On Tue, Feb 10, 2026 at 05:32:22PM +0000, Boris Staletic wrote:
> On Tuesday, February 10th, 2026 at 4:49 PM, Marek Polacek 
> <[email protected]> wrote:
> 
> > On Tue, Feb 10, 2026 at 03:26:37PM +0000, Boris Staletic wrote:
> > > On Tuesday, February 10th, 2026 at 2:05 PM, Jason Merrill 
> > > <[email protected]> wrote:
> > >
> > > >
> > > >
> > >
> > > Snipped older discussions for brevity.
> > >
> > > > >
> > > > > Here's a v2 of the patch, bootstrapped and tested on 
> > > > > x86_64-pc-linux-gnu.
> > > > > -- >8 --
> > > > > In case of expressions like `&[:expr:]` where `expr` depends on a
> > > > > template parameter, and the splice expression represents a 
> > > > > `FIELD_DECL` or
> > > > > a non-static member `FUNCTION_DECL`, that's exactly what we'd pass on.
> > > > > However, `build_x_unary_op()` for these expressions is expecting an
> > > > > `OFFSET_REF`. `OFFSET_REF` is also what gets passed to
> > > > > `build_x_unary_op()` when templates are not involved.
> > > > >
> > > > > There's also a difference between the template argument being a type 
> > > > > and
> > > > > using `members_of()` to get to the reflections of members (in which 
> > > > > case
> > > > > evaluating the `SPLICE_EXPR` returns a `FUNCTION_DECL` - `splice10.C`
> > > > > test) and passing `^^T::member` as the template argument (in which 
> > > > > case
> > > > > evaluating the `SPLICE_EXPR` returns a `BASELINK` - `splice11.C`).
> > > > >
> > > > > Signed-off-by: Boris Staletic [email protected]
> > > > >
> > > > > PR c++/123660
> > > > > PR c++/123661
> > > > >
> > > > > gcc/cp/ChangeLog:
> > > > >
> > > > > * pt.cc (tsubst_splice_expr): Handle pointers to non-static members
> > > > > from splice expressions
> > > > >
> > > > > gcc/testsuite/ChangeLog:
> > > > >
> > > > > * g++.dg/reflect/splice10.C: New test.
> > > > > * g++.dg/reflect/splice11.C: New test.
> > > > > ---
> > > > > gcc/cp/pt.cc | 14 ++++++++++++
> > > > > gcc/testsuite/g++.dg/reflect/splice10.C | 28 ++++++++++++++++++++++++
> > > > > gcc/testsuite/g++.dg/reflect/splice11.C | 29 +++++++++++++++++++++++++
> > > > > 3 files changed, 71 insertions(+)
> > > > > create mode 100644 gcc/testsuite/g++.dg/reflect/splice10.C
> > > > > create mode 100644 gcc/testsuite/g++.dg/reflect/splice11.C
> > > > >
> > > > > diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
> > > > > index 049bbf07e0..0c0076c9a4 100644
> > > > > --- a/gcc/cp/pt.cc
> > > > > +++ b/gcc/cp/pt.cc
> > > > > @@ -16755,6 +16755,20 @@ tsubst_splice_expr (tree t, tree args, 
> > > > > tsubst_flags_t complain, tree in_decl)
> > > > > op = splice (op);
> > > > > if (op == error_mark_node)
> > > > > return error_mark_node;
> > > > > +
> > > > > + if (SPLICE_EXPR_ADDRESS_P (t) && !TREE_STATIC (TREE_TYPE (op)))
> > > >
> > > >
> > > > TREE_STATIC on a type isn't meaningful, the checks below are enough.
> > >
> > > I thought that was fishy... You're right, everything works without that 
> > > check.
> > >
> > > >
> > > > > + {
> > > > > + if (TREE_CODE (op) == BASELINK
> > > > > + && TREE_CODE (TREE_TYPE (op)) == METHOD_TYPE)
> > > >
> > > >
> > > > It shouldn't be necessary to check for METHOD_TYPE; any class member is
> > > > suitable for passing to build_offset_ref. If the argument is static,
> > > > it'll return it unchanged rather than actually build an OFFSET_REF.
> > >
> > > If I drop the METHOD_TYPE check, I get an ICE in both splice10.C and 
> > > splice11.C.
> > > splice10.C ICEs when calling `build_offset_ref()` on a static member 
> > > function (s::get_y).
> > > splice11.C ICEs when calling `build_offset_ref()` on a free function 
> > > (::f).
> > >
> > > That's why I was initially looking for a way to do this only for 
> > > non-static members.
> > 
> > I think Jason meant the check for METHOD_TYPE for a BASELINK, which
> > is certainly not needed.  Please also use BASELINK_P instead of
> > TREE_CODE == BASELINK.
> > 
> > We shouldn't call build_offset_ref on static member functions or
> > free functions.  So I think the second call to build_offset_ref
> > should be guarded by
> > 
> >   TREE_CODE (op) == FIELD_DECL || DECL_OBJECT_MEMBER_FUNCTION_P (op)
> > 
> > But maybe Jason prefers something else.
> > 
> > 
> > I would also move the SPLICE_EXPR_ADDRESS_P block after checking
> > dependent_splice_p and after check_splice_expr.
> 
> Okay, all of that works. One question:
> Would it be more correct to guard the whole block with
> 
> if (SPLICE_EXPR_ADDRESS_P (t))
> 
> or
> 
> if (SPLICE_EXPR_ADDRESS_P (op))
> 
> Given what's in the dependent_splice_p block, I'm thinking latter, but I'm 
> not sure.

It should check t, because op won't be a SPLICE_EXPR (except in the
dependent_splice_p block).

Marek

Reply via email to