hugin-app: failure to compile when including boost headers

[mojca (Mojca Miklavec)]

hugin-app (both 2012.0.0 and the latest version from RCS at the time of writing) fails to compile with boost @1.54.0, at least on Lion. The list of errors is long, but here's the first one, most likely related to a weird mixture of Apple and boost headers conflicting with each other:

In file included from /work/hugin-2012.0.0/src/hugin_base/huginapp/ImageCache.h:30:
In file included from /opt/local/include/boost/signal.hpp:27:
In file included from /opt/local/include/boost/signals/signal0.hpp:24:
In file included from /opt/local/include/boost/signals/signal_template.hpp:21:
In file included from /opt/local/include/boost/signals/slot.hpp:15:
In file included from /opt/local/include/boost/signals/trackable.hpp:13:
In file included from /opt/local/include/boost/type_traits.hpp:35:
In file included from /opt/local/include/boost/type_traits/has_operator.hpp:12:
In file included from /opt/local/include/boost/type_traits/has_bit_and.hpp:43:
/opt/local/include/boost/type_traits/detail/has_binary_operator.hpp:157:121: error: expected expression
   BOOST_STATIC_CONSTANT(bool, value = (sizeof(check(((make<Lhs>() BOOST_TT_TRAIT_OP make<Rhs>()),make<has_operator>())))==sizeof(::boost::type_traits::yes_type)));
                                                                                                                        ^
/opt/local/include/boost/config/suffix.hpp:393:72: note: expanded from macro 'BOOST_STATIC_CONSTANT'
#     define BOOST_STATIC_CONSTANT(type, assignment) static const type assignment
                                                                       ^

[mojca (Mojca Miklavec)]

build log from hugin-app 2012 against boost 1.54.0

[macports@…]

Cc Me!

[michaelld (Michael Dickens)]

I'm trying to track down a similar bug. Have you looked at ​this Boost ticket? Seems like if Apple headers are included before Boost, then issues can arise because of the "#define check" in AssertMacros.h; "check" is then used by Boost's "operator_exists" in has_binary_operator.hpp (and, maybe, elsewhere). Solution is to either include boost first, or to

#define __ASSERT_MACROS_DEFINE_VERSIONS_WITHOUT_UNDERSCORES 0

or

-D__ASSERT_MACROS_DEFINE_VERSIONS_WITHOUT_UNDERSCORES=0

for compiling. I wish this was my issue, but it's not; maybe it will work for this ticket?

[mojca (Mojca Miklavec)]

Interesting. I had exactly that line, but I was trying to set it to one. I upgraded boost to 1.55.0b1, set this value to zero and stumbled across an weird issue with libtiff:

In file included from
/path/to/hugin-app/work/hugin-2013.0.0/src/hugin1/stitch_project/hugin_stitch_project.cpp:46:

In file included from /opt/local/include/tiffio.h:33:
/opt/local/include/tiff.h:78:23: error: typedef redefinition with different
types ('unsigned long' vs 'uint64_t' (aka 'unsigned long long'))
typedef TIFF_UINT64_T uint64;
                      ^
//System/Library/Frameworks/Security.framework/Headers/cssmconfig.h:53:18:
note: previous definition is here
typedef uint64_t uint64;
                 ^
2 warnings and 1 error generated.

but after I manually fixed that one, Hugin.app compiled fine and it seems to work.

Now, I'm not exactly sure which one of the two changes (upgrading boost or setting the value to 0) made the compilation possible, but it worked for me.

[mojca (Mojca Miklavec)]

I committed the changes in r113119 and filed a new ticket for tiff in #41280.

[ryandesign (Ryan Carsten Schmidt)]

I'm not seeing this problem with hugin-app @2013.0.0 and boost @1.55.0.

[mojca (Mojca Miklavec)]

True, this particular problem must be gone, most probably -D__ASSERT_MACROS_DEFINE_VERSIONS_WITHOUT_UNDERSCORES=0 in r113119 fixed it, I just didn't close the ticket because of the problem with tiff back than and because I couldn't reliably confirm a successful compilation (a successful build on a buildbot would count for example ;).

Thanks for closing it.