These lcov reports look great (in terms of how they're presented,
not how Stan fared!).
I'm not surprised at the coverage. We should definitely
try for 100% coverage. As always, it's a matter of priorities.
I'd be happy to prioritize more testing and refactoring of
the code that we have. I'm going to start on gm/ast today,
which has very few tests currently.
I believe the next agrad/fwd pull request is blocked on
binomial_coefficient_log, where it's failing in the code branch
for inputs > 1000-ish.
Can we somehow restrict reports to src/stan/** suffixes?
Or remove anything with lib/**? I'm finding it hard to see
our files among all the Boost and Eigen files, which we don't
try to test.
Any idea how lcov decides what is covered? Some of the
things that are listed as not covered, like agrad/rev/op,
are base classes that are used all over the place in tests.
We should make mock objects that extend them and test those.
So the question is how we would make sure that counts as
coverage.
- Bob
> --
> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to
stan-dev+u...@googlegroups.com.
> For more options, visit
https://groups.google.com/groups/opt_out.
>
>