Arithmetic errors on some field calculations

[Patch]

Or round to a precision consistent with the internal number representation.

Fix precision works well when adding units of currency. But not as well with other algebra operations.

Sounds like a prior issue

@Patch

Double precision floating point numbers have a precision of about 14 decimal digits. As this is the most common floating point format, it is probably what is used by Manger.

Lubos may be able to display number rounded to say 10 significant decimal digits and hide most floating point calculation limitations.