[Bug tree-optimization/102586] [12 Regression] ICE in clear_padding_type, at gimple-fold.c:4798 since r12-3433-ga25e0b5e6ac8a77a

[qing.zhao at oracle dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102586

--- Comment #27 from Qing Zhao <qing.zhao at oracle dot com> ---
> On Feb 11, 2022, at 11:47 AM, jakub at gcc dot gnu.org <gcc-bugzilla@gcc.gnu.org> wrote:
> But if I do:
> struct C { char a; int b; char c; long d; C () : a (42), b (42), c (42), d (42)
> {} };
> void bar (C *);
> void
> foo ()
> {
>  C c;
>  bar (&c);
> }
> then *.gimple is:
>      c = .DEFERRED_INIT (24, 1, &"c"[0]);
>      __builtin_clear_padding (&c, 0B, 1);
>      C::C (&c);
>      bar (&c);
> ...
> void C::C (struct C * const this)
> {
>  *this = {CLOBBER};
>  {
>    this->a = 42;
>    this->b = 42;
>    this->c = 42;
>    this->d = 42;
>  }
> }
> After einline this is:
>  c = .DEFERRED_INIT (24, 1, &"c"[0]);
>  MEM <char[3]> [(struct C *)&c + 1B] = {};
>  MEM <char[7]> [(struct C *)&c + 9B] = {};
>  c ={v} {CLOBBER};
>  c.a = 42;
>  c.b = 42;
>  c.c = 42;
>  c.d = 42;
>  bar (&c);
> and that keeps until dse1 which optimizes that out:
>  c ={v} {CLOBBER};
>  c.a = 42;
>  c.b = 42;
>  c.c = 42;
>  c.d = 42;
>  bar (&c);
> so there is no zero padding initialization at all.

Does this issue only exist with -flifetime-dse=2?
When -flifetime-dse=2, the call to __builtin_clear_padding should be inserted
AFTER the
start point of the constructor of the object, otherwise it’s dead and will be
eliminated by DSE. 
And with -flifetime-dse=2, the padding initialization should be done by C++ FE
instead of middle end.
Is this understanding correct?