dmy (12/15/82)
/*
The following program contains, as one line, the smallest version of a
c-optimizer bug that I could construct. I looked at the erroneous assembler
code produced by the optimizer, and this is definitely a compiler problem.
Anyone know of this bug before?
On our 4.1absd, running a.out after cc -O test.c produces 840.
After cc test.c, it produces the correct value of 50.
The bug (in the optimized object code) is a missing cvtdl (I believe).
*/
main()
{
float f;
int n;
f = 50.0;
n = (int)f & 07777; /* this statement is the bug */
printf("%d\n",n);
}
/*
--dmy--
*/