Maurizio Vitale wrote:
Another data point: even though the above code compile with 8 and 7, in my real code I get a match failure on number<_,_,_,_,_,_,_,_> [8] or number<_,_,_,_,_,_,_> [7]. As far as I can tell, the only difference between the working code and the non-working code is the additional template arguments to number<>, which are not used.
This is only a problem on GCC because of a compiler bug wrt template template parameters. Proto reuses a work-around from the MPL, which is conditioned on an MPL macro called BOOST_MPL_LIMIT_METAFUNCTION_ARITY. If you define this to be at least as large as BOOST_PROTO_MAX_ARITY, the problem goes away. I've added some code to check for this condition and fail loudly at compile time. Also, in another mail, you say:
I'm really interested in having patterns with those many arguments, so once the variable controlling the maximum arity and the one controlling the maximum number of subpatterns in a pattern get decoupled I may be fine, but still giving a clear error message would be helpful.
Depending on what "number of subpatterns in a pattern" means, you may already have what you want. There is a separate macro called BOOST_PROTO_MAX_LOGICAL_ARITY which controls how many parameters proto::or_ and proto::and_ accept. It is independent of BOOST_PROTO_MAX_ARITY. For the time being, whether a pattern such as number<_,_,_,_,_,_,_,_> is accepted is controlled by BOOST_PROTO_MAX_ARITY. -- Eric Niebler Boost Consulting www.boost-consulting.com