[Bug c/116357] [12/13/14/15 Regression] The item's address of the array is not correct if aligned is used

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116357

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|wrong-code                  |accepts-invalid
           Priority|P3                          |P2

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
Well, then the frontend should reject this.  We reject

typedef int int64 __attribute__((aligned (8)));
int64 a[4];

but not

typedef volatile int int64 __attribute__((aligned (8)));
int64 a[4];

somehow the latter arrives as 'int' in c_build_array_type but the former
has 'int64'.  This is because of grokdeclarator doing

  if ((TREE_CODE (type) == ARRAY_TYPE
       || first_non_attr_kind == cdk_array)
      && TYPE_QUALS (element_type))
    {   
      orig_qual_type = type;
      type = TYPE_MAIN_VARIANT (type);
    }

which not only drops the volatile qualifier from int64 but also the alignment.
This later confuses the rest of the frontend with how it treats the array.