Wikitext:
<blockquote> Foo Bar </blockquote>
Parsoid renders this as:
<blockquote data-parsoid='{"stx":"html","dsr":[0,35,12,13]}'>
<p data-parsoid='{"dsr":[13,16,0,0]}'>Foo</p>
<p data-parsoid='{"dsr":[18,21,0,0]}'>Bar</p>
</blockquote>PHP renders this as:
<blockquote> Foo Bar </blockquote>
which displays as "Foo Bar"
Version: unspecified
Severity: normal
See Also:
T8200
T3857
T17491
Ugh ... Tidy messes with us again.
<blockquote>
foo
</blockquote>
In the test case above and the the one posted by Roan, parserTests specifies no outer p-wrapping, but Tidy adds an outermost p-wrapper inside blockquote (check in sandbox). Only one p-tag is added around plain text insde blockquote independent of # of newlines. Adding it will fail parser tests, and not adding it will not match Tidy's output.
So, should we or should we not add that single p-wrapper around text as a child of <blockquote>?
The content semantics are really those with tidy, so whenever those make some sense we should follow them rather than broken PHP-only parser behavior. We can always mark the test Parsoid-only.
Discussing this in IRC, we resolved that this was Not A Bug in Parsoid... but in fact the PHP parser should be changed to be consistent with parsoid (ie, double-newlines inside blockquotes should indicate division into paragraphs).