#967 uinteger (64 bit) seems to be limited at (2^63) for uinteger variables only

[Vincent Chapman]

Confusing behaviour for uniteger, in uniteger variables only, on FBC Linux 64 bit platform

[Vincent Chapman]

Aside from dumb mistakes in my example, clarifying the problem as:

print (2 ^ 64) - 1 'this works

dim a as uinteger = ((2 ^ 64 )  - 1)
print a ' returns a zero

now

print (2 ^ 63) - 1: print (2 ^ 64) - 1 'both work for signed integers

dim min as integer = ((2 ^ 64) - 1) 
dim max as integer = (2 ^ 63) - 1)

print min; max'this works

Does not appear to be handling uinteger declared variables correctly pas 63 bits

root@abaddon:/usr/local/src/basic# fbc --version
FreeBASIC Compiler - Version 1.09.0 (2022-01-01), built for linux-x86_64 (64bit)
Copyright (C) 2004-2021 The FreeBASIC development team.

[TeeEmCee]

Not a bug. The ^ operator returns a double, as your own examples show. Double precision floats have only 53 bits of precision. They can't actually store the number 2 ^ 64 - 1!

var x = 2 ^ 64
var y = 2 ^ 64 - 1
dim t as boolean = (x = y)
print t  'true

If you look more closely at your examples, the lines you say "work" actually do not. print max prints -9223372036854775808 instead of the correct value 9223372036854775807.

The solution is to write 1 shl 63 instead of 2 ^ 63. However! You can't write 1 shl 64 because on x86 CPUs (but IIRC not ARM) that's equivalent to 1 shl 0 which is equal to 1 instead of the expected result, 0. Thankfully fbc will warn you about the overflow.

[Jeff Marshall]

I have reviewed and reworded some notes about this in the documentation.
https://www.freebasic.net/wiki/KeyPgOpExponentiate