Re: AmigaOS port of Boost 1.27.0

From: smsolie (ssolie_at_[hidden])
Date: 2002-03-25 11:00:02

--- In Boost-Users_at_y..., Darin Adler <darin_at_b...> wrote:
> The BOOST_DISABLE_THREADS define shouldn't be set for a platform.
> If the platform doesn't have threads, we just don't define 
> BOOST_HAS_THREADS for that platform. So you shouldn't need this
> line. If you do need it, can you explain why?

I defined BOOST_DISABLE_THREADS because AmigaOS does not support
multithreading and I assumed the opposite (threads enabled) was the
default.

> > #define BOOST_NO_CWCHAR
> > #define BOOST_NO_CWCTYPE
> > #define BOOST_NO_SWPRINTF
> 
> You only have to define BOOST_NO_CWCHAR.

I'll try removing BOOST_NO_CWCTYPE and BOOST_NO_SWPRINTF.

I went through the "Boost Macro Reference" and picked off all the
definitions I thought applied.  If there are dependencies between
the definitions, I didn't see them.

> > #define BOOST_NO_INTRINSIC_WCHAR_T
> 
> This is usually a property of the compiler, not the platform, and
> so it's the responsibility of the boost/config/compiler/xxx.hpp
> file to set this. But I see that you're using GCC 2.95.3. I think
> that GCC 2.95.3 does have an intrinsic wchar_t. What's going on
> here.

I hit this problem in boost/integer_traits.hpp line 109.
The message "No WCHAR_MIN and WCHAR_MAX present, please adjust
integer_traits<> for your compiler." appeared when I tried to
compile the graph library regression test suite.

Either GCC 2.95.3 doesn't support intrinsic wchar_t or the AmigaOS
port of GCC 2.95.3 doesn't.  I suspect the AmigaOS port has been
modified to remove the intrinsic wchar_t support.  I'll ask on
the Amiga GCC mailing list and see if anyone knows for sure.

> > 3) libs/graph/test/isomorphism.cpp
> >
> > - inline bool verify_isomorphism(const Graph1& g1, const Graph2& 
g2,
> 
> I'm not sure why you removed this function template. It's used in 
> test_isomorphism below. Could you elaborate?

That template is already defined in boost/graph/isomorphism.hpp
starting on line 394.  Looks like a possible cut & paste error to
me.

> > + #include <cassert>
> 
> Since <cassert> and <assert.h> are identical and some platforms 
lack 
> <cassert> we might want to use <assert.h> here even though it's a 
> deprecated feature.

Here's a list of files that include <assert.h>:
./assert.hpp
./compatibility/cpp_c_headers/cassert
./graph/graph_utility.hpp

Here's a list of files that include <cassert>:
./graph/adjacency_matrix.hpp
./graph/detail/bitset.hpp
./graph/minimum_degree_ordering.hpp
./graph/properties.hpp
./graph/push_relabel_max_flow.hpp
./graph/subgraph.hpp
./min_rand.hpp
./pending/bucket_sorter.hpp
./random/bernoulli_distribution.hpp
./random/detail/const_mod.hpp
./random/exponential_distribution.hpp
./random/geometric_distribution.hpp
./random/inversive_congruential.hpp
./random/linear_congruential.hpp
./random/lognormal_distribution.hpp
./random/normal_distribution.hpp
./random/shuffle_output.hpp
./random/triangle_distribution.hpp
./random/uniform_int.hpp
./random/uniform_real.hpp
./random/uniform_smallint.hpp
./regex/detail/fileiter.hpp
./tokenizer.hpp
./token_functions.hpp

It seems that the majority is using <cassert> so I think it is
safe to assume it will work for everyone.

Note: My lists may not be exhaustive because I just did a simple grep.