Bug 73472

Summary:

[MutationObservers] Make WebKitMutationObserver::deliverAllMutations() tolerant of re-entrant calls

Product:

WebKit

Reporter:

Rafael Weinstein <rafaelw>

Component:

DOM

Assignee:

Rafael Weinstein <rafaelw>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

adamk, ojan, rniwa, sam, webkit.review.bot

Priority:

Version:

528+ (Nightly build)

Hardware:

Unspecified

OS:

Unspecified

Bug Depends on:

Bug Blocks:

68729

Attachments:

Description	Flags
Patch	none

Description Rafael Weinstein 2011-11-30 11:01:54 PST

In effect, deliverAllMutations() needs to prevent each delivery from, itself, causing deliverAllMutations().

The code happens to be structured right now in such a way that it's behavior is still semantically correct, but we should avoid having multiple calls on this static on the stack.

Comment 1 Rafael Weinstein 2011-11-30 11:06:27 PST

Created attachment 117228 [details]
Patch

Comment 2 Ojan Vafai 2011-11-30 11:10:15 PST

Comment on attachment 117228 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=117228&action=review

> Source/WebCore/dom/WebKitMutationObserver.cpp:139
> +        return;

Should we have an ASSERT_NOT_REACHED() before the return here? It's a bug in the C++ code if we ever hit this, right?

Comment 3 Rafael Weinstein 2011-11-30 11:19:35 PST

No. We probably will reach it. deliveryAllMutations is likely going to invoke script for each observer. The easiest thing is for each embedder to simply hook ALL outer-most script invocations. This guard just means that for each delivery, the embedder will *try* and quietly fail to re-start delivery (which is already underway).

Comment 4 WebKit Review Bot 2011-11-30 21:00:16 PST

Comment on attachment 117228 [details]
Patch

Clearing flags on attachment: 117228

Committed r101597: <http://trac.webkit.org/changeset/101597>

Comment 5 WebKit Review Bot 2011-11-30 21:00:20 PST

All reviewed patches have been landed.  Closing bug.