float->double->float conversion with 19991122 snapshot (Linux/Intel)

[Paul Kimoto]

On Tue, Nov 23, 1999 at 02:33:01PM -0500, Paul Kimoto wrote:
> I am trying the 19991122 snapshot on Linux/Intel with glibc-2.1.2.

This problem is still present in the 19991201 snapshot.  
A test function is

/* begin C code */
float
fdf (float xx)
{
   double x = xx;
   return x;
}
/* end C code */

When compiled with "-O0", this gives me results like 
fdf(0.5) = 2.145722.  The right(?) thing happens when
I compile with "-O1", even with "-O1 -fno-defer-pop 
-fno-thread-jumps -fno-inline".

-- 
Paul Kimoto <kimoto@lightlink.com>