Hi!
On Fri, Dec 12, 2025 at 09:39:18AM -0500, Patrick Palka wrote:
> + TREE_TYPE (templ)
> + = build_typename_type (TYPE_CONTEXT (TREE_TYPE (templ)),
> + TYPE_NAME (TREE_TYPE (templ)),
> + fullname,
> + get_typename_type_tag (TREE_TYPE (templ)));
> template_id = templ;
> }
> else
This change causes ICE e.g. on the following testcase.
The problem is that build_typename_type expects IDENTIFIER_NODE as the
second argument, e.g. it uses it as
tree d = build_decl (input_location, TYPE_DECL, name, t);
argument. But TYPE_NAME doesn't have to be an IDENTIFIER_NODE, it can
be a TYPE_DECL too and when we build a TYPE_DECL with TYPE_DECL as
DECL_NAME, it breaks all kinds of assumptions everywhere in the FE as well
as middle-end.
Fixed by using TYPE_IDENTIFIER instead.
So far bootstrapped successfully on x86_64-linux and i686-linux, ok
for trunk if it passes also regtest on both (in progress)?
2025-12-18 Jakub Jelinek <[email protected]>
PR c++/123186
* parser.cc (cp_parser_template_id): Use TYPE_IDENTIFIER instead of
TYPE_NAME in second build_typename_type argument.
* g++.dg/template/crash133.C: New test.
--- gcc/cp/parser.cc.jj 2025-12-17 15:21:21.045719368 +0100
+++ gcc/cp/parser.cc 2025-12-18 11:11:56.074952675 +0100
@@ -20435,7 +20435,7 @@ cp_parser_template_id (cp_parser *parser
fullname, arguments);
TREE_TYPE (templ)
= build_typename_type (TYPE_CONTEXT (TREE_TYPE (templ)),
- TYPE_NAME (TREE_TYPE (templ)),
+ TYPE_IDENTIFIER (TREE_TYPE (templ)),
fullname,
get_typename_tag (TREE_TYPE (templ)));
template_id = templ;
--- gcc/testsuite/g++.dg/template/crash133.C.jj 2025-12-18 13:22:56.797502427
+0100
+++ gcc/testsuite/g++.dg/template/crash133.C 2025-12-18 13:22:49.365630906
+0100
@@ -0,0 +1,6 @@
+// PR c++/123186
+
+template <class T>
+struct A : T {
+ typename A <T>::template B <42> C;
+};
Jakub
