Re: Templated method call problem

[jon hanson]

Bernd Strieder <strieder@informatik.uni-kl.de> wrote in message 
news:<ciscm7$38d$1@news.uni-kl.de>...
> Hello,
> 
> the question and the answer belong into gnu.g++.help.
> 
> 
> jon hanson wrote:
> 
> >     template<typename X>
> >     void test2 (const X&)
> >     {
> >         // This gives the error.
> >         std::cout << "r=" <<  TagNames<X>::name<1> ();
> 
> This should read
> 
>          std::cout << "r=" <<  TagNames<X>::template name<1> ();
> 
> Due to two-phase lookup the compiler must interpret the code above as
> some comparisions between some member name and the number 1 compared to
> something behind, which stops to make any sense when reaching the empty
> braces.
> 
> g++ 3.4 produces different error messages, which make equally littly
> clear, why the compiler cannot understand this. See gnu.gcc.org for the
> changes in g++ 3.4, there are some clear examples, when to insert those
> famous "template" and "typename" keywords. As soon as you are in a
> template, you have to take special care, i.e. you have to tell the
> compiler whether something is a dependent type or a template, which is
> only known at instantiation time (second phase of name lookup). OTW it
> will be interpreted as a member, like here, or something completely
> different at parsing time (first phase of lookup).
> 
> 
> Bernd Strieder

Bernd & the others - thanks for the help. I should have spotted that
however the cryptic error message foxed me.

Sorry if the example itself was a bit cryptic but it was a cutdown
version of the actual code which exhibited the problem and i didn't
want to post it in it's entirety.

I thought the code was otherwise standard conformant however i could
be wrong.  The code in question has to compile on g++/linux,
Forte/Solaris, & VC/WinXp and since they're all at different levels of
lag behind the current standard writing conformant code is tricky at
the best of times.

jon