Use [TreatNullAs=EmptyString] for DOMString attributes with [Reflect]

Use [TreatNullAs=EmptyString] for DOMString attributes with [Reflect]

In cases where the HTML specification uses [TreatNullAs=EmptyString] on
an IDL attribute that reflects a content attribute, use the same in our
IDL files to get correct behavior when assigning null.

Previous incorrect behavior when assigning null was to remove the
reflected content attribute, if present, and otherwise do nothing.

The new behavior matches Firefox and Opera/Presto. IE11 behaves the same
when the content attribute is present, but doesn't create the content
attribute if it's missing and null is assigned to the IDL attribute.

This change impacts performance significantly (~4x slowdown) in the case
of assigning null to the IDL attribute when the reflected content
attribute is not present, i.e. the previous "do nothing" case. This ought
to be considered an irrelevant micro-benchmark, and the reason it slows
down is that the incorrect (web-facing) behavior is a no-op while the
correct behavior is not.

Performance in other cases, in particular the case of assigning null to
the IDL attribute when the reflected content attribute is present, show
no significant change.

BUG=391194
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=177728

[Jens Widell, 2014-07-07 09:59:02]

PTAL

The modified (and tested) attributes are all those that have
[TreatNullAs=EmptyString] in the current HTML specification
(http://whatwg.org/html). Firefox nightly passes all added tests, as does
Opera/Presto.

I've left some reflected attributes with [TreatNullAs=NullString] for now. They
are typically non-standard, and either not supported by Firefox or has
null->"null" behavior in Firefox.

[haraken, 2014-07-07 10:42:06]

LGTM

Let's land this CL and see what benchmarks regress in perf bots. Overall, I
don't think people are interested in performance regressions caused by this
change.

Let's get approvals from multiple reviewers before landing this change.

+arv, +tkent, +philipj

[tkent, 2014-07-07 23:58:20]

lgtm

[philipj_slow, 2014-07-08 19:57:25]

LGTM, getting these and other IDL files in lines with the spec was really the
reason I took a shot at [TreatNullAs=EmptyString] a while back.

[Jens Widell, 2014-07-09 09:09:04]

On 2014/07/08 19:57:25, philipj wrote:
> LGTM, getting these and other IDL files in lines with the spec was really the
> reason I took a shot at [TreatNullAs=EmptyString] a while back.

Thanks you all.

I'll queue this; let's hope arv isn't of a differing opinion. :-)

[Jens Widell, 2014-07-09 09:09:08]

The CQ bit was checked by jl@opera.com

[commit-bot: I haz the power, 2014-07-09 09:09:27]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/jl@opera.com/371833004/20001

[commit-bot: I haz the power, 2014-07-09 09:14:08]

Change committed as 177728