Brendan has debugged this and is patching, he asked me to file this.

See also bug 326466 comment 126.

At least these two: 1. XPCWrappedNative::GetNewOrUsed calls the scriptable helper's PostCreate method and ignores it r.v. 2. nsGlobalWindow::RunTimeout ignores the r.v. from CallEventHandler. Testing more, I'll patch as soon as the bloatcycle page works again. /be

Actually, nsGlobalWindow::RunTimeout is ok, it's private to that class (protected at any rate), and called only from a timer callback (from the main event loop). So there is no way JS can be active on the JSContext it uses, and therefore the JS API will report any pending exception as it unwinds back to CallEventHandler. I've beefed RunTimeout up with an assertion to this effect. Still not bloatcycling for some reason... /be

Not seeing bloatcycle.html actually cycle, yet. /be

mrbkap has been tracking the XPCWrappedNative::GetNewOrUsed fix in bug 326497. /be

(In reply to comment #5) > mrbkap has been tracking the XPCWrappedNative::GetNewOrUsed fix in bug 326497. Rather, I've dup'ed bug 326497, in part. I'll build on mrbkap's patch as soon as he refreshes it. /be

This should go in ASAP to cure the orange on balsa. I'll leave it to mrbkap to make the better nsDOMClassInfo.cpp fix that his comment describes, for bug 326497. /be

Fixes a botch in nsScriptLoader.cpp's patch that Blake pointed out. /be

Comment on attachment 224407 [details] [diff] [review] patch, v2 r=mrbkap on the parts that I didn't write. The parts that I did already have r=jst/shaver, so this really isn't patch-cest.

To get balsa back, I checked in with r=mrbkap. /be

this checkin didn't show up on the trunk tinderboxen (they're closed anyway)