Re: [Boost-build] Intel's /Qvc option

From: Sean Huang (huangsean_at_[hidden])
Date: 2006-09-12 12:07:45

----- Original Message ----- 
From: "Vladimir Prus" <ghost_at_[hidden]>
To: "Boost.Build list" <boost-build_at_[hidden]>
Sent: Tuesday, September 12, 2006 3:41 AM
Subject: [Boost-build] Intel's /Qvc option

>
> Hi,
> can somebody tell me what exactly the /QvcNN option to intel compier does?
> Docs have this:
>   Description
>   This option specifies compatibility with Visual C++ or Visual Studio.
>
>   Default
>   OFF  No compatibility with Microsoft Visual C++ or Visual Studio is
>      enabled.
>   However, on IA-32 systems, if Visual C++ 6.0, Visual Studio .NET 2002, 
> or
>   Visual Studio .NET 2003 is installed, compatibility is enabled by 
> default.
>
> This leaves me with several questions.
>
> 1. Is intel capable to working without using vc as backend compiler?
It is just a compiler and needs MSVC's runtime libraries. So I think the 
answer is no.

> 2. If no, should /Qvc option match the version of vc backend?
Yes. You'll most likely get link errors otherwise.

> 3. Does /QvNN affects the language that is accepted by the compiler?
I think so. If I remember correctly, _MSC_VER will be affected too.

> 4. If all the code will be compiled with intel, does this option matter?
I think so because it needs to link with vc runtime libraries.

> 5. Does intel link with it's own runtime, or with vc's runtime library?
Intel ships with some of the runtime but it needs vc's runtime libraries.

> 6. Finally, why one would care about this option at all, and want it to 
> have a
> specific value.
When the compiler is installed, it asks for VC compatibility and writes it 
to a configuration file (icl.cfg). iclvars.bat is also updated to call the 
corresponding vcvars32.bat. Because of this, we do not use iclvars.bat 
directly  since we have mutiliple versions of MSVC installed and need to 
switch on different VC versions.

>
> The reason I'm asking is that intel-win currently has some code to set the
> value, and while I though I understand why it's needed, I'm no longer 
> sure.
>
I think it is needed but the current toolset does not do a complete job yet.

Thanks,
Sean