After getting the upgrade to beta2, my Firefox has been frequently going into "I'm going to eat half your CPU mode." I haven't yet figured out what triggers this to happen, but in my Console I see dozens of these messages per second Dec 21 14:46:51 lilblackbk /Applications/Firefox.app/Contents/MacOS/firefox-bin: CGImageCreate: invalid image provider: NULL.

I strongly suspect that roc's post addresses this bug. I did have gmail open each time I saw the CPU issue. If he has a better bug open for this, feel free to dupe. http://weblogs.mozillazine.org/roc/archives/2007/12/driveby_debuggi.html

Confirming reported error with SeaMonkey (on MacOSX 10.4): { http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1238867971.1238873759.28604.gz&fulltext=1 OS X 10.4 comm-central unit test on 2009/04/04 10:59:31 *** 12854 INFO Running /tests/content/canvas/test/test_2d.composite.clip.destination-atop.html... CGImageCreate: invalid image provider: NULL. *** 12858 INFO Running /tests/content/canvas/test/test_2d.composite.clip.destination-in.html... CGImageCreate: invalid image provider: NULL. *** 12874 INFO Running /tests/content/canvas/test/test_2d.composite.clip.source-in.html... CGImageCreate: invalid image provider: NULL. *** 12878 INFO Running /tests/content/canvas/test/test_2d.composite.clip.source-out.html... CGImageCreate: invalid image provider: NULL. } Slightly different message with Firefox 3.6/3.5 (on MacOSX 10.5): { Sat Apr 4 10:29:57 bm-xserve19.build.mozilla.org firefox-bin[67288] <Error>: CGImageCreate: invalid image size: 0 x 0. Sat Apr 4 11:31:07 bm-xserve16.build.mozilla.org firefox-bin[20623] <Error>: CGImageCreate: invalid image size: 0 x 0. }

er, I'm confused -- those are two different reports; Joey's report is of dozens of the CGImageCreate messages per second, your report is one message per test for a few tests, no?

FWIW, I haven't seen this in 3.0.x in a long while. The "dozens per second" part should probably be considered worksforme, if no one else can reproduce.

Minusing per Joey's comments -- we should probably still look at the CGImageCreate warnings, but they're not critical (CGImageCreate returns null correctly in those cases).. but in a separate bug.