This is probably due to a bug in recording, but the general idea is: $ cat testa.js var obj = {} var k; var i; var f = obj.f = function() { for (i = 0; i < 5; ++i) k = this; return k; } obj.f() assertEq(f().isInner, false); $ cat testb.js var obj = {} var k; var i; var f = obj.f = function() { print(); for (i = 0; i < 5; ++i) k = this; return k; } obj.f() assertEq(f().isInner, false); $ tracer-obj/js -j -z testa.js $ tracer-obj/js -j -z testb.js testb.js:6: TypeError: Assertion failed: got true, expected false

I just noticed: initializing |k = {}| and |i = 0| before the calls makes the bug disappear.

The tests in comment 0 are unit tests for bug 513160. Both tests now pass. To reproduce this bug you must revert the fix for bug 513160: hg export 2b8f660c66cd | patch -p1 -R Then you might reasonably think both tests should fail. Actually testa passes and testb fails. I can't reproduce what Blake says in comment 1. So I blame this on bug 513160. I suspect what's happening here is that calling print(); does something to the bottommost stack frame that causes the bug not to happen. So of course the test then passes. Note that it's not anything specific about print(). Call any global function and the test is disrupted. Tentatively this WORKSFORME. Blake, is this analysis completely bogus?