tracking bug for bugs blocking default setting of fastback to on

Adding a few bugs I know need to be fixed; more coming soon probably.

mac blocker: bug 298112

plus to get it on the list at least.

pre-requesting review and approval to enable this once the blocking bugs are landed

sicking and I decided that 293235 is not really an alpha blocker

talking with bryner https://bugzilla.mozilla.org/showdependencytree.cgi?id=274784&hide_resolved=1 has the remaining outstanding issues that could be worked out after flipping the switch.

the current state of that list looks like * bug 292890 [---, adamlock@eircom.net] -Investigate interaction of bfcache/fastback and midas. * bug 292938 [---, nobody@mozilla.org] -Slight hitch when using back/forward with bfcache/fastback. * bug 292945 [---, general@dom.bugs] -bfcache/fastback code can synchronously trigger off a timeout. * bug 292955 [---, bryner@brianryner.com] -fastback/bfcache doesn't work well with framesets. * bug 292958 [---, bryner@brianryner.com] -Blink timers not disabled in cached presentations. * bug 292959 [---, general@dom.bugs] -What to do with computed style objects for DOM nodes in a cached presentations?. * bug 292965 [---, nobody@mozilla.org] -Improve eviction algorithm for fastback/bfcache. * bug 292969 [---, nobody@mozilla.org] -Thawing presshell could synchronously reenter nsDocShell::RestorePresentation. * bug 292970 [---, nobody@mozilla.org] -Firing restore events can reenter nsDocShell::RestorePresentation. * bug 292976 [---, nobody@mozilla.org] -Various PLEvents and timers can fire on cached presentations. * bug 292977 [---, general@dom.bugs] -Restoring window state can reenter state restoration. * bug 292978 [---, nobody@mozilla.org] -Firing restore events can mutate the docshell tree. * bug 292980 [---, nobody@mozilla.org] -Recursion into frames when freezing/thawing presshell should happen in the document. * bug 292998 [---, nobody@mozilla.org] -fastback/bfcache should be invalidated after a text resize. * bug 293175 [---, general@dom.bugs] -Inconsistent document container handling. * bug 293235 [---, nobody@mozilla.org] -when using the back button (or keyboard), visited links are not marked as visited. * 293386 [---, bryner@brianryner.com] -Audit UnbindFromTree codepaths to make sure teardown change is OK. * bug 293572 [---, nobody@mozilla.org] -Local Proxy (AdMuncher) breaks bfcache. * bug 294610 [---, bryner@brianryner.com] -fastback/bfcache remembers DOM state after event processing on page quit. * bug 295591 [---, nobody@mozilla.org] -blazingly fast for/backward overwrites? svg objects. * bug 295931 [---, nobody@mozilla.org] -With bfcache enabled, rocker navigation stops text entry into textarea. * bug 296575 [---, nobody@mozilla.org] -RSS statusbar icon not (re)enabled after pressing back using bfcache. * bug 296760 [---, nobody@mozilla.org] -[MAS, v1.8b2-0605++] (new) "Various" crashes with bfcache ON. * bug 297488 [---, nobody@mozilla.org] -Irregular crashes when closing tab/windows or quiting Camino with bfcache enabled [@ nsChildView::GetParent() + 0x24]. * bug 297887 [---, bryner@brianryner.com] -Form values are not correct with bfcache enabled. * bug 298077 [---, events@dom.bugs] -Link remains focused when going back to the previous page using the back button and the focus can not be undone.. * bug 298125 [---, nobody@mozilla.org] -Back/forward cache becomes confused on bug lists. * bug 298622 [---, nobody@mozilla.org] -find fails on page with url typed.

bug 298377 DP 1.1a1 breaks dojo.io.bind's back-button interception should also get sorted out before 1.1 final too...

anything else that has turned up in recent testing with the pref turned on?

I'm not aware of any problems that aren't on the 274784 list. I think we should go ahead and flip the switch.

also see bug 298622 (recent bfcache-only regression).

byner can you look at https://bugzilla.mozilla.org/show_bug.cgi?id=299008

that bug claims to be reproducable with fastback disabled.

OK, it looks like we're ready to set fastback pref'd on for 1.8b3. We're using this alpha release as an oppportunity to obtain wider testing and quality feedback, to evaluate whether or not it's enabled in 1.1. final code review to be lead by bz and brendan by 1.8b4.

Comment on attachment 187171 [details] [diff] [review] patch 3 not 5? ok, r=darin

checked in, thanks guys.

I feel that bug 292965 should have been a blocker for this... The current eviction algorithm leads to quite painful results any time multiple tabs/windows are used, and if we're going to test we should test with the "final" eviction algorithm.

This broke screen reader usage, see bug 300642.