...
I can get it to repro more frequently if I increase the load on the machine.
It's a FTL miscompile on chainTest#BRCzIm, which has 430 bytecode instructions.
Reduced test case: function Foo(a, b) { this.f = a.f; this.g = b.f + 1; } function foo(a, b) { return new Foo(a, b); } noInline(foo); for (var i = 0; i < 100000; ++i) { var result = foo({f:1}, {f:2}); if (result.f != 1) throw "Error: bad result.f: " + result.f; if (result.g != 3) throw "Error: bad result.g: " + result.g; } var result = foo({f:1}, {f:2.5}); if (result.f != 1) throw "Error: bad result.f: " + result.f; if (result.g != 3.5) throw "Error: bad result.f: " + result.g;
Created attachment 223681 [details] the patch
Comment on attachment 223681 [details] the patch Attachment 223681 [details] did not pass mac-wk2-ews (mac-wk2): Output: http://webkit-queues.appspot.com/results/6337216930381824 New failing tests: http/tests/misc/last-modified-parsing.html
Created attachment 223685 [details] Archive of layout-test-results from webkit-ews-11 for mac-mountainlion-wk2 The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews. Bot: webkit-ews-11 Port: mac-mountainlion-wk2 Platform: Mac OS X 10.8.5
Comment on attachment 223681 [details] the patch r=me
Landed in http://trac.webkit.org/changeset/163789