>No nothing changes on this areas.... But another thing your code doesn't in fact works. Check with sample data that I post.
It's the implementation of the algorithm as it was back then, although the one for SDK, i.e. account numbers. If they applied a different version of the algorithm to J/EMBG, then of course this one won't work.
And you're right, I checked against some old demo data I have on disk, and the check simply isn't the right one for the JMBG. I've found the algorithm at
http://www.regos.hr/UserDocsImages/downloads/algoritmiobrazacr-s.pdf (it's a PDF) and it is a variation of mod11() - it uses two groups of six digits, multiplying 1st with 7, 2nd with 6, 3rd with 5 etc then 7th with 7, 8th with 6... 12th with 2, adding it all up, then calculating the mod(sum,11). The last digit needs to be such that this mod() is zero (or 10).
It's in Croatian, but we know how different that is :).