mmuegel@fwhnm02.fwrdc.rtsg.mot.com (Michael S. Muegel) (05/15/91)
Just wanted to warn everyone of some major cc6.8 bugs we have been running
into recently. Many have to do with optimization (imagine that ;-). Some,
however, rear their ugly heads with some very simple code. All occur only on
the 680x0 architecture.
Here is one example:
#include <stdio.h>
main()
{
long value;
value = 0x10000L;
#ifdef PRINT_INITIAL
printf(" Initial value = %lX\n", value);
#endif
value &= 0xffffL;
printf(" Final value = %lX\n", value);
}
And some results:
% cc -DPRINT_INITIAL -DRA -g -W0,-opt,4 -o test2 test2.c ;test2
Initial value = 10000
Final value = 0
% cc -DRA -g -W0,-opt,4 -o test2 test2.c ;test2
Final value = FF00
In the above example, the OPT level is 4, and the result is correct if the
printf() is compiled, but incorrect if the printf() is not compiled.
We have some more complex examples that show cc errors when optimization is
turned OFF! After calling this in to HP/Apollo (and it was verified by them
as a bug) I received a new rev of cc6.8. Here are the relevant time stamps from
the executables in /usr/apollo/lib:
Ver Name Time Stamp File Name
--------------------------------------------------------------
c 1 cc 1990/12/06 19:33:07 CDT (Thu) cc.m68k.old
c 1 cc 1991/04/24 12:57:23 CDT (Wed) cc.m68k
The newest cc.m68k is version 6.82(191).
-Mike
--
+-----------------------------------------------------------------------------+
| Mike Muegel | Internet: mmuegel@mot.com |
| Software Tools Group | UUCP: uunet!motcid!muegel |
| Fort Worth Research & Development Center | Voice: (817) 232-6129 |
| Radio Telephone and Systems Group | ... My opinions are surely not |
| Motorola, Inc. | shared by Motorola :-& |
+-----------------------------------------------------------------------------+