On Thu, May 14, 2026 at 10:40:25PM -0400, Jason Merrill wrote:
> On 5/14/26 12:50 PM, Marek Polacek wrote:
> > Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk/16.2?
> >
> > -- >8 --
> > Thanks to DR 696 (r253266), this works:
> >
> > int g;
> > void fn ()
> > {
> > int &c = g;
> > auto l = [] { c++; };
> > l();
> > }
> >
> > because `c` in the lambda body is not an odr-use because we can
> > evaluate it to a constant and so there's no capture. But when
> > fn is a template, we reject the code and crash. This patch fixes
> > both.
> >
> > Outside a template, the call to maybe_constant_value in mark_use
> > evaluates `c` to `(int&) &g` but in a template, it remains `c`.
> > Then we emit an error, and crash on the error_mark_node from
> > process_outer_var_ref. One of the reasons is
> > else if (TYPE_REF_P (TREE_TYPE (expression)))
> > /* FIXME cp_finish_decl doesn't fold reference initializers. */
> > return true;
> > in value_dependent_expression_p but even if that changed, we still
> > wouldn't get the referent because decl_really_constant_value wouldn't
> > give it to us; the DECL_INITIAL is not a TREE_CONSTANT yet.
> >
> > So I stopped trying to make this work in a template, and instead
> > I'm delaying the processing to instantiating when we know that
> > maybe_constant_value wouldn't even try to evaluate.
>
> But it sounds like maybe_constant_value is fine?
That's right, the call to maybe_constant_value doesn't cause
any problems, it's just that it wouldn't evaluate anything.
> > The is_capture_proxy change is due to lambda-scope10.C: we are
> > checking is_nondependent_constant_expression before the enclosing
> > function has been built up.
> >
> > PR c++/123536
> >
> > gcc/cp/ChangeLog:
> >
> > * expr.cc (mark_use) <case VAR_DECL>: Check
> > is_nondependent_constant_expression before
> > process_outer_var_ref.
> > * lambda.cc (is_capture_proxy): Check DECL_CONTEXT.
> >
> > gcc/testsuite/ChangeLog:
> >
> > * g++.dg/cpp0x/lambda/lambda-const12.C: New test.
> > ---
> > gcc/cp/expr.cc | 3 +-
> > gcc/cp/lambda.cc | 1 +
> > .../g++.dg/cpp0x/lambda/lambda-const12.C | 48 +++++++++++++++++++
> > 3 files changed, 51 insertions(+), 1 deletion(-)
> > create mode 100644 gcc/testsuite/g++.dg/cpp0x/lambda/lambda-const12.C
> >
> > diff --git a/gcc/cp/expr.cc b/gcc/cp/expr.cc
> > index 4d017d530ef..15036cf7cad 100644
> > --- a/gcc/cp/expr.cc
> > +++ b/gcc/cp/expr.cc
> > @@ -131,7 +131,8 @@ mark_use (tree expr, bool rvalue_p, bool read_p,
> > }
> > }
> > if (outer_automatic_var_p (expr)
> > - && decl_constant_var_p (expr))
> > + && decl_constant_var_p (expr)
> > + && is_nondependent_constant_expression (expr))
>
> I think we want this condition to stay parallel with the one in
> process_outer_var_ref:
>
> > /* Only an odr-use of an outer automatic variable causes an
> > error, and a constant variable can decay to a prvalue
> > constant without odr-use. So don't complain yet. */
> > else if (!odr_use && decl_constant_var_p (var))
> > return var;
>
> ...though that comment should refer to mark_use, and vice versa.
>
> The idea is that first we try to process the ref from finish_id_expression,
> and if that's going to fail we wait and try again from mark_use.
>
> Patrick's suggestion
>
> > Maybe we can make both testcases work if we instead give
> > process_outer_var_ref
> > an early exit for when the outer variable is from a template?
>
> makes sense to me.
Just posted something to that effect. Thanks,
Marek
