15875 – [3.4/4.0 Regression] rejects pointer to member in template

Description Andrew Pinski 2004-06-08 14:16:14 UTC

struct foo
  {
  int f1(int) const;
  };
template<typename T>  
  void func()
    {
    typedef int (foo::*fptr)(int) const;
    fptr ptr = &foo::f1;
    }

Comment 1 Andrew Pinski 2004-06-08 14:17:25 UTC

Mark I think this might have been caused by one of your recent patches.

Comment 2 Wolfgang Bangerth 2004-06-08 15:36:29 UTC

This was introduced before 2004-06-02, since that is my last build. I 
think it is significant enough that we have to fix it before 3.4.1. 
 
W. 

Comment 3 Mark Mitchell 2004-06-09 18:20:27 UTC

No, this is not becuase of my patch -- it is because of the patch for PR 13092.

Jason, would you please revert that patch, since it is causing this problem? 
Then, please close this PR and reopen that one?

Thanks!

Comment 4 Giovanni Bajo 2004-06-10 10:49:50 UTC

Isn't it worth a little investigation on this bug before? We're trading a 
regression with another regression, and the patch in PR 13092 looks like a step 
forward anyway.

Comment 5 Mark Mitchell 2004-06-10 13:40:16 UTC

Subject: Re:  [3.4/3.5 Regression] rejects pointer to member
 in template

giovannibajo at libero dot it wrote:

> ------- Additional Comments From giovannibajo at libero dot it  2004-06-10 10:49 -------
> Isn't it worth a little investigation on this bug before? We're trading a 
> regression with another regression, and the patch in PR 13092 looks like a step 
> forward anyway.

If someone can easily fix the 13092 patch, that's fine, of course.

However, if we have to choose, it will be better to have regressions in 
3.4.1 that are already in 3.4.0 that new regressions.  A primary goal of 
the point releases is stability.

Comment 8 Andrew Pinski 2004-06-10 18:55:16 UTC

Fixed.

Comment 9 Andrew Pinski 2004-06-11 21:18:11 UTC

*** Bug 15949 has been marked as a duplicate of this bug. ***