[Vm-dev] Float math in VMMaker-dtl.123

[Nicolas Cellier]

2009/7/2 Nicolas Cellier <nicolas.cellier.aka.nice at gmail.com>:
> 2009/7/2 Eliot Miranda <eliot.miranda at gmail.com>:
>> So before we thrash breaking what looked a few weeks ago to be the right fix can we please write the relevant tests.  I nominate Nicholas as he wrote the bug fixes and revamped the tests we have in the first place.
>
> That's fair.
> I'll have a look this week end.
>
> Nicolas
>

OK, my week-end assignments could not start before Sunday night, but I
lately confirm VMMaker-dtl.124 is correct wrt Float comparison.

What I did:
- took a fresh 3.10.2 image.
- loaded LPF to get FFI installed
- loaded VMMaker-dtl.124
- initialized various VMMaker classes by hand because LPF Monticello
version did not do it
- downloaded latest platform sources with svn
- generated a VM for linux (all plugins external, otherwise would not
compile...)
- configured for src32 compiled and installed
- (Installer ensureFix: 6719) for fixing comparison
- added additional non regression tests for comparison
( M7260-FloatCompareNonRegressionTest-nice.2.cs at
http://bugs.squeak.org/view.php?id=7260 )
- ran all the KernelTests-Number and got 154 green tests, no failure, no error.

If you do not apply mantis fix 6719, then
1>=Float nan -> false. (thanks to new primitives)
But:
1 perform: #>= with: Float nan. -> true. (Magnitude not fixed...)
Optimizations are tricky...

Nicolas