On 1/29/26 4:32 PM, Jakub Jelinek wrote:
On Wed, Jan 28, 2026 at 05:10:18PM -0500, Marek Polacek wrote:
Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
-- >8 --
As discussed in
<https://gcc.gnu.org/pipermail/gcc-patches/2026-January/705756.html>,
we should check TYPE_BEING_DEFINED along with COMPLETE_TYPE_P in
eval_define_aggregate. It seems that with this check added, we don't
need this code anymore:
if (c == type)
{
auto_diagnostic_group d;
error_at (loc, "%<define_aggregate%> evaluated from "
"%<consteval%> block enclosed by %qT being "
"defined", type);
//...
}
but I'm not removing that in this patch.
Why not?
gcc/cp/ChangeLog:
* reflect.cc (eval_define_aggregate): Also give an error when
TYPE_BEING_DEFINED is true for the first argument.
gcc/testsuite/ChangeLog:
* g++.dg/reflect/define_aggregate3.C: Adjust expected diagnostic.
* g++.dg/reflect/p2996-21.C: Likewise.
---
gcc/cp/reflect.cc | 8 ++++++++
gcc/testsuite/g++.dg/reflect/define_aggregate3.C | 16 ++++++++--------
gcc/testsuite/g++.dg/reflect/p2996-21.C | 4 ++--
3 files changed, 18 insertions(+), 10 deletions(-)
diff --git a/gcc/cp/reflect.cc b/gcc/cp/reflect.cc
index bdeec2f0f38..4845e648fdf 100644
--- a/gcc/cp/reflect.cc
+++ b/gcc/cp/reflect.cc
@@ -5896,6 +5896,14 @@ eval_define_aggregate (location_t loc, const
constexpr_ctx *ctx,
*non_constant_p = true;
return call;
}
+ if (TYPE_BEING_DEFINED (type))
+ {
+ if (!cxx_constexpr_quiet_p (ctx))
+ error_at (loc, "first %<define_aggregate%> argument is a reflection "
+ "of a class type being defined");
I wonder if this message shouldn't mention the type like the diagnostics
that is obsoleted by this.
While in cases like S3 it is obvious, sometimes the define_aggregate might
be in a consteval call with completely different locus and figuring out
what is the type being defined might be harder.
Please.
Jason
