Re: FAIL: gcc.c-torture/execute/950628-1.c execution, -O1

["John David Anglin" <dave at hiauly1 dot hia dot nrc dot ca>]

>     > However, I still come back to the issue of whether it's worth imposing
>     > the restriction of only one store.
> 
>     And I'll come back to the issue of changing "unchanging" memory.
> 
> Oh, I certainly agree that it's a conceptually-confusing issue, but the
> cost is significant.
> 
> If you have two constant variables that would normally be able to occupy the
> same stack slot, they won't be able to anymore.  That probably doesn't matter
> at all in C, is likely more significant in C++ and even moreso in Ada.

I have to wonder if this debate is missing the mark with respect to the
/u flag.  This is the function where the original problem was noted:

typedef struct
{
  char hours, day, month;
  short year;
} T;

T f (void)
{
  T virk;

  virk = g ();
  return virk;
}

As ~r noted, there is no "const" anywhere and yet the stores to virk have
the /u bit set.  I changed f to set `virk = g ()' twice and the /u bits
are there for both assignments.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)