CODEWITHSUNDEEP
divide-by-zeroforthgforth
utamdekq
utamdekq

Reputation: 31

Why do division by zero errors get raised for non-zero divisors?

Consider:

9223372034707292159. 1 um/mod

For example, it raises a division-by-zero error on my machine. Why? The divisor is 1 (not 0!) and 9223372034707292159 fits within an unsigned double.

The same goes for fm/mod and sm/rem.

I'm using Gforth, but these are core Forth words.

Upvotes: 2

Views: 147

Answers (1)

ruvim
ruvim

Reputation: 8554

See the specification for UM/MOD ( ud u1 -- u2 u3 ):

An ambiguous condition exists if u1 is zero or if the quotient lies outside the range of a single-cell unsigned integer.

If you use a Forth system having a cell size of 32 bits, then in the given case the quotient lies outside the rage of a single-cell unsigned integer, and then it's an ambiguous condition and the behavior of a system is not specified by the standard.

Certainly, a better error message would be "integer overflow" than "division by zero" in this case.

[update] As it later appeared, in Gforth "the -fast engine does make some compromises speed vs. accuracy of error messages".

Upvotes: 4

Related Questions