#697 floor and truncate round up for some numbers

[Leo]

(floor 13158008.6)
13158009
It should be 13158008.
Same with truncate.

[Bruno Haible]

It's not the FLOOR function which rounds up, it's the parsing of the number:

(rationalize 13158008.6) => 13158009
read-default-float-format => SINGLE-FLOAT

It you want more than 6 digits of precision with floating-point numbers, you need to force double-precision, either explicitly

(floor 13158008.6d0) => 13158008 ; 0.599999999627471d0

or by changing read-default-float-format :

(setq read-default-float-format 'double-float) => DOUBLE-FLOAT
(floor 13158008.6) => 13158008 ; 0.599999999627471d0