Page 1 of 1

2.02rc1 busts Back Button on arstechnica.com?

Posted: Tue Aug 10, 2010 7:19 pm
by BarryM
This seems to have started in the 2.01 timeframe, but happens as well with 2.02RC1.

1) Go to arstechnica.com
2) Click an article
3) Hit the back button

Doesn't want to navigate away and keeps putting the same page back in.

Mozilla/5.0 (Windows NT 5.1; rv:2.0b4pre) Gecko/20100810 Minefield/4.0b4pre (.NET CLR 3.5.30729)

Everything except for Google Analytics and doubleclick.net are trusted. I'm also using an older version of http://someonewhocares.org/hosts/ (Nov 5th, 2007 at 15:34)

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Wed Aug 11, 2010 1:01 am
by therube
I can confirm what you're seeing.

Though see if this gives any insight, Problem with Back Commands in IE8.

Or maybe not.
Cause it does not occur in -safe-mode. So perhaps there is something NoScript related causing it?

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Wed Aug 11, 2010 4:58 am
by al_9x
It is NS related, started in 2.0.1rc5.

Fx 3.6.8, new profile, block iframe on trusted, allow arstechnica.com
  1. If you look at the activity indicator on the tab, as you load an article, you can see what appear to be several refreshes (stop/starts). These were there prior 2.0.1rc5, but still not normal.
  2. starting with 2.0.1rc5 the back bug appears
  3. starting with 2.0.1rc5, arstechnica pages (articles and others) are added to the back/forward history (button drop down) multiple times (3)
  4. allowing iframes seems to make the above less frequent, but doesn't eliminate it completely.
  5. unrelated to the bug: the root page has a blocked flash object with the url of the page itself. I was curious as to what this was, did not look normal. Turns out this is a probe node that their script later (timer) tries to remove "FBD.fake_embed_1.parentNode.removeChild(FBD.fake_embed_1)" The removal fails because a) parentNode is null b) this dummy node is no longer in the dom anyway (calling body.removeChild fails with "Node was not found") Does it makes sense to fake parentNode and intercept removeChild (on the prototype) to allow these probe objects to be removed? This may be a recurring pattern.

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Wed Aug 11, 2010 9:42 am
by Giorgio Maone
Does the problem go away if you disable surrogates (noscript.surrogate.enabled = false in about:config)?

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Wed Aug 11, 2010 10:07 am
by al_9x
Giorgio Maone wrote:Does the problem go away if you disable surrogates (noscript.surrogate.enabled = false in about:config)?
yes, back works and multiple history entries are gone, however the multiple stop/start thing is still there, do you know why it's doing that

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Wed Aug 11, 2010 10:23 am
by Giorgio Maone
al_9x wrote: however the multiple stop/start thing is still there, do you know why it's doing that
Why are you saying it's not normal?

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Wed Aug 11, 2010 11:31 am
by al_9x
Giorgio Maone wrote:
al_9x wrote: however the multiple stop/start thing is still there, do you know why it's doing that
Why are you saying it's not normal?
I must have only tried it once without NS. Now I tried it multiple times and it does happen without NS as well, although less frequently and it seems only single stop/start vs. two with NS. So, not a problem, then.

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Wed Aug 11, 2010 7:22 pm
by BarryM
Yes, turning off surrogates makes the back button problem go away for me.

Upgraded to 2.02RC2 and problem persists with surrogates on. Also happens with NYtimes.com (I only looked at the Business section).

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Thu Aug 12, 2010 9:08 am
by Giorgio Maone

Re: 2.02rc1 busts Back Button on arstechnica.com?

Posted: Thu Aug 12, 2010 2:01 pm
by al_9x
Giorgio Maone wrote:Fixed in latest development build
confirmed