https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122283
Bug ID: 122283
Summary: Incomplete error and bailing out if a consteval ctor
delegates to a non-constexpr one
Product: gcc
Version: unknown
Status: UNCONFIRMED
Keywords: error-recovery, ice-on-invalid-code
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: blubban at gmail dot com
Target Milestone: ---
template<int a>
struct foo {
int b;
foo(int) {}
consteval foo() : foo(1) {}
};
void bar()
{
foo<2> c;
}
-std=c++20
Expected:
<source>: In function 'void bar()':
<source>:11:16: error: call to consteval function 'a.foo<2>::foo()' is not a
constant expression
11 | foo<2> a;
| ^
<source>:11:16: error: 'consteval foo<a>::foo() [with int a = 2]' called in a
constant expression
<source>:6:19: note: 'consteval foo<a>::foo() [with int a = 2]' is not usable
as a 'constexpr' function because:
6 | consteval foo() : foo(1) {}
| ^~~
<source>:6:32: error: call to non-'constexpr' function 'foo<a>::foo(int) [with
int a = 2]'
6 | consteval foo() : foo(1) {}
| ^
<source>:5:9: note: 'foo<a>::foo(int) [with int a = 2]' declared here
5 | foo(int) {}
| ^~~
Compiler returned: 1
Actual:
<source>: In function 'void bar()':
<source>:11:16: error: call to consteval function 'a.foo<2>::foo()' is not a
constant expression
11 | foo<2> a;
| ^
<source>:11:16: error: 'consteval foo<a>::foo() [with int a = 2]' called in a
constant expression
<source>:6:19: note: 'consteval foo<a>::foo() [with int a = 2]' is not usable
as a 'constexpr' function because:
6 | consteval foo() : foo(1) {}
| ^~~
<source>:6:19: internal compiler error: in build_data_member_initialization, at
cp/constexpr.cc:466
0x28de9b8 diagnostics::context::diagnostic_impl(rich_location*,
diagnostics::metadata const*, diagnostics::option_id, char const*,
__va_list_tag (*) [1], diagnostics::kind)
???:0
0x28d37ab internal_error(char const*, ...)
???:0
0xb05622 fancy_abort(char const*, int, char const*)
???:0
0xb8850a explain_invalid_constexpr_fn(tree_node*)
???:0
0xb82bb3 cxx_constant_value(tree_node*, tree_node*, int)
???:0
0x189208c walk_tree_1(tree_node**, tree_node* (*)(tree_node**, int*, void*),
void*, hash_set<tree_node*, false, default_hash_traits<tree_node*> >*,
tree_node* (*)(tree_node**, int*, tree_node* (*)(tree_node**, int*, void*),
void*, hash_set<tree_node*, false, default_hash_traits<tree_node*> >*))
???:0
0x18928f6 walk_tree_1(tree_node**, tree_node* (*)(tree_node**, int*, void*),
void*, hash_set<tree_node*, false, default_hash_traits<tree_node*> >*,
tree_node* (*)(tree_node**, int*, tree_node* (*)(tree_node**, int*, void*),
void*, hash_set<tree_node*, false, default_hash_traits<tree_node*> >*))
???:0
0xbb18dc cp_fold_function(tree_node*)
???:0
0xc00ea9 finish_function(bool)
???:0
0xd3e213 c_parse_file()
???:0
0xeae8b9 c_common_parse_file()
???:0
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.
Compiler returned: 1
Removing the member variable makes it not ICE, but instead, it prints the final
error and note twice.
https://godbolt.org/z/KnfsKWrez
Like bug 102262, it's an incomplete error due to missing constexpr keyword, but
that one doesn't ICE, so I'm guessing it's a different root cause.
