Bigint : better ensure canonicity of arrays of int blocks

In the chosen representation of negative numbers, (-base) is
actually legal as first element of an int array. This situation
was wrongly handled by many functions in this module, leading to
possible non-canonical representation of a big number, and hence
faulty "equal" answers or buggy output of "to_string".

For instance, on a 32-bit machine :

open Big_int;;
to_string (sub (of_string "-10000") (of_string "-1000000"));;
(* was displaying 9810000 instead of 990000 *)

Also simplified "of_string", which now rely on "neg" for handling
negative numbers.

Btw, improved checks comparing w.r.t OCaml's Big_int instead of float.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/coq/trunk@15665 85f007b7-540e-0410-9357-904b9bb8a0f7

0 files changed, 0 insertions, 0 deletions