Re: [C++ PATCH] [PR17447] Check incomplete types of parameters intemplates

[Mark Mitchell <mark at codesourcery dot com>]

Giovanni Bajo wrote:

> Gabriel Dos Reis wrote:
>
>  
>
> > I thought the common practice had been to avoid uses_template_parms,
> > in preference of dependent_xxx_p.
> >    
>
> AFAICT, not after Mark rewrote it before 3.4 to avoid using walk_tree and
> allocating a hash table (that is, painfully slow). dependent_type_p won't work
> here as-is, I would have to put ++processing_template_decl before
> and --processing_template_decl after its use, which is exactly what
> uses_template_parms does.
>  

> It could be argued that dependent_type_p should work also outside templates,
> and then fix all its uses. I would support that, but it's a cleanup for 4.1 at
> this point, and surely this patch can live with uses_template_parms at the
> moment.
>
>  
The fact that dependent_type_p does not check with 
processing_template_parms is zero is an important optimization., so I 
would not like this change.

I think that probably processing_template_decl should be getting set 
earlier.  Normally, it is set right after "template <...>" -- but for 
"preparsed" functions something is not setting this, I take it.  What 
does the call stack look like when we are getting here with a template 
function, but with processing_template_decl not set?

Thanks,

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com