Ignore the HTML parser's borrow flag in GC tracing

[kmcallister]

r? @mbrubeck (who hit the bug), @pcwalton (who advised me on the safety of possible solutions)

[hoppipolla-critic-bot]

Critic review: https://critic.hoppipolla.co.uk/r/2900

This is an external review system which you may optionally use for the code review of your pull request.

In order to help critic track your changes, please do not make in-place history rewrites (e.g. via git rebase -i or git commit --amend) when updating this pull request.

[kmcallister]

r=mbrubeck

[mbrubeck]

@bors: retry

[pcwalton]

@bors: retry

[bors-servo]

saw approval from mbrubeck
at kmcallister@0debc87

[bors-servo]

merging kmcallister/servo/tokenizercell = 0debc87 into auto

[bors-servo]

kmcallister/servo/tokenizercell = 0debc87 merged ok, testing candidate = 3eea7ed

[bors-servo]

some tests failed:
failure: http://build.servo.org/builders/linux/builds/363

[bors-servo]

saw approval from mbrubeck
at kmcallister@0debc87

[bors-servo]

some tests failed:
failure: http://build.servo.org/builders/linux/builds/379

[bors-servo]

saw approval from mbrubeck
at kmcallister@0debc87

[bors-servo]

merging kmcallister/servo/tokenizercell = 0debc87 into auto

[bors-servo]

kmcallister/servo/tokenizercell = 0debc87 merged ok, testing candidate = 2d1996d

[bors-servo]

some tests failed:
failure: http://build.servo.org/builders/linux/builds/381

[mbrubeck]

/old-tests/submission/Opera/script_scheduling/039.html
------------------------------------------------------
FAIL  scheduler: IFRAMEs added with DOM
OK expected CRASH [Parent]

This looks like a genuine fix caused by this patch. You'll need to git rm tests/wpt/metadata/old-tests/submission/Opera/script_scheduling/039.html.ini.

[jdm]

It involves iframes, which makes me suspect an intermittent pass. Hopefully that will be corrected by #3759.

[mbrubeck]

Needs a rebase now that #3759 has landed. If old-tests/submission/Opera/script_scheduling/039.html is still intermittent with this patch, we might need to skip the test for now. #3721 also seems to affect that test case, so maybe we'll be able to re-enable it once both PRs are in.

[kmcallister]

Hm, maybe I should just add the ability to break the rules to DOMRefCell, since it will be in upstream RefCell eventually (rust-lang/rust#18212)