public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug c++/104163] New: Misleading diagnostic for incompletely initialized constinit variable
@ 2022-01-21 13:36 redi at gcc dot gnu.org
2022-01-21 13:38 ` [Bug c++/104163] " redi at gcc dot gnu.org
0 siblings, 1 reply; 2+ messages in thread
From: redi at gcc dot gnu.org @ 2022-01-21 13:36 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104163
Bug ID: 104163
Summary: Misleading diagnostic for incompletely initialized
constinit variable
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Keywords: diagnostic
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: redi at gcc dot gnu.org
Target Milestone: ---
struct B
{
int i = 0;
};
struct C
{
void* p;
B b;
constexpr C() = default;
};
constinit C c{};
This is invalid because C::p is left uninitialized by the constructor, but
GCC's warning is confusing:
mut.C:13:13: error: 'constinit' variable 'c' does not have a constant
initializer
13 | constinit C c{};
| ^
mut.C:13:15: error: 'C{B{0}}' is not a constant expression
13 | constinit C c{};
| ^
mut.C:13:15: error: 'C()' is not a constant expression because it refers to an
incompletely initialized variable
The "'C{B{0}}' is not a constant expression" makes it look like the problem is
related to initialization of B. There is no mention of the p member. I spent
quite a bit of time trying to figure out what was wrong in a more complex,
deeply nested example.
Clang tells you exactly what's wrong:
mut.C:13:13: error: variable does not have a constant initializer
constinit C c{};
^~~
mut.C:13:1: note: required by 'constinit' specifier here
constinit C c{};
^~~~~~~~~
mut.C:13:13: note: subobject of type 'void *' is not initialized
constinit C c{};
^
mut.C:8:9: note: subobject declared here
void* p;
^
1 error generated.
The notes about the void* are clear and precise.
EDG is less clear, but at least still identifies C::p as the problem:
"mut.C", line 13: error: expression must have a constant value
constinit C c{};
^
"mut.C", line 13: note: access to uninitialized subobject (member "C::p")
constinit C c{};
^
1 error detected in the compilation of "mut.C".
^ permalink raw reply [flat|nested] 2+ messages in thread
* [Bug c++/104163] Misleading diagnostic for incompletely initialized constinit variable
2022-01-21 13:36 [Bug c++/104163] New: Misleading diagnostic for incompletely initialized constinit variable redi at gcc dot gnu.org
@ 2022-01-21 13:38 ` redi at gcc dot gnu.org
0 siblings, 0 replies; 2+ messages in thread
From: redi at gcc dot gnu.org @ 2022-01-21 13:38 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104163
Jonathan Wakely <redi at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2022-01-21
Status|UNCONFIRMED |NEW
Ever confirmed|0 |1
--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> ---
The deeply nested case was:
at.C:9:48: error: 'std::atomic<std::shared_ptr<int>
>{std::_Sp_atomic<std::shared_ptr<int> >{std::_Sp_atomic<std::shared_ptr<int>
>::_Atomic_count{std::__atomic_base<long unsigned int>{0}}}}' is not a constant
expression
The real problem was a member of _Sp_atomic, so everything about
std::_Sp_atomic<std::shared_ptr<int> >::_Atomic_count{std::__atomic_base<long
unsigned int>{0}} was a red herring.
Simplified:
struct atomic_base
{
constexpr atomic_base(int i) : i(i) { }
int i{0};
};
struct Sp_atomic
{
struct Atomic_count
{
constexpr Atomic_count() = default;
atomic_base a{0};
};
void* ptr;
Atomic_count c{};
constexpr Sp_atomic() = default;
};
struct atomic
{
constexpr atomic() = default;
Sp_atomic s;
};
constinit atomic a{};
mut.C:27:18: error: 'constinit' variable 'a' does not have a constant
initializer
27 | constinit atomic a{};
| ^
mut.C:27:20: error:
'atomic{Sp_atomic{Sp_atomic::Atomic_count{atomic_base{0}}}}' is not a constant
expression
27 | constinit atomic a{};
| ^
mut.C:27:20: error: 'atomic()' is not a constant expression because it refers
to an incompletely initialized variable
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2022-01-21 13:38 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-01-21 13:36 [Bug c++/104163] New: Misleading diagnostic for incompletely initialized constinit variable redi at gcc dot gnu.org
2022-01-21 13:38 ` [Bug c++/104163] " redi at gcc dot gnu.org
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).