Or, well, one could argue that it was already broken but my content sink changes and a suitably crafted test case just exposed the layout issues that were already there.
After I saw the breakage, I added “IsMonolithicContainer”
checks to the XML content sink and made the mtr element
monolithic. It didn’t help, so I made math monolithic.
Now the results depend on interaction with the event loop.
There’s a test case, but it is useless without my content sink changes.
Actual results (varying pointing device waving throughout parsing):
Expected results (also actual results from the same build but with vigorous pointing device waving):
I constructed an XHTML
table test case that has the same table structure as the MathML
test case. It is laid out properly regardless of how I wave the
pointing device. (I made XHTML tr monolithic as well.)
Will the Reflow Branch fix this?
There’s a thread for comments in mozilla.dev.tech.mathml.