syntactic diffs, but will avoid semantic diffs. The patch also adds escaping for newly created entity-like text content.
And merged.