[Bug c++/120506] New: [16 Regression] Missing reason for failed constinit since r16-57

[redi at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120506

            Bug ID: 120506
           Summary: [16 Regression] Missing reason for failed constinit
                    since r16-57
           Product: gcc
           Version: 16.0
            Status: UNCONFIRMED
          Keywords: diagnostic
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: redi at gcc dot gnu.org
                CC: jason at gcc dot gnu.org
  Target Milestone: ---

Given this code to be compiled with -std=c++20:

struct A
{
  constexpr A(int c) : counter(c) { }

  int counter;
};


struct B : A
{
  constexpr B(int c) : A(c) { }

  int i; // OOPS, not initialized
};

struct C
{
  B sem;

  constexpr C(int c) : sem(c) { }
};

constinit C s(0);


GCC 15 explains the problem:


sem.cc:23:13: error: 'constinit' variable 's' does not have a constant
initializer
   23 | constinit C s(0);
      |             ^
sem.cc:23:16: error: 'C{B{A{0}}}' is not a constant expression
   23 | constinit C s(0);
      |                ^
sem.cc:23:16: error: 's.C::C(0)' is not a constant expression because it refers
to an incompletely initialized variable


Since r16-57-g9ac98b5742ebce we're missing the last error:


sem.cc:23:13: error: 'constinit' variable 's' does not have a constant
initializer
   23 | constinit C s(0);
      |             ^
sem.cc:23:16: error: 'C{B{A{0}}}' is not a constant expression
   23 | constinit C s(0);
      |                ^


I refactored the class B to move the member into a base class A, but forgot to
remove the B::i member, and the error was not at all helpful in understanding
why I couldn't use a constinit variable!


r16-57 was:

    c++: reorder constexpr checks

    My proposed change to stop setting TREE_STATIC on constexpr heap
    pseudo-variables led to a diagnostic regression because we would get the
    generic "not constant" diagnostic before the "allocated storage"
diagnostic.
    So let's move the generic verify_constant down a bit.

    gcc/cp/ChangeLog:

            * constexpr.cc (cxx_eval_outermost_constant_expr): Move
            verify_constant later.