bug#15320: [2.0.7] Rational number representation problem

bug#15320: [2.0.7] Rational number representation problem

[Josh Stokes]

[-- Attachment #1: Type: text/plain, Size: 254 bytes --]

It seems that certain small numbers are represented as integers, as (*
4294967296 4294967296) results in 0, but (* 3294967296 3294967296) returns
the correct value of 10856809481709551616. In addition, (ash 1 64) gives 0,
but (ash 1 65) works correctly.

[-- Attachment #2: Type: text/html, Size: 276 bytes --]

bug#15320: [2.0.7] Rational number representation problem

[Mark H Weaver]

Josh Stokes <jsstokes2@googlemail.com> writes:

> It seems that certain small numbers are represented as integers, as (*
> 4294967296 4294967296) results in 0, but (* 3294967296 3294967296)
> returns the correct value of 10856809481709551616. In addition, (ash 1
> 64) gives 0, but (ash 1 65) works correctly.

This is fixed in Guile 2.0.9.

In case you're curious: this bug came into existence when C compilers
started optimizing out overflow checks, on the theory that if a signed
integer overflow occurs then the behavior is unspecified and thus the
compiler can do whatever it likes.

http://stackoverflow.com/questions/14495636/strange-multiplication-behavior-in-guile-scheme-interpreter/14498437#14498437

    Thanks,
      Mark