07:29 <inode> hosewiejacke: are you using gforth compiled for x86-64?

07:30 <hosewiejacke> Yes, it's from the OpenBSD packages.

07:33 <inode> ok, i found that the Division by Zero message seems to occur not only for division by zero, but for the error raised by the idiv instruction when the result overflows as boru suggested

07:34 <hosewiejacke> I'm just discovered there are snapshots on https://www.complang.tuwien.ac.at/forth/gforth/Snapshots/

07:50 <hosewiejacke> Compiling the gforth snapshot from 2020-12-03 seems to require a swig version that has a -forth option. swig 4.0.2 doesn't seem to have this option...

07:52 <dave0> inode: the x86 unsigned divide only overflows if the high word of the dividend is greater or equal to the divisor

07:54 <dave0> inode: you can check for overflow in advance (rdx:rax / rcx) cmp rdx,rcx ; jae overflow ; div rcx

07:54 <dave0> that also catches divide by zero (because if rcx=0, rdx is always >= rcx)

07:55 <dave0> don't know about signed divide :-(

08:05 <inode> i was going on the operation described in https://www.felixcloutier.com/x86/idiv (probably should be looking at intel's own spec though)

08:09 <inode> and some debugging session info @ https://pastebin.com/qvfSwN95

08:09 <dave0> inode: oh that looks like it's from the intel spec anyways

08:11 <dave0> c is so hard to read when it's in assembly :-p

08:11 <inode> heh

08:12 <hosewiejacke> Esp. when it's amd64 assembly :)