> If tests are expected to fail, then they should be in "%!xtest" sections instead of "%!test" sections.
This depends on how one interprets "expected to fail". The ones
failing are not expected to fail, they are bugs that we are real bugs
that we are aware of. They should still be fixed, it just happens
that we haven't got around fixing those issues either because we do
not know what Matlab is actually doing to achieve those results or
because no one has found the time to work on them.
My feeling is that marking tests for known bugs as expected to fail
helps in having them never fixed but it also seems to be causing a lot
of confusion. It makes really hard to use the test suite to identify
regressions which reduces a lot the usefulness of tests so I will be
marking this as xtest.
|
I think those two failing tests in edge.m (for the automatic threshold value in the "Canny" method) are intentional.
When I remember Carne right, they should document the remaining Matlab-incompatibility of this Octave function. See bug #46099 for details.
(Note: I think I heard Carne say that intentionally failing tests in the image package are not so uncommon...)
|
The two last unit tests for function edge in the image package are failing. The test log is below.
|