[Bug target/58416] Incorrect x87-based union copying code

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58416

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |wrong-code
   Last reconfirmed|                            |2024-02-20
                 CC|                            |jamborm at gcc dot gnu.org,
                   |                            |rguenth at gcc dot gnu.org

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
Reconfirmed with the comment#5 testcase.  ESRA does

 int main ()
 {
+  long double x$d;
   unsigned char * p;
   union u y;
   union u x;
@@ -25,9 +20,9 @@
   int _2;

   <bb 2> :
-  x.d = 0.0;
-  x.s.s = "xxxxxxxxxxxxxxxx";
-  y = x;
+  x$d_3 = 0.0;
+  x$d_16 = MEM[(char[16] *)"xxxxxxxxxxxxxxxx"];
+  MEM[(union u *)&y] = x$d_16;

and the value re-interpretation goes off.  Martin - we may have dups of this
but SRA shouldn't do this (it possibly gets confused by the x.d store)?