Re: C++ err msgs [ was Re: building libstdc++-v3 from the 20000105 snapshot dies with multiple def errors. ]

[llewelly at dbritsch dot dsl dot xmission dot com]

On Thu, 11 May 2000, Petter Urkedal wrote:

[snip]
> I don't see how this will look like if we have to split nested
> templates.  How about
> 
>     std::basic_string
>       < char,
>         std::char_traits<char>,
>         std::allocator<char>
>       >::basic_string(unsigned int, char, std::allocator<char> const &)
> 
> A more compilcated symbol would then look like
> 
>     std::map
>       < std::basic_string
>           < char,
>             std::char_traits<char>,
>             std::allocator<char>
>           >,
>         std::pair<int, int>
>       >::operator[](std::basic_string
>                       < char,
>                         std::char_traits<char>,
>                         std::allocator<char>
>                       >)

So far I think I like this approach best.

> 
> Of course, only part which do not fit on a line is split.
> 
> In my own code, I have been using a more schemish approach, which
> BTW is closer to your suggestion,
>     
>     std::map
>       < std::basic_string
>           < char,
>             std::char_traits<char>,
>             std::allocator<char> >,
>         std::pair<int, int> >
>       ::operator[](std::basic_string
>                      < char,
>                        std::char_traits<char>,
>                        std::allocator<char> >)

I do not want this one unless/untill emacs matches angle brackets in
  c++-mode.

[snip]

I think the threshold for breaking up names should either be
  user-adjustable (say, via -fbreak-names-over-length=N), or should
  correspond to -fmessage-length. (both?)