Hi!

The following testcase ICEs since r12-6080.  The problem is that
lookup_fnfields can return NULL_TREE on failure, but the maybe_incomplete
handling was added before the if (!baselink) handling and assumes that
baselink is non-NULL (and BASELINK).

The following patch reorders the if (maybe_incomplete) handling with
if (!baselink).

Bootstrapped/regtested on x86_64-linux and i686-linux, approved in the
PR by Patrick, committed to trunk (so far).

2025-11-20  Jakub Jelinek  <[email protected]>

        PR c++/120876
        * pt.cc (tsubst_baselink): Move maybe_incomplete handling after
        !baselink handling.

        * g++.dg/parse/crash81.C: New test.

--- gcc/cp/pt.cc.jj     2025-11-17 09:39:39.627473470 +0100
+++ gcc/cp/pt.cc        2025-11-19 16:35:48.922884412 +0100
@@ -17625,19 +17625,6 @@ tsubst_baselink (tree baselink, tree obj
       bool maybe_incomplete = BASELINK_FUNCTIONS_MAYBE_INCOMPLETE_P (baselink);
       baselink = lookup_fnfields (qualifying_scope, name, /*protect=*/1,
                                  complain);
-      if (maybe_incomplete)
-       {
-         /* Filter out from the new lookup set those functions which didn't
-            appear in the original lookup set (in a less specialized form).
-            This is needed to preserve the consistency of member lookup
-            performed in an incomplete-class context, within which
-            later-declared members ought to remain invisible.  */
-         BASELINK_FUNCTIONS (baselink)
-           = filter_memfn_lookup (fns, BASELINK_FUNCTIONS (baselink),
-                                  binfo_type);
-         BASELINK_FUNCTIONS_MAYBE_INCOMPLETE_P (baselink) = true;
-       }
-
       if (!baselink)
        {
          if (complain & tf_error)
@@ -17655,6 +17642,19 @@ tsubst_baselink (tree baselink, tree obj
          return error_mark_node;
        }
 
+      if (maybe_incomplete)
+       {
+         /* Filter out from the new lookup set those functions which didn't
+            appear in the original lookup set (in a less specialized form).
+            This is needed to preserve the consistency of member lookup
+            performed in an incomplete-class context, within which
+            later-declared members ought to remain invisible.  */
+         BASELINK_FUNCTIONS (baselink)
+           = filter_memfn_lookup (fns, BASELINK_FUNCTIONS (baselink),
+                                  binfo_type);
+         BASELINK_FUNCTIONS_MAYBE_INCOMPLETE_P (baselink) = true;
+       }
+
       fns = BASELINK_FUNCTIONS (baselink);
     }
   else
--- gcc/testsuite/g++.dg/parse/crash81.C.jj     2025-11-19 16:41:21.651130105 
+0100
+++ gcc/testsuite/g++.dg/parse/crash81.C        2025-11-19 16:40:35.614787909 
+0100
@@ -0,0 +1,14 @@
+// PR c++/120876
+// { dg-do compile { target c++11 } }
+
+template <typename T>
+struct S {
+  static bool foo (decltype (bar (T {})));     // { dg-error "'bar' was not 
declared in this scope; did you mean 'baz'\\\?" }
+  static constexpr bool s = foo (0);           // { dg-error "declaration of 
'S<int>::foo' depends on itself" }
+};
+
+void
+baz ()
+{
+  S <int>::s;
+}

        Jakub

Reply via email to