Platform/HTML5 parser tasks

This page outlines what needs to be done to 1) enable the HTML5 parser by default on mozilla-central and 2) ship a release with the HTML5 parser enabled by default.

Things that need to be done before turning the HTML5 parser on by default

• Frequent timing-sensitive test failures on the real tinderbox
  • 562328 test_frames.html depends on SimpleTest.waitForFocus(), which makes the test timing-sensitive (I suggest disabling this test -- hsivonen)
• Less frequent unmasked random oranges on the real tinderbox
  • 562326 bug451286_window.xul (used by test_bug451286_window.xul) fails randomly (I suggest leaving this on to see how often it happens and disabling the test if it happens often -- hsivonen)
• Waiting for review:
  • 562319 test_sibling-matching-parent.html and test_bug270414.html fail intermittently (bzbarsky)
  • 562635 make the first flush timer delay equally long to the subsequent delays (sicking)
  • 562631 layout/generic/test/test_bug496275.html makes mochitest-plain time out when the HTML5 parser is enabled (mats) (introduced after the HTML5 parser was briefly turned on)
• Assumed to be fixed:
  • 562333 jsreftest.html randomly fails with "timed out waiting for onload to fire"

Things that need to be done before shipping the HTML5 parser in a final Firefox release

• Fix "[HTML5]" bugs marked P2. (hsivonen)
  • 482913 Make sure document.write() order for nested scripts is right
    • Waiting for Hixie to respond to feedback about a test case looking broken.
    • Time estimate: Hixie's test cases for this can be run in less than a day. If something fails, working out what failed, why it failed and if we want to change code, tests or the spec will take multiple email round-trips through the WHATWG list spanning at least a week.
  • 483209 The HTML5 parser needs limits on internal buffer growth
    • Time estimate: Less than a day.
  • 552908 Implement spec changes to SVG </a> and handling
    • Waiting for Hixie to resolve bugs discovered in the spec text while implementing this.
  • 552938 Implement the SVG load event in text/html
    • Time estimate: Less than a day, but waiting for the SVG WG to decide whether to remove this event instead.
  • 557197 &#13; in element content expands to CR and shows up as a non-collapsing space in layout
    • Time estimate: Less than a day if hacked around in the parser. Unknown if fixed more properly in layout.
• Security review.
  • Talked to dveditz about how to do the security review for the HTML5 parser, and it's not clear what the best format for the review would be. But we did decide that it's not something we need to do before turning the parser on by default, but we need to do it before we ship it. --jst
• 544994Create a "slow tp" test suite for exercising speculative parsing. (bnewman)
  • What if this suite reveals something bad?

Suspended items that may re-enter the list of release blockers

• Suspended bugs that were previously P2
  • 482917 Make HTML5 parsing not regress stream data availability to extensions
    • There's now a patch but no way to test it.
    • Suspended, because there are no known extensions that use the API in question and are compatible with Firefox 3.6.
    • Time estimate: Unknown. (Need to find an extension that actually uses this. The alternative is writing an extension just for testing.)
  • 540480 Hootsuite.com dashboard not loading with HTML5 parser turned on
    • Moved over to evangelism.
  • 537948 The procedure to fix a <a> around a element is acting weird.
    • Spec issue. No indication so far on whether this is a problem for real-world sites. The HTML5 behavior seems well enough aligned with the IE8 behavior.