Re: [v3] c++0x <system_error> CD updates

[Paolo Carlini <paolo dot carlini at oracle dot com>]

Hi,
> Updated patch and changelog. Regtested x86_64 linux (with and without -Wall)
>   
Patch is Ok, thanks. Note, we are exporting two symbols, system_category
and generic_category, which depend on the resolution of DR 890, and, in
principle, we will not be able tor "unexport" until the next
ABI-breakage (even if for C++0x features the rules are more lax...
Actually, C++0x exports should be probably in a class apart, we briefly
discussed that with Benjamin already). Thus, let's make sure to review
the issue when the release of gcc4.4.0 approaches.
> One thing I'm not clear on is why error_code's default ctor uses
> system_category while error_condition uses generic_category. I think
> if the integral value maps to an errc, then you use generic_category
> and since 0 isn't, the error_code default ctor uses system_category
> but I have no idea why error_condition's doesn't.
>   
Yes, I was also slightly surprised by that (well the counterpart in the
previous WD). You may want to ask on the library reflector...

Thanks again,
Paolo.