> On Dec. 8, 2016, 2:35 p.m., Till Toenshoff wrote: > > src/tests/slave_tests.cpp, line 1491 > > <https://reviews.apache.org/r/54205/diff/1/?file=1573012#file1573012line1491> > > > > Would it make sense to not cast into an int at all here, now that we > > got the epsilon covered?
Oh it definitely would. - Benjamin ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/54205/#review158520 ----------------------------------------------------------- On Dec. 8, 2016, 2:57 p.m., Benjamin Bannier wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/54205/ > ----------------------------------------------------------- > > (Updated Dec. 8, 2016, 2:57 p.m.) > > > Review request for mesos, Joseph Wu and Till Toenshoff. > > > Bugs: MESOS-4695 > https://issues.apache.org/jira/browse/MESOS-4695 > > > Repository: mesos > > > Description > ------- > > This test was checking two floating point values for equality by only > examining their integer part. Since one of the values went through > a number of conversions, even its integer part could have changed for > certain starting values. > > This patch introduces an epsilon into the comparison to account for > possible changes in the integer part. > > > Diffs > ----- > > src/tests/slave_tests.cpp 967c7307b6fc8e9eb4490f51938bdcf286b5e7a6 > > Diff: https://reviews.apache.org/r/54205/diff/ > > > Testing > ------- > > `make check` passes. > > I was not able to make this test fail even without the fix. After the fix it > continued to work for me, even for a large number of test iterations. > > > Thanks, > > Benjamin Bannier > >
