Re: [C++ PATCH] [PR14028] Parser accepts invalid unbalanced bracket(regression 3.4/3.5)

[Mark Mitchell <mark at codesourcery dot com>]

Jason Merrill wrote:

> On Fri, 6 Feb 2004 14:44:33 +0100, "Giovanni Bajo" <giovannibajo@libero.it> wrote:
>
>  
>
> > the parser accepts invalid code such as:
> > ----------------------------------------------------------
> > template <int> struct A {};
> > template <typename TP> class B : public A<4  {};  // missing bracket!
> > ----------------------------------------------------------
> >
> > This happens because the check is done by cp_parser_require, which does not
> > emit an error during tentative parsing. The argument list is not invalidated by
> > this, and a CPP_TEMPLATE_ID is created, so that subsequent (non-tentative)
> > parsings fail to diagnose the problem. Fixed by forcing a hard error.
> >    
>
> I'm not very familiar with this code, but it seems to me that giving a hard
> error during tentative parsing isn't right.
That was my original concern too.

Basically, we must have conceptually committed by this point if issuing 
the error message make sense.

The problem is that we don't have nested parser committment; either 
we're fully committed, or we're not.  So, we might be sure that we're 
looking a template argument list, but not sure what larger construct 
contains that argument list.

That's implicit in the way that cp_parser_template_id compacts a stream 
of tokens into a CPP_TEMPLATE_ID token.  So, Giovanni's patch should be 
safe, even though it's non-intuitive in this way.

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