Change Details

These tests pass consistently with `make check` but fail `ninja check` consistently with the following setup: ``` gcc 8.1.0 clang 7.0.1 boost 1.58.0 ``` The failure occurs at step `i = 9`, where `registerPoll()` is called, setting inflight to `10` and then immediately after, the call to `shouldPoll()` returns `false`. Based on reading of the test, it seems clear that shouldPoll() should be evaluated before registerPoll() is called, but apparently compiler differences evaluate the BOOST_CHECK_EQUAL macro differently.

These tests pass consistently with `make check` but fail `ninja check` consistently with the following setup: ``` gcc 8.1.0 clang 7.0.1 boost 1.58.0 ``` The failure occurs at step `i = 9`, where `registerPoll()` is called, setting inflight to `10` and then immediately after, the call to `shouldPoll()` returns `false`. Based on reading of the test, it seems clear that shouldPoll() should be evaluated before registerPoll() is called, but apparently compiler differences evaluate the BOOST_CHECK_EQUAL macro differently. Before writing this diff, I attempted to write the test in such a way as to evaluate shouldPoll() and registerPoll(), in that order, in intermediate variables. This test fails for ninja and make as expected.

These tests pass consistently with `make check` but fail `ninja check` consistently with the following setup: ``` gcc 8.1.0 clang 7.0.1 boost 1.58.0 ``` The failure occurs at step `i = 9`, where `registerPoll()` is called, setting inflight to `10` and then immediately after, the call to `shouldPoll()` returns `false`. Based on reading of the test, it seems clear that shouldPoll() should be evaluated before registerPoll() is called, but apparently compiler differences evaluate the BOOST_CHECK_EQUAL macro differently. Before writing this diff, I attempted to write the test in such a way as to evaluate shouldPoll() and registerPoll(), in that order, in intermediate variables. This test fails for ninja and make as expected.