data: URL parsing and the event loop

[Henri Sivonen]

I've identified multiple mochitests that depend on how exactly the old
HTML parser in Gecko, the stream data pumping code and the docshell
work. Specifically, the test cases assume that scriptless data: URLs
parse synchronously.

With the HTML5 parser, no URL loads parse synchronously. (This is
spec-compliant; the spec allows the parser to spin the event loop pretty
much anywhere.)

To make new mochitests work with the HTML5 parser, please use the onload
handler to continue the test after data: URLs have been loaded. That is,
please treat data: URLs like http: URLs as far as the event loop goes.

(I'll fix the existing mochitests that depend on data: URLs parsing
synchronously.)

--
Henri Sivonen
hsiv...@iki.fi
http://hsivonen.iki.fi/