On Wed, 19 Nov 2025, Jason Merrill wrote:

> On 11/18/25 3:27 AM, Patrick Palka wrote:
> > Bootstrapped and regtested on x86_64-pc-linux-gnu, does this
> > look OK for trunk/15?
> 
> OK.
> > -- >8 --
> > 
> > Since we don't implement deferred noexcept-spec parsing of a friend
> > declaration, the r15-2117 change to diagnose name lookup failure for
> > the current instantiation ahead of time needs to compensate by
> > suppressing the name lookup failure in this case.
> > 
> >     PR c++/122668
> >     PR c++/114764
> > 
> > gcc/cp/ChangeLog:
> > 
> >     * pt.cc (dependentish_scope_p): Return true for the current
> >     instantiation from within an immediately parsed noexcept-spec.
> > 
> > gcc/testsuite/ChangeLog:
> > 
> >     * g++.dg/cpp0x/noexcept91.C: New test.
> > ---
> >   gcc/cp/pt.cc                            | 10 +++++++++-
> >   gcc/testsuite/g++.dg/cpp0x/noexcept91.C | 15 +++++++++++++++
> >   2 files changed, 24 insertions(+), 1 deletion(-)
> >   create mode 100644 gcc/testsuite/g++.dg/cpp0x/noexcept91.C
> > 
> > diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
> > index b7cb807d8daa..f132f8b39d24 100644
> > --- a/gcc/cp/pt.cc
> > +++ b/gcc/cp/pt.cc
> > @@ -29057,7 +29057,15 @@ dependent_scope_p (tree scope)
> >   bool
> >   dependentish_scope_p (tree scope)
> >   {
> > -  return dependent_scope_p (scope) || any_dependent_bases_p (scope);
> > +  return dependent_scope_p (scope) || any_dependent_bases_p (scope)
> > +    /* A noexcept-spec is a complete-class context, so this should never
> > hold.
> > +       But since we don't implement deferred noexcept-spec parsing of a
> > friend
> > +       declaration (PR114764) we compensate by treating the scope as
> > dependent
> > +       in this case to avoid bogus name lookup failures (PR122668).  */
> > +    || (cp_noexcept_operand

Oops, cp_noexcept_operand detects noexcept-expressions not
noexcept-specifiers so this workaround is insufficient.

I think we should just bite the bullet and defer parsing of these
friends' noexcept-specs.. it should be easy enough to do so at class
template scope at least since we don't need to match the friends
until instantiation time anyway.

> > +   && CLASS_TYPE_P (scope)
> > +   && TYPE_BEING_DEFINED (scope)
> > +   && dependent_type_p (scope));
> >   }
> >     /* T is a SCOPE_REF.  Return whether it represents a non-static member
> > of
> > diff --git a/gcc/testsuite/g++.dg/cpp0x/noexcept91.C
> > b/gcc/testsuite/g++.dg/cpp0x/noexcept91.C
> > new file mode 100644
> > index 000000000000..590dfa6162a3
> > --- /dev/null
> > +++ b/gcc/testsuite/g++.dg/cpp0x/noexcept91.C
> > @@ -0,0 +1,15 @@
> > +// PR c++/122668
> > +// { dg-do compile { target c++11 } }
> > +
> > +template<class T>
> > +struct A {
> > +  friend void f1(A& a) noexcept(noexcept(a.g(0))) { }
> > +  friend void f2(A& a) noexcept(noexcept(A::g(0))) { }
> > +  static void g(int) noexcept;
> > +};
> > +
> > +int main() {
> > +  A<int> a;
> > +  static_assert(noexcept(f1(a)), "");
> > +  static_assert(noexcept(f2(a)), "");
> > +}
> 
> 

Reply via email to