As was done for the jsreftest suite in bug 637400, but for the jit-test suite.

If we look at jit-tests (js/src/jit-test/jit_tests.py) it calls jittests.run_test_remote and jittests.run_tests so we need to do check in the both the functions separately that the output for different flag is not different. Am I correct?

How about this one?

Comment on attachment 8780642 [details] [diff] [review] try_v1.diff The patch file is empty. Don't forget to qref.

I forgot to refresh the patch earlier :P

Comment on attachment 8780710 [details] [diff] [review] try_v1.diff Review of attachment 8780710 [details] [diff] [review]: ----------------------------------------------------------------- ::: js/src/tests/lib/jittests.py @@ +513,5 @@ > + if res.test.path in output_dict.keys(): > + if output_dict[res.test.path] != res.out: > + pb.message("FAIL - OUTPUT DIFFERS {}".format(res.test.relpath_tests)) > + else: > + output_dict[res.test.path] = res.out This is almost perfect. However, by putting this above where we print the program output, it might not get seen by a reader. Instead let's move it down below to where we emit errors with pb.message. How about just above |doing = 'after {}'.format(res.test.relpath_tests)|?

Fixed it :)

Comment on attachment 8781172 [details] [diff] [review] try_v2.diff Review of attachment 8781172 [details] [diff] [review]: ----------------------------------------------------------------- Excellent!

Pushed by ryanvm@gmail.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/772ec8c912ac Check that the output of all JITFLAGS is the same on jit-tests. r=terrence