1360230 - MutationObserver recursively triggers on non-mutating element.classList function calls

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[The 8472]

The following code snippet causes Nightly to execute observer callbacks forever, on stable the observer callback is only fired once:

el=document.createElement("div"); mutate = () => {el.classList.toggle("foo", true)};obs = new MutationObserver((x) => {console.log(x); mutate()}); obs.observe(el, {"attributes": true}); mutate();

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

TokenList impl has changed recently. Note, setting attribute value to its currently value is supposed to trigger mutation observer.

Comment 2

[The 8472]

Chrome and Edge both behave like FF stable and I have spotted this behavior by code in the wild relying on the non-recursive behavior.

Comment 3

[Olli Pettay [:smaug][[e-mail]]]

Per spec we're doing right now, as far as I see.

https://dom.spec.whatwg.org/#dom-domtokenlist-toggle always calls 
6. Run the update steps. 


"A DOMTokenList object’s update steps
are to set an attribute value for the associated element using associated attribute’s local name and the result of running the ordered set serializer for token set. "

And since attribute value is set, MutationRecord is scheduled.


( Looks like some patch has landed recently and searchfox hasn't been updated yet. )

Comment 4

[Olli Pettay [:smaug][[e-mail]]]

Do you know what Safari is doing? Since I believe Safari was supposed to be updated to follow the latest spec.

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

Aryeh, if the spec isn't web compatible, could you please find some time to back out problematic patches. please.

Comment 6

[The 8472]

(In reply to Olli Pettay [:smaug] from comment #4)
> Do you know what Safari is doing? Since I believe Safari was supposed to be
> updated to follow the latest spec.

No, I don't have a safari to test with.

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

I think this is a regression from Bug 1359780.

Comment 8

[Olli Pettay [:smaug][[e-mail]]]

I need to still verify the regression range.

Comment 9

[Olli Pettay [:smaug][[e-mail]]]

Backed out bug 1359780.

Comment 10

[Aryeh Gregor (:ayg) (no longer with Mozilla)]

I don't understand, what was the problem here?  This behavior is expected and follows the standard.  Did a real-world site break?  I don't see anywhere in the bug report that says why we don't want this behavior.

Comment 11

[The 8472]

(In reply to Aryeh Gregor (:ayg) (next working August 2-22) from comment #10)
> Did a real-world site break?

I mentioned that in comment 2.

> This behavior is expected and follows the standard. 

Wasn't a spec-change which triggered this change in the first place?

Comment 12

[Aryeh Gregor (:ayg) (no longer with Mozilla)]

Oops, I missed that, sorry.  Okay, then the backout is fine.