On 8/3/25 12:23 PM, Patrick Palka wrote:
On Sun, 3 Aug 2025, Jason Merrill wrote:

On 8/1/25 11:16 AM, Patrick Palka wrote:
Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK
for trunk and 15/14 after 15.2 is released?

-- >8 --

r13-3299 changed our internal declaration of __dynamic_cast to reside
inside the abi / __cxxabiv1 namespace instead of the global namespace,
matching the real declaration.  This inadvertently made us now attempt
constexpr evaluation of user-written calls to abi::__dynamic_cast since
cxx_dynamic_cast_fn_p now also returns true for them, but we're not
prepared to handle arbitrary calls to __dynamic_cast, and therefore ICE.

This patch restores cxx_dynamic_cast_fn_p to return true only for
synthesized calls to __dynamic_cast, which can be distinguished by
DECL_ARTIFICIAL.

        PR c++/120620

gcc/cp/ChangeLog:

        * constexpr.cc (cxx_dynamic_cast_fn_p): Return true only
        for synthesized __dynamic_cast.

gcc/testsuite/ChangeLog:

        * g++.dg/cpp2a/constexpr-dynamic19.C: New test.
---
   gcc/cp/constexpr.cc                              |  6 +++++-
   gcc/testsuite/g++.dg/cpp2a/constexpr-dynamic19.C | 10 ++++++++++
   2 files changed, 15 insertions(+), 1 deletion(-)
   create mode 100644 gcc/testsuite/g++.dg/cpp2a/constexpr-dynamic19.C

diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
index e051a50fe16e..4cf94afb9ca5 100644
--- a/gcc/cp/constexpr.cc
+++ b/gcc/cp/constexpr.cc
@@ -3240,7 +3240,11 @@ cxx_dynamic_cast_fn_p (tree fndecl)
   {
     return (cxx_dialect >= cxx20
          && id_equal (DECL_NAME (fndecl), "__dynamic_cast")
-         && CP_DECL_CONTEXT (fndecl) == abi_node);
+         && CP_DECL_CONTEXT (fndecl) == abi_node
+         /* Only consider implementation-detail __dynamic_cast calls that
+            correspond to an actual dynamic_cast, and ignore direct calls
+            to abi::__dynamic_cast.  */
+         && DECL_ARTIFICIAL (fndecl));

With this patch, does constexpr dynamic_cast still work after including
cxxabi.h?

Yes AFAICT.  I don't see duplicate_decls being called on __dynamic_cast
at all when including <cxxabi.h>, so presumably both declarations are
coexistiing.  Maybe because the synthesized declaration is pushed with
hidden=true or because it's not declared as extern "C" like the real
declaration?

Then please add a testcase for that case or add #include <cxxabi.h> to an existing constexpr dynamic_cast test. OK with that addition.

Jason

Reply via email to