We are migrating our forecasting database from Oracle 10.2 to Oracle 12.1. This means that we have to replace the Oracle ProC 9.2.0.3.0 pre-compiler with ProC 10.2.0.5.0.
All of our application programs are built and run under OpenVMS Alpha 8.4. We have traditionally compiled our programs with /FLOAT=G_FLOAT, the native VAX floating-point format.
- A program built with Oracle Pro*C 9.2.0.3.0 and VAX floating-point gives correct results.
- A program built with Oracle Pro*C 10.2.0.5.0 and VAX floating-point gives invalid results. We got errors and NaNs for the same programs reading the same data.
- A program built with Oracle Pro*C 10.2.0.5.0 and IEEE floating-point gives correct data results.
From the results it appears that the Oracle Pro*C 10.2.x pre-compiler requires that host variables use the IEEE floating-point format.
What do others think? Is there any Oracle documentation on this?