Failing to build stan-math unit tests

test.out

bbales2@frog:~/smath$ make test/unit/math/mix/mat/prob/categorical_test
make: Warning: Archive 'test/libgtest.a' seems to have been created in deterministic mode. 'test/gtest.o' will always be updated. Please consider passing the U flag to ar to avoid the problem.
ar rv test/libgtest.a test/gtest.o
r - test/gtest.o
make: Warning: Archive 'test/libgtest.a' seems to have been created in deterministic mode. 'test/gtest.o' will always be updated. Please consider passing the U flag to ar to avoid the problem.
g++ -I . -isystem lib/eigen_3.2.9 -isystem lib/boost_1.62.0 -isystemlib/cvodes_2.9.0/include -Wall -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE -DBOOST_DISABLE_ASSERTS -DNO_FPRINTF_OUTPUT -pipe -fPIC   -c -O3 -DGTEST_USE_OWN_TR1_TUPLE -DGTEST_HAS_PTHREAD=0 -isystem lib/gtest_1.7.0/include -isystem lib/gtest_1.7.0 test/unit/math/mix/mat/prob/categorical_test.cpp -o test/unit/math/mix/mat/prob/categorical_test.o
In file included from ./stan/math/fwd/core.hpp:4:0,
                 from ./stan/math/mix/mat/fun/typedefs.hpp:5,
                 from ./stan/math/mix/mat.hpp:4,
                 from test/unit/math/mix/mat/prob/categorical_test.cpp:1:
./stan/math/fwd/core/fvar.hpp: In substitution of ‘template<class V> stan::math::fvar<T>::fvar(const V&, typename boost::enable_if_c<stan::math::ad_promotable<V, T>::value>::type*) [with V = stan::math::fvar<stan::math::var>]’:
./stan/math/fwd/scal/fun/log.hpp:19:52:   required from ‘stan::math::fvar<T> stan::math::log(const stan::math::fvar<T>&) [with T = stan::math::fvar<stan::math::var>]’
./stan/math/prim/mat/prob/categorical_lpmf.hpp:77:27:   required from ‘typename boost::math::tools::promote_args<T2>::type stan::math::categorical_lpmf(const std::vector<int>&, const Eigen::Matrix<RhsScalar, -1, 1>&) [with bool propto = false; T_prob = stan::math::fvar<stan::math::fvar<stan::math::var> >; typename boost::math::tools::promote_args<T2>::type = stan::math::fvar<stan::math::fvar<stan::math::var> >]’
./stan/math/prim/mat/prob/categorical_log.hpp:50:37:   required from ‘typename boost::math::tools::promote_args<T>::type stan::math::categorical_log(const std::vector<int>&, const Eigen::Matrix<Scalar, -1, 1>&) [with T_prob = stan::math::fvar<stan::math::fvar<stan::math::var> >; typename boost::math::tools::promote_args<T>::type = stan::math::fvar<stan::math::fvar<stan::math::var> >]’
test/unit/math/mix/mat/prob/categorical_test.cpp:73:3:   required from here
./stan/math/fwd/core/fvar.hpp:33:7: error: ambiguous template instantiation for ‘struct stan::math::ad_promotable<stan::math::fvar<stan::math::var>, stan::math::fvar<stan::math::var> >’
       fvar(const V& v,
       ^
In file included from ./stan/math/fwd/scal/meta/ad_promotable.hpp:4:0,
                 from ./stan/math/fwd/core/fvar.hpp:6,
                 from ./stan/math/fwd/core.hpp:4,
                 from ./stan/math/mix/mat/fun/typedefs.hpp:5,
                 from ./stan/math/mix/mat.hpp:4,
                 from test/unit/math/mix/mat/prob/categorical_test.cpp:1:
./stan/math/prim/scal/meta/ad_promotable.hpp:30:12: note: candidates are: template<class T> struct stan::math::ad_promotable<T, T> [with T = stan::math::fvar<stan::math::var>]
     struct ad_promotable<T, T> {
            ^
In file included from ./stan/math/fwd/core/fvar.hpp:6:0,
                 from ./stan/math/fwd/core.hpp:4,
                 from ./stan/math/mix/mat/fun/typedefs.hpp:5,
                 from ./stan/math/mix/mat.hpp:4,
                 from test/unit/math/mix/mat/prob/categorical_test.cpp:1:
./stan/math/fwd/scal/meta/ad_promotable.hpp:18:12: note:                 template<class V, class T> struct stan::math::ad_promotable<V, stan::math::fvar<T> > [with V = stan::math::fvar<stan::math::var>; T = stan::math::var]
     struct ad_promotable<V, fvar<T> > {
            ^
In file included from ./stan/math/rev/core/operator_unary_plus.hpp:7:0,
                 from ./stan/math/rev/core.hpp:34,
                 from ./stan/math/mix/mat/fun/typedefs.hpp:6,
                 from ./stan/math/mix/mat.hpp:4,
                 from test/unit/math/mix/mat/prob/categorical_test.cpp:1:
./stan/math/prim/scal/fun/constants.hpp:65:18: warning: ‘stan::math::NEGATIVE_EPSILON’ defined but not used [-Wunused-variable]
     const double NEGATIVE_EPSILON
                  ^
In file included from ./stan/math/rev/core.hpp:42:0,
                 from ./stan/math/mix/mat/fun/typedefs.hpp:6,
                 from ./stan/math/mix/mat.hpp:4,
                 from test/unit/math/mix/mat/prob/categorical_test.cpp:1:
./stan/math/rev/core/set_zero_all_adjoints.hpp:14:17: warning: ‘void stan::math::set_zero_all_adjoints()’ defined but not used [-Wunused-function]
     static void set_zero_all_adjoints() {
                 ^
make/tests:24: recipe for target 'test/unit/math/mix/mat/prob/categorical_test.o' failed
make: *** [test/unit/math/mix/mat/prob/categorical_test.o] Error 1

There are a couple ways to fix this. I'm surprised it's working on other platforms. The problem is instantiating

Base template class:

    template <typename V, typename T>
    struct ad_promotable {
      enum { value = false };
    };

and we're trying to match ad_promotable<fvar<var>, fvar<var> >, and we have ambiguity between

From: stan_math/stan/math/prim/scal/meta/ad_promotable.hpp:

    template <typename T>
    struct ad_promotable<T, T> {
      enum { value = true };
    };

From: stan_math/stan/math/fwd/scal/meta/ad_promotable.hpp:

    template <typename V, typename T>
    struct ad_promotable<V, fvar<T> > {
      enum { value = ad_promotable<V, T>::value };
    };

So it must be that this compiler is treating the resolution slightly differently. I see two resolutions:

1. We might be able to get away with just removing that first template given how explict we are in the other specializations.

2. We could use Boost's disable_if and condition on testing that V and fvar<T> are the same type (I think there's an is_same metaprogram)

3. We could specialize further for template <typename T> struct ad_promotable<fvar<T>, fvar<T> > in the fvar directory.

If (1) works that's cleanest. I'd actually prefer (3) to (2) because all the disable-if and enable-if stuff is hard to follow.