[boost] Re: Recent changes to intel compiler config

From: David Abrahams (dave_at_[hidden])
Date: 2003-11-13 09:41:23

"John Maddock" <john_at_[hidden]> writes:

>> I believe recent changes to the intel compiler config are wrong, or
>> at the very least incompatible with what Boost.Build is doing:
>
> That hasn't changed in the last 3 months or so has it?

I was away from a net connection when I wrote that email, so let me
see...

OK, no it hasn't.

Well, it just started reporting errors for me because the PP symbol is
getting redefined.  Maybe I haven't tested with that configuration in
some time.

>> The above unconditionally decides that intel5 and intel6 don't support
>> intrinsic wchar_t.  I believe that to be in error.  The Boost.Build
>> logic is:
>
> [snipped]
>
>> Which says that if Intel C++ is being used on windows over msvc6 and
>> its standard library, there is no wchar_t support.  Otherwise,
>> versions of intel C++ > 5 over vc7 or vc7.1 get explicit and automatic
>> support for wchar_t.
>
> Only if the right command line option is used - I guess Boost.build does
> that, but what about users using Intel from their IDE?

I guess that's their problem, since AFAICT we're unable to properly
detect all the conditions which say whether wchar_t is supported.
I'm sympathetic to users who want to do that, but we have one build
system which we're claiming to support.  It seems to me that we
should support it, no?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com