The reason for dirty diffs in infoboxes and other templates being seen currently on en-wp VE edits is because of incorrect DSR on reused template expansions.
Ex:
- dsr":[2076,2080,2,2] for the infobox on en:Libreoffice
- dsr":[878,882,2,2] for the infobox on en:Bleak House
This then marks the infobox dirty during serialization (although not sure why it should if the same cached html (with buggy dsr) is used by the serializer -- so, maybe something is off there as well) which causes it to be serialized via the regular serializer -- which then normalizes whitespace, etc.
NOTE: Ignoring data-parsoid attribute during dom-diff would cause more problems since the incorrect DSR values would be used to return original wikitext which would have broken serialize WT more seriously.
Related: we need a way of testing this functionality as well.
Version: unspecified
Severity: major
See Also:
https://bugzilla.wikimedia.org/show_bug.cgi?id=49608