Re: [boost] Argument-dependent lookup for Intel/Win32

From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-06-02 20:14:13

At 10:01 AM 6/2/2003, David Abrahams wrote:

 >I just checked in changes to the intel-win32-tools.jam file which
 >enable argument-dependent lookup for all versions, since we were were
 >incorrectly operating as though it was enabled for v7.1 anyway.  I
 >think we are still out-of-synch with the config, though.  The config
 >for the intel compiler contains:
 >
 >#  if defined(_MSC_VER) && (_MSC_VER <= 1300) // added check for <= VC 7
 >(Peter Dimov)
 >
 >      // Intel C++ 5.0.1 uses EDG 2.45, but fails to activate Koenig 
lookup
 >      // in the frontend even in "strict" mode, unless you use
 >      // -Qoption,cpp,--arg_dep_lookup.  (reported by Kirk Klobe & Thomas 

 >Witt)
 >      // Similarly, -Qoption,cpp,--new_for_init enables new-style "for"
 >loop
 >      // variable scoping. (reported by Thomas Witt)
 >      // Intel C++ 6.0 (currently in Beta test) doesn't have any 
front-end
 >      // changes at all.  (reported by Kirk Klobe)
 >      // That can't be right, since it supports template template
 >      // arguments (reported by Dave Abrahams)
 >#     ifndef BOOST_NO_ARGUMENT_DEPENDENT_LOOKUP
 >#        define BOOST_NO_ARGUMENT_DEPENDENT_LOOKUP
 >#     endif
 >#     define BOOST_NO_SWPRINTF
 >#  endif
 >
 >I have two questions about this:
 >
 >1. Is there a reason we're checking _MSC_VER here?
 >
 >2. Can we just eliminate the #define for
 >   BOOST_NO_ARGUMENT_DEPENDENT_LOOKUP?

I came to the same conclusion before reading this post and have already 
committed the change. Also changed the comment to better describe the 
situation as regards BOOST_NO_ARGUMENT_DEPENDENT_LOOKUP

 >I ran the config tests for intel 5,6,7 and got the following
 >interesting results, but I am at a loss to interpret them.  Do the
 >link errors with Intel7 mean that we failed to get the whchar_t
 >support settings right?

I just did an intel-win32 clean and then limits_test. It compiled, linked, 
and ran without error.

 >What does it mean that intel5,6 still report
 >BOOST_NO_ARGUMENT_DEPENDENT_LOOKUP?  I don't know what to do with
 >this information.

Because # define BOOST_NO_ARGUMENT_DEPENDENT_LOOKUP is (or was, until I 
removed it) hardwired into intel.hpp when used with msvc?

Be careful about making changes at this point. AFAICS, the 
intel-win32-tools.jam and intel.hpp stuff is working far better than it 
ever has.

--Beman