Reliability of `bc' (or: is `bc' buggy?)

[Luuk Uljee]

In article <1991Mar1.093606.3280 at glance.ch>,
	david at glance.ch (David Mosberger) writes:
> A little test program to check a 64 bit division algorithm with `bc'
> unexpectedly turned into a test program for `bc' itself (after I found
> two bugs in my program :-).
> ..
> Does anybody know more about the reliability of `bc,' and what
> may cause these strange results with control characters in the output?


The following diff fixes this problem. It says nothing about the reliability
of bc/dc though.
The problem was caused by a miscalculation of the current high order digit of
the quotient. It could become one too high and thus the intermediate remainder
could get negative. Since this was unaccounted for, strange things resulted.
By printing digits that were out of range the minuses and the control chars
appeared. The original code only corrected for the digit being one too low.

(Warning: These diffs are t.o.v. SunOS 3.2 /usr/bin/dc.) 
------------------------------------------------------------------------------
*** dc.c	Mon Mar  4 13:08:11 1991
--- bdc.c	Mon Mar  4 18:51:48 1991
***************
*** 655,672 ****
  			}
  			salterc(divd,d);
  		}
! 		divcarry = carry;
  		sbackc(p);
  		salterc(p,dig);
  		sbackc(p);
  		if(--offset >= 0)divd->wt--;
- 	}
- 	if(divcarry != 0){
- 		salterc(p,dig-1);
- 		salterc(divd,-1);
- 		ps = add(divr,divd);
- 		release(divd);
- 		divd = ps;
  	}
  
  	rewind(p);
--- 655,677 ----
  			}
  			salterc(divd,d);
  		}
! 		if(carry){
! 			rewind(divr);
! 			seekc(divd,offset);
! 			dig --;
! 			carry=0;
! 			while(sfeof(divr) == 0){
! 				d = sgetc(divr)+sgetc(divd)+carry;
! 				carry = d / 100;
! 				sbackc(divd);
! 				salterc(divd,d%100);
! 			}
! 		}
! 
  		sbackc(p);
  		salterc(p,dig);
  		sbackc(p);
  		if(--offset >= 0)divd->wt--;
  	}
  
  	rewind(p);
-------------------------------------------------------------------------------
Luuk (luuk at cs.vu.nl).