9 Jan
2007
9 Jan
'07
1:21 a.m.
Hello.
Given a conforming implementation, the following scenario is possible (I think... at least, that is the case with SBCL):
1. Some integers cannot be coerced into single-float.
2. FORMAT floating-point printer ~E chooses to coerce integers into single-float.
3. FORMAT fails to coerce a given integer and chooses to error.
The SLIME Inspector uses ~E when viewing integers, and fails to consider this scenario.
For example, try to inspect the result of evaluating (expt 2 1024).