[webkit-dev] Unprefixing DOM MutationObservers, and looking for help from port maintainers

[Adam Klein]

On Fri, Jun 15, 2012 at 11:15 AM, Ryosuke Niwa <rniwa at webkit.org> wrote:
> Great!
>
> What is the difference in handling the style attribute?

It's this test:

http://trac.webkit.org/browser/trunk/LayoutTests/fast/mutation/observe-attributes.html#L741

Firefox mutates the attribute even though nothing changed. Per Olli on
#whatwg, they're just following the HTML spec here
(http://www.whatwg.org/specs/web-apps/current-work/multipage/elements.html#the-style-attribute):

"Mutating the CSSStyleDeclaration object must create a style attribute
on the element (if there isn't one already) and then change its value
to be a value representing the serialized form of the
CSSStyleDeclaration object."

I don't think this is likely to cause any problems that haven't
already cropped up over the years from this difference.

> Also, can we convert existing tests to use testharness.js and submit them to
> W3C? That way, we can help other browser vendors implement it "correctly"
> when they do. (of course, this shouldn't block unprefixing the API)

It'd definitely be nice to push these upstream, I agree. And I also
don't think this should block unprefixing :)

>
> - Ryosuke
>
> On Jun 15, 2012 11:09 AM, "Adam Klein" <adamk at chromium.org> wrote:
>>
>> I was working closely with Olli Pettay when he was putting together
>> the tests for Firefox, and we passed all of those (as of a couple
>> months ago; our code hasn't changed since then). Just ran through our
>> tests with Firefox Aurora and everything passed modulo differences in
>> supported features (they don't support WebSQL or FileSystem API which
>> we use in a few tests; there were some minor differences in handling
>> of the style attribute, but nothing that looks likely to cause
>> compatibility issues (whitespace)).
>>
>> I've uploaded a proposed patch at
>> https://bugs.webkit.org/show_bug.cgi?id=89231 (wanted to give it a
>> shot through the EWS bots). I've yet to implement a way to turn off
>> the unprefixing for ports that aren't yet done, suggestions on how to
>> do that most simply are appreciated.
>>
>> - Adam
>>
>> On Fri, Jun 15, 2012 at 10:48 AM, Ryosuke Niwa <rniwa at webkit.org> wrote:
>> > Before unprefixing it, can we make sure we also pass Mozilla's tests and
>> > vice versa? Mutation events was a huge mess partially because browsers
>> > didn't interoperate. I'd like to make sure this API interoperates well
>> > from
>> > day 1.
>> >
>> > - Ryosuke
>> >
>> > On Fri, Jun 15, 2012 at 10:31 AM, Adam Klein <adamk at chromium.org> wrote:
>> >>
>> >> An update: it appears that the GTK and EFL ports have added at least
>> >> basic support for end-of-task delivery, in
>> >> http://trac.webkit.org/changeset/108628 and
>> >> http://trac.webkit.org/changeset/110568. They now pass all
>> >> MutationObserver tests. Do these tests pass in the actual browsers
>> >> built from GTK and EFL, or only DRT?
>> >>
>> >> On Fri, Jun 8, 2012 at 2:16 PM, Adam Klein <adamk at chromium.org> wrote:
>> >> > Hi webkit-dev,
>> >> >
>> >> > DOM MutationObservers (see meta bug
>> >> > https://bugs.webkit.org/show_bug.cgi?id=68729) have been shipping as
>> >> > WebKitMutationObserver in Chromium since earlier this year. The
>> >> > feature is fully specced as part of DOM4
>> >> > (http://www.w3.org/TR/domcore/#mutation-observers) and is implemented
>> >> > in Firefox. Mozilla has also recently unprefixed their version of the
>> >> > API (https://bugzilla.mozilla.org/show_bug.cgi?id=749920), and I'd
>> >> > like to do the same (https://bugs.webkit.org/show_bug.cgi?id=85161).
>> >> >
>> >> > The tricky part is that while the Chromium version is complete and
>> >> > compatible with Firefox, there's one piece missing from all other
>> >> > ports (https://bugs.webkit.org/show_bug.cgi?id=78290). The short
>> >> > version is that each port needs to be able to run some code
>> >> > (delivering mutations) at the end of every task (see
>> >> >
>> >> >
>> >> > http://www.whatwg.org/specs/web-apps/current-work/multipage/webappapis.html#processing-model-2,
>> >> > step 4 "Perform a microtask checkpoint"). Without this code,
>> >> > mutations
>> >> > due to user input are not delivered in a timely fashion.
>> >> >
>> >> > It's easy for Chromium to do this because we have our own MessageLoop
>> >> > abstraction wrapping the native event loops on various platforms, so
>> >> > our definition of end-of-task is easy to define. But implementing
>> >> > this
>> >> > properly is likely to be slightly different for every port.
>> >> >
>> >> > In the short term, my plan is to add the unprefixed version (in
>> >> > addition to the prefixed) of MutationObserver when PLATFORM(CHROMIUM)
>> >> > is enabled. But I'd like to help other ports implement this
>> >> > appropriately, and provide the unprefixed version there too. Please
>> >> > let me know how I can be of assistance.
>> >> >
>> >> > Let me know if you have questions or concerns,
>> >> > Adam Klein
>> >> _______________________________________________
>> >> webkit-dev mailing list
>> >> webkit-dev at lists.webkit.org
>> >> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>> >
>> >