Running imvirt multiple times results in unstable and incorrect output about imvirt

[alexbarton]

Hm … probably this is a duplicate of (or at least related to) #14 – the report looks quite similar:

• $VAR1 looks similar between runs, numbers are all the same, but the sorting differs.
• and the "greather than" test uses different numers for the good vs. bad case …

[liske]

Hi Alex,

I think it is a duplicate of #15 ... but your details on both cases let me found the issue:

ImVirt: imv_get: 0.653333333333333 - sqrt(0.1024) > 0.0133333333333333

vs.

ImVirt: imv_get: 0.653333333333333 - sqrt(0.1024) > 0.333333333333333

The $VAR1 hash (virtualization => probility) is sorted, the candidate with the highest probability is returned if, and only if it has enough distance to the second top-most candidate. I've looked long time and intensively on the code and everything seems to be OK. But than I realized that the sort function is broken:

ImVirt.pm:261    my @res = sort { $rtree{$b} > $rtree{$a} } keys %rtree;

should read:

ImVirt.pm:261    my @res = sort { $rtree{$b} <=> $rtree{$a} } keys %rtree;

to sort correctly!

[alexbarton]

Thanks a lot for fixing this @liske!