[cfe-dev] clang: not deterministic anymore?

[Giangiacomo Mariotti]

Hi all,
         today I was doing some more compilation tests with clang(both
in debug and release version) and I found out that clang, as of
version 85898, isn't deterministic anymore. Is that acceptable?
I compiled the same code with the same options with clang in both
debug and release version(of clang) and each time I got a different
file(checked with sha1sum).

49dfb5940f87f9d471b45a788523dbf4480d1e0c  time_tests_clang-debug_O0
d5c793b07e723517d7d896e4013921d67ebdc55b  time_tests_clang-debug_O0_2
995027f7b9cbd7b0aba81f8c6da0ee4c14e3fa23  time_tests_clang-debug_O0_3
63c9f13d11e43fd5821ff075be5d652a1af2f160  time_tests_clang-release_O0
4b5dff1f450e3c96627c8e30516c949e0060719d  time_tests_clang-release_O0_2
e6b3db5b666a45ac0bfc30481b45d9e668ad76c9  time_tests_clang-release_O0_3
2b8e87d9e0eeb1e8e213d5ebf064949b643accde  time_tests_gcc_O0
2b8e87d9e0eeb1e8e213d5ebf064949b643accde  time_tests_gcc_O0_2
2b8e87d9e0eeb1e8e213d5ebf064949b643accde  time_tests_gcc_O0_3

I guess there's a bug somewhere, which was introduced after version
85507, which I previously tested and didn't seem to have this problem.

Giangiacomo