1007649 - querySelector for id on shadowRoot returns null

Status: RESOLVED FIXED

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

Description

[Wilson Page [:wilsonpage]]

myElement.shadowRoot.querySelector('#some-id'); //=> null
myElement.shadowRoot.getElementById('some-id'); //=> HTMLElement

Comment 1

[Boris Zbarsky [:bzbarsky]]

This is because querySelector makes certain optimizations for ID selectors when the node is in a document (mapping to document.getElementById in that case).  And shadow roots lie about being in a document: see bug 992521.

That said, we should perhaps disable that querySelector optimization when aRoot is anonymous!

Wilson, can you please attach the testcase you were using to reproduce this bug?

Comment 2

[Wilson Page [:wilsonpage]]

http://jsbin.com/mayuxu/1/edit?html,javascript,console

Comment 3

[Boris Zbarsky [:bzbarsky]]

OK, this stuff is just totally busted.  Clearly no one has ever written a test trying to querySelector on a shadowroot, since it immediately fails an assertion in FindMatchingElements:

  NS_ASSERTION(aRoot->IsElement() || aRoot->IsNodeOfType(nsINode::eDOCUMENT) ||
               !aRoot->IsInDoc(),
               "The optimization below to check ContentIsDescendantOf only for "
               "elements depends on aRoot being either an element or a "
               "document if it's in the document.");

Furthermore, the shadow root does not claim to be in an anonymous subtree, so the obvious fix I was going to try won't even work.

We really need to fix bug 992521.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Attachment: The anonymous content thing plus test

Of course this doesn't fix the bug, because ShadowRoot is just completely fucked up

Comment 5

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Boris, would this have to do with the crash seen here: 
https://crash-stats.mozilla.com/report/index/f834ff19-e794-4eed-a9e3-4266a2140617
We only see the crash on the 6/11 build.

Comment 6

[Boris Zbarsky [:bzbarsky]]

No.  In fact, I would expect this bug to be fixed now that bug 992521 is fixed.  Wilson, can you confirm?

That crash is some other issue that you should probably file a separate bug on.

Comment 7

[Wilson Page [:wilsonpage]]

Confirmed. Original test case in comment 2 is now working as expected.

Comment 8

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Thanks Boris.  I'll file it.