[Bug c++/98130] New: [11 regression] placement new fails on webkit-gtk-2.28.4

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98130

            Bug ID: 98130
           Summary: [11 regression] placement new fails on
                    webkit-gtk-2.28.4
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: slyfox at gcc dot gnu.org
  Target Milestone: ---

Initial problem is observed on liferea crash at startup against
webkit-gtk-2.28.4 built with gcc-11.

If I reduced original source correctly here is minimized version of the crash:

//$ cat a.cc
/*

  $ g++-10.2.0 -O1 -fno-strict-aliasing a.cc -o a-10 && ./a-10
  &a=0x7ffc83475894
  $ g++-11.0.0 -O1 -fno-strict-aliasing a.cc -o a-11 && ./a-11
  Illegal instruction     (core dumped) ./a-11

*/

#if 0
#  include <memory>
#else
inline void* operator new(unsigned long, void* __p) { return __p; }
#endif

#include <stdio.h>

typedef int* T;

static T storage;
static T* p = &storage;

// '__attribute__((__always_inline__))' seems to be the trigger
static inline __attribute__((__always_inline__)) void append(T value)
{
    new (p) T(value);
}

int main() {
    int a;
    append(&a);

    if (!*p)
        __builtin_trap();
    fprintf(stderr, "&a=%p\n", *p);
}