Re: can't specialize a template

From:

Ulrich Eckhardt <eckhardt@satorlaser.com>

Newsgroups:

microsoft.public.vc.language

Date:

Thu, 01 Jun 2006 16:28:28 +0200

Message-ID:

<i5a3l3-ceh.ln1@satorlaser.homedns.org>

Arkadiy wrote:
[...]

In my case using unnamed namespace reduces number of contexts where ODR
can arise (If you are interested in more details I can provide them or
direct you to the thread on the boost list where this issue was
discussed in relation to typeof library).

Hehe, in fact when I read this it reminded me of a discussion that took
place there, so there is probably not much I could add to it. ;)

Every translation unit that includes an anonymous namespace gets its
own 'instance' thereof. IIRC the problem with precompiled headers is that
they are generated as part of the associated translation unit, but when
included in other translation units anonymous namespaces are not handled
correctly, i.e. to the compiler it seems like two distinct 'instances' of
the anonymous namespace which are intended to not conflict, but in your
case they are impossible to make interoperable.

Yes, that's what I thought... However, I don't think this is right.
Shouldn't the use of PCH be transparent?

IMO yes.

The fact that they are generated as a part of another TU is an
implementation detail that should not prevent legal C++ code from
being compiled. And the fact that it (partially) works in VC71
proves that it can be done.

I totally agree.

So, if this cannot be done, do you think it should be considered a bug
in VC80?

Yes.

Uli