Reintegrate 552 r66225-r66645....

Reintegrate 552 r66225-r66645.

------
66466 17.11.2010 19:09:22, by bauerb@chromium.org
Tentative compile fix after merge error.

TBR=kerz
-----
66453 17.11.2010 18:17:09, by bauerb@chromium.org
Merge 65953 -- Move click-to-play to about:flags.

XIB changes: Add an outlet |pluginDefaultSettingMatrix_| to ContentSettingsDialogController, hooked up to the associated matrix, to remove the click-to-play radio button.

While I'm at it, clean up a bit:
* Remove the old --disable-click-to-play flag that reverted to the M6 behavior for blocked plugins
* Make ContentExceptionsWindowController use ContentSettingComboModel for the action popup.
* Make HostContentSettingsMapTest use AutoReset to reset command line switches.

BUG=62091
TEST=unit_tests

Review URL: http://codereview.chromium.org/4643007
-----
66581 18.11.2010 03:23:22, by cevans@chromium.org
Merge 66567 - Fix XPath bug from upstream.

BUG=63444
TEST=See bug

Review URL: http://codereview.chromium.org/5196003

TBR=cdn@chromium.org
Review URL: http://codereview.chromium.org/5216001
-----
66642 18.11.2010 19:14:57, by thakis@chromium.org
Merge 66631 - Mac: Only clear the background of CoreAnimation plugins if we're going to draw into them.

Previously, the logic was:

1.) Clear plugin background
2.) If the plugin didn't update, return early
3.) Paint plugin
4.) "Swap buffers"

But the "Swap buffers" step only unbound and rebound an FBO object If the plugin didn't change, its backing store would contain transparent black, and if the graphics driver decided to flush the FBO for another reason than a "swap buffers" call, the blackness would show up in the browser.

This CL swaps steps 1 and 2, so even if the FBO is flushed for some unrelated reason, we display something valid.

BUG=60341
TEST=Open the file attached to the bug. Resize the window for a few minutes, put the computer to sleep and back on, resize window for a few more minutes. The plugin area (the 191x60px rect in the middle) shouldn't become black. YouTube should still work. CPU usage shouldn't be worse than it was before for the browser, plugin, and renderer processes.

Review URL: http://codereview.chromium.org/5220002

TBR=thakis@chromium.org
Review URL: http://codereview.chromium.org/5167003
-----
66645 18.11.2010 19:30:53, by willchan@chromium.org
Merge 66630 - Fix SPDY crash on race when canceling a stream that just got created.

When I fixed the code not to be re-entrant (since that caused crashes) in r61880, I created a window when the pending create callback was posted to the MessageLoop to be run on the next iteration. In this window before it actually gets invoked, if the pending stream creation got cancelled, then the callback wasn't cancelled, so we would execute a callback on a cancelled stream creation, which can cause crashes.

The fix is to keep track of the pending callbacks. Cancellation of pending stream creations check this pending callback map first.

BUG=63532
TEST=See bug thread for repro steps. New unit test added as well.

Review URL: http://codereview.chromium.org/5174005

TBR=willchan@chromium.org
Review URL: http://codereview.chromium.org/5216002
-----



Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=66868