Jeff Muizelaar [:jrmuizel] Reporter
	Description • 13 years ago

A profile shows us spending our time here:

6498.0ms   27.1%	 	js::PropertyTable::search(long, bool)
4830.0ms   20.1%	 	Encode(JSContext*, JSString*, unsigned short const*, unsigned short const*, JS::Value*)
3068.0ms   12.8%	 	js_LookupProperty(JSContext*, JSObject*, long, JSObject**, JSProperty**)

This doesn't happen on Aurora.

	Alice0775 White
	Comment 1 • 13 years ago

Confirmed On Windows7 too.
Browser unresponsive for a while and use RAM more than 900MB. 
However browser become responsive after 50-60sec.

http://hg.mozilla.org/mozilla-central/rev/46a6d0fd13d5
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:10.0a1) Gecko/20111013 Firefox/10.0a1 ID:20111013030957


Regression window(m-c)
Works:
http://hg.mozilla.org/mozilla-central/rev/6751379365e3
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:10.0a1) Gecko/20111010 Firefox/10.0a1 ID:20111010024953
Hangs:
http://hg.mozilla.org/mozilla-central/rev/aee2bf4eb5f8
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:10.0a1) Gecko/20111010 Firefox/10.0a1 ID:20111010025555
Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=6751379365e3&tochange=aee2bf4eb5f8
Suspected bug:
Bug 648801

	Alex Keybl [:akeybl]
	Comment 3 • 13 years ago

[Triage Comment]
What are the next steps in the investigation of this bug? We'd likely take a patch on Aurora if it landed before the cutover.

	Boris Zbarsky [:bzbarsky]
	Comment 4 • 13 years ago

Hrm.  Normally the next step would be a profile, but I can't seem to reproduce the issue on m-c.  Is there something to be done here to reproduce other than loading the page in the url field?  Jeff, Alice0775?

Note that I think not tracking this is the wrong call: if it's really a regression from the nodelist patch and we don't have a fix for it, we should consider preffing off the nodelist code for 10.  That's why we have a preference there!  Renominating based on that.

	Alex Keybl [:akeybl]
	Comment 5 • 13 years ago

(In reply to Boris Zbarsky (:bz) from comment #4)
> Hrm.  Normally the next step would be a profile, but I can't seem to
> reproduce the issue on m-c.  Is there something to be done here to reproduce
> other than loading the page in the url field?  Jeff, Alice0775?
> 
> Note that I think not tracking this is the wrong call: if it's really a
> regression from the nodelist patch and we don't have a fix for it, we should
> consider preffing off the nodelist code for 10.  That's why we have a
> preference there!  Renominating based on that.

I have no issue with taking a fix for this bug - but the tracking flag implies that we would slam on the brakes if this was not resolved in some way before shipping. Do we have reason to believe that more than just dartlang is affected by this bug?

	Boris Zbarsky [:bzbarsky]
	Comment 6 • 13 years ago

My take on it is that _if_ this is a regression from the nodelist changes then we should disable the new nodelist code before shipping, because we have no reason to believe that only dartlang is affected...

	Jeff Muizelaar [:jrmuizel] Reporter
	Comment 7 • 13 years ago

I can't reproduce this anymore. If someone is interested, they could try to do a bisection to see what fixed it, otherwise we might as well close it.