Re: [Boost-users] boost/numeric/ublas/matrix_sparse.hpp does not play well with RICE

On Thu, May 20, 2010 at 4:04 PM, John O. Woods wrote:

I posted earlier today, and I think now that the problem might not be RICE at all. I just tried to compile something that I know should compile properly, which does not use RICE, and it didn't work with a different boost library (in this case ). I'm not really sure how to troubleshoot this. I know I upgraded recently from boost 1.41 to 1.43, but I can't see how that would be the issue. Perhaps I somehow have multiple versions of boost hidden here and there and it's trying to use the headers from one but libraries from another? For example, compiling this file: #include int main() { return 0; }

Gives these errors: $ g++ -I. -I. -I/usr/lib/ruby/1.8/i486-linux -I. -D_FILE_OFFSET_BITS=64  -fPIC -fno-strict-aliasing -g -g -O2  -fPIC    -Wall -g -c main.cppIn file included from /usr/local/include/boost/unordered/detail/equivalent.hpp:11,                  from /usr/local/include/boost/unordered/unordered_map.hpp:19,                  from /usr/local/include/boost/unordered_map.hpp:16,                  from main.cpp:4: /usr/local/include/boost/unordered/detail/extract_key.hpp:31: error: template argument required for ‘struct set_extractor’ /usr/local/include/boost/unordered/detail/extract_key.hpp:77: error: template argument required for ‘struct map_extractor’ In file included from /usr/local/include/boost/unordered/unordered_map.hpp:19,                  from /usr/local/include/boost/unordered_map.hpp:16,                  from main.cpp:4: /usr/local/include/boost/unordered/detail/equivalent.hpp:19: error: wrong number of template arguments (4, should be 1) /usr/local/include/boost/unordered/detail/fwd.hpp:661: error: provided for ‘template<class T> class boost::unordered_detail::hash_equivalent_table’ (continues for a few pages) It does not do this with all boost libraries. I have no problems with lexical_cast. Grateful for any assistance, John

On Thu, May 20, 2010 at 11:32 AM, John O. Woods wrote:

...

Hi there, Cross-posted this to the RICE listserv, but they cannot reproduce. I can reproduce it on karmic and lucid (ubuntu), separate machines, in boost 1.43.0. RICE is a C++ interface for Ruby (not unlike boost python). If I include boost numeric ublas matrix_sparse and either rice/Data_Type.hpp or rice/Constructor.hpp, I get a lot of errors -- even if I do nothing with them. For example, this works: #include int main() {   return 0; }

So does this: #include int main() {   return 0; } But if I put both #includes in, it gives the following errors: $ make g++ -I. -I. -I/usr/lib/ruby/1.8/i486-linux -I. -D_FILE_OFFSET_BITS=64 -I/usr/lib/ruby/gems/1.8/gems/rice-1.3.2/ruby/lib/include -fPIC -fno-strict-aliasing -g -g -O2  -fPIC    -Wall -g -c main.cpp In file included from /usr/local/include/boost/numeric/ublas/storage_sparse.hpp:23,                  from /usr/local/include/boost/numeric/ublas/vector_sparse.hpp:16,                  from /usr/local/include/boost/numeric/ublas/matrix_sparse.hpp:16,                  from main.cpp:5: /usr/local/include/boost/numeric/ublas/storage.hpp:60: error: expected primary-expression before ‘)’ token /usr/local/include/boost/numeric/ublas/storage.hpp:65: error: expected primary-expression before ‘)’ token /usr/local/include/boost/numeric/ublas/storage.hpp:79: error: expected primary-expression before ‘)’ token /usr/local/include/boost/numeric/ublas/storage.hpp: In member function ‘typename ALLOC::size_type boost::numeric::ublas::unbounded_array::max_size() const’: /usr/local/include/boost/numeric/ublas/storage.hpp:174: error: expected primary-expression before ‘)’ token /usr/local/include/boost/numeric/ublas/storage.hpp:174: error: expected ‘;’ before ‘ruby_xmalloc’ /usr/local/include/boost/numeric/ublas/storage.hpp:174: error: expected primary-expression before ‘)’ token /usr/local/include/boost/numeric/ublas/storage.hpp: In member function ‘typename ALLOC::size_type boost::numeric::ublas::bounded_array::max_size() const’: /usr/local/include/boost/numeric/ublas/storage.hpp:357: error: expected primary-expression before ‘)’ token /usr/local/include/boost/numeric/ublas/storage.hpp:357: error: expected ‘;’ before ‘ruby_xmalloc’ /usr/local/include/boost/numeric/ublas/storage.hpp:357: error: expected primary-expression before ‘)’ token In file included from /usr/local/include/boost/numeric/ublas/vector_sparse.hpp:16,                  from /usr/local/include/boost/numeric/ublas/matrix_sparse.hpp:16,                  from main.cpp:5: /usr/local/include/boost/numeric/ublas/storage_sparse.hpp: At global scope: /usr/local/include/boost/numeric/ublas/storage_sparse.hpp:240: error: expected primary-expression before ‘)’ token make: *** [main.o] Error 1

ruby_xmalloc does not seem to exist in that file, so I gather it's having some kind of syntax hiccup that is leaking into later #includes. Here are other relevant details: $ sudo gem list --local |grep rice rice (1.3.2, 1.3.1) I seem to get the same bug with both boost 1.41 and boost 1.43, though sometimes I have trouble telling which one I'm including. Should this be a bug report, or am I missing something stupid? (I'd be eternally grateful if you'd reply to me as well as to the list.) Thanks so much, John

PS: extconf.rb below. Works fine for other projects, as long as I don't use matrix_sparse.

require 'rubygems' require 'mkmf-rice' dir_config("boost") have_library("stdc++") have_library("boost") have_library("boost_filesystem") if RUBY_VERSION =~ /1.9/ then   $CPPFLAGS += " -DRUBY_19" end create_makefile('main')

Try compiling this: #ifdef ValueType # undef ValueType #endif #ifdef key_equal # undef key_equal #endif #include int main() { return 0; } If that works, somehow a macro is getting set that should not be set. If it does not work, that could still be what is happening if the compiler enforces it. That seems to be the common thing amongst what you posted for unordered_map. Regardless, it seems kind of like something is defining some macros. Generate a preprocessed file and see how it looks.