garyd@chemabs (garyd) (03/22/85)
C Preprocessor BUG in IS/3 and System V with Hexidecimal Constants.
When using hexadecimal constants as part of a #IF <constant-expression> the
preprocessor evaluates the hexadecimal digits A through F (or a - f) not as 10
to 15 (decimal), but rather as 0 to 5. This holds true for each hexidecimal
nybble. Thus 0xff is really interpreted by the preprocessor as a binary
01010101. And 0xa is evaluated as zero, which can easily be demonstrated by
the following C program which by all rights should print true:
main()
{
#if (0XA)
printf("true\n");
#else
printf("false\n");
#endif
}
This bug was first discovered in the IS/3 version of the C Preprocessor, and
was found to be hiding in the System V source code also. The Whitesmiths' C
preprocessors handle this properly. This error is in the yylex.c source
code file which is part of the cpp program. The function containing the
error is tobinary(), which follows:
73 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
74 t = c-'a'; if (b>10) break;
75 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
76 t = c-'A'; if (b>10) break;
This code might be corrected as follows:
73 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
74 t = c-'a'+10; if (b>10) break;
75 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
76 t = c-'A'+10; if (b>10) break;
Gary A. Davis
Senior Engineer
Chemical Abstracts Service
UUCP: ..{ihnp4|mhuxl|ho95b|ulysses}!cbosgd!osu-eddie!apr!chemabs!garyd
AT&T: (614) 421-3600 X2355
USPS: Chemical Abstracts Service
P.O. Box 3012
Columbus, Ohio 43210hansen@pegasus.UUCP (Tony L. Hansen) (03/23/85)
Re: #if 0xAA == 0 Neither System V release 2 nor 4.2BSD contain this bug. Tony Hansen pegasus!hansen