Re: [boost] Re: metrowerks codewarrior regression failures appeared

From: Howard Hinnant (hinnant_at_[hidden])
Date: 2002-12-20 17:04:45

On Friday, December 20, 2002, at 04:40  PM, Samuel Krempp wrote:

> ok. So, you mean the codewarrior compiler that was previously used for 
> the
> regression must have been linked statically to the C++ lib, while the 
> new
> one is linked dynamically, and thus locale-disabled ?

That was my working theory, but Beman says I'm wrong.

> I come to think, wouldn't it be better if the no-locale situation was
> implemented by having a fake 'locale' object and empty related 
> functions ?
> (allows writing locale-enabled code, and the compiler just ignores it)

We're working on a better solution for Pro 9.

> in anycase, the lines causing the error here were the only ones I had 
> left
> out of  #ifndef BOOST_NO_STD_LOCALE  blocks, so adding such #ifndef 
> should
> fix the failure, with or without locales enabled in the codewarrior 
> being
> used. I'll try that..

That sounds like a reasonable thing to do.  But it also seems like we 
have a minor mystery here.

In metrowerks.hpp I see:

#   ifdef _MSL_NO_LOCALE
#     define BOOST_NO_STD_LOCALE
#   endif

which seems right.  But if this is statically linking, (and thus not 
including <UseDLLPrefix.h.>), I see no reason BOOST_NO_STD_LOCALE 
should be getting turned on.

-Howard