Regression test RHEAD failing

[ghannington]

The test RHEAD in the regression suite (Report Header test ViewID) is failing with RCA.

I ran it and compared it to C++. It’s the SKE column that is changed to Zoned 16 for C++, but Java is leaving it as alpha 16.
Strangely the results of the Java one look better than the results of the C++ one, because of the sign nibble left in the zoned number.

[ghannington]

There are two issues here.

1. Java RCA is acting differently to MR91 in that the SKE column is not being flipped to Zoned.

2. the final output from the MR91 version is signed Zoned numeric, which looks unreadable in some cases (as opposed to the output from the RCA version), but the column is defined as unsigned.

I have traced this to an error in the MR95 model code.
There is no model for SKE Binary length 8 unsigned to Numeric unsigned.
It is using the model for SKE Binary length 8 unsigned to Numeric signed.

[ghannington]

I am creating more tests to see what happens for Binary 4, binary 2 and binary 1 sources.

[ghannington]

See issue in Performance Engine repo:
genevaers/Performance-Engine#27