Currently, we have a test coverage on coveralls of about 77% which is indeed not bad. The remainder of about 23% comes mostly from two problems:
we do not check for failures and exception-raising yet. Thus all our sanity checs are never triggered
due to some coveralls bug the definition line of cdef functions counts as not visited even though the function body itself was examined as visited.
we simply do not cover a particular piece of code yet
The first issue can be resolved by add failure-checks where we'd execute some code provoking some specific failure and report if exceptions are not raised as we would expect them to be. The second issue however is up to coveralls but we might see if we could so something to avoid the situation it. The third issue could be addressed quite easily by having a thorough look at the coverage output
Overall, this will increase reliability of sanity checks for new (and old) classes and detect some common pitfalls on sanity checks during initialization.
Currently, we have a test coverage on coveralls of about 77% which is indeed not bad. The remainder of about 23% comes mostly from two problems:
The first issue can be resolved by add failure-checks where we'd execute some code provoking some specific failure and report if exceptions are not raised as we would expect them to be. The second issue however is up to coveralls but we might see if we could so something to avoid the situation it. The third issue could be addressed quite easily by having a thorough look at the coverage output
Overall, this will increase reliability of sanity checks for new (and old) classes and detect some common pitfalls on sanity checks during initialization.