See attached testcase. It triggers aborts like so: abort: 5500: fp->scopeChain is not global or active call object Abort recording of tree /Users/bzbarsky/test.js:3@17 at /Users/bzbarsky/test.js:12@0: name. Note that if I inline looper() in the testcase (so replace that function call to looper() with two nested loops with f() called in the inner loop) then we trace this.

fwiw, I tried that patch on the go game. It fixed all the board.js scope chain aborts except for the two that actually change the value of a variable the close over.

Comment on attachment 377336 [details] testcase Not this bug's test, naming f changes semantics and differs from the go game. I'll put this in a new bug. /be

(In reply to comment #4) > I'll put this in a new bug. bz kindly offered to do the deed. /be

Filed bug 492918 on that first testcase.

Fixed in tm: http://hg.mozilla.org/tracemonkey/rev/010768e6119c /be

We should take this for 1.9.1 after a bit of baking. /be

This has been backed out from 1.9.1: http://hg.mozilla.org/releases/mozilla-1.9.1/rev/f497e113346f

(In reply to comment #11) > This has been backed out from 1.9.1: > > http://hg.mozilla.org/releases/mozilla-1.9.1/rev/f497e113346f yep, sorry I missed the note here.

This caused regression bug 495045.