We have multiple html2wt regressions in the "Italics and bold: multiple quote sequences" tests, which were hidden because these tests were on the blacklist (the <nowiki> insertion previously made the html2wt tests fail).
The failing tests, along with their wt2wt output (as "After") are:
Italics and bold: multiple quote sequences: (3,4,2)
Before: '''foo''''bar''
<b>foo'</b>bar<i></i>
After: '''<nowiki>foo'</nowiki>'''bar''''
<b>foo'</b>bar'<b></b>
Italics and bold: multiple quote sequences: (3,4,3)
Before: '''foo''''bar'''
<b>foo'</b>bar<b></b>
After: '''<nowiki>foo'</nowiki>'''bar''''''
'<i>foo<b>bar'</b></i>
In the process of fixing these bugs, we should probably also write "clean" versions of the tests which include the proper <nowiki> tags in the input and therefore round-trip correctly, to ensure that we are better protected against regressions in the future.
These next tests are ok, but should have a clean version added. The "Before" version is what's in the test currently; after one round of --wt2wt it becomes the equivalent "Clean" version, which subsequently round-trips correctly:
Italics and bold: multiple quote sequences: (2,4,2)
Before: ''foo''''bar''
Clean: ''<nowiki>foo'</nowiki>'''bar'''''
Italics and bold: multiple quote sequences: (2,4,3)
Before: ''foo''''bar'''
Clean: ''<nowiki>foo'</nowiki>'''bar'''''
Italics and bold: multiple quote sequences: (2,4,4)
Before: ''foo''''bar''''
Clean: ''<nowiki>foo'</nowiki>'''<nowiki>bar'</nowiki>'''''
Version: unspecified
Severity: normal