Re: [boost] [filesystem] gcc3.2.3/bcc regression fixes

From: Beman Dawes (bdawes_at_[hidden])
Date: 2006-02-20 16:05:01

"Tobias Schwinger" <tschwinger_at_[hidden]> wrote in message 
news:dt306t$pp5$1_at_sea.gmane.org...
> 1. path.hpp
> ===========
>
> 1.1. The BCC workaround is quite clean. If it doesn't make any sense to 
> (statically) override the operator=(string_type) or the string() member 
> functions of basic_path<...> it would be safe to use it as the default 
> implementation, I figure.
>
> 1.2. If VC7.0 can handle this kind of deduction it would be worth a try to 
> see if bypassing can be avoided (I don't have this compiler installed).
>
> 1.3. There was workaround bypassing the operators for BCC 5.8.1, I don't 
> think it's a good idea because it's likely my workaround (tested with 
> 5.6.4, without bypassing) works for the newer version, too. Someone should 
> adjust TESTED_AT, after it's done.
>
> 2. convenience.hpp
> ==================
>
> The workaround was originally posted by Alisdair Meredith but seems it 
> hasn't been applied, yet (so I added it for completeness and kept 
> TESTED_AT with the version number of the new compiler -- I couldn't 
> double-check because I don't have this version).
>
> 3. fstream.hpp
> ==============
>
> The tweak makes things compile with some versions of the GCC parser, that 
> mistake the open bracket of the explicit function template argument list 
> for an operator (which seems to be the case with that gcc-3.2.3-linux 
> regression).
>
>
> OK to commit?

Done. Thanks!

--Beman