Re: How much testing required after CVTRPGSRC?

[Dan <dan27649@xxxxxxxxx>]

Interesting bug.  Fortunately, while the conversion report doesn't catch
this, the compile of the converted source will issue a severity=30 error,
causing the compile to fail.

- Dan

On Wed, Nov 4, 2015 at 7:31 AM, Wilson, Jonathan <piercing_male@xxxxxxxxxxx>
wrote:

> One "gotcha" I hit was the use of *like Defn *IN Afield. After
> conversion the Afield becomes a type in its own right, instead of
> behaving like a 1 char field. Moving constants, and constant fields, are
> type checked to be *on or *off (or equivalent) and the field also has a
> default auto-value of *off instead of, if I recall correctly, *blank (At
> the least in the programs being changed moving *blank in was allowed
> prior to conversion).
>
> Not much of a gotcha, but enough to cause problems when its usage was
> other than "as an indicator"... in the case of the programs I was
> working on the field was "tri-state" blank=no status (and/or not yet
> used/initialised) and then *on/*off as a status. (The actual test was
> used as *on=RollUp, *off=RollDown, and *blank=neither).
>
> While technically a "bug", the incorrect usage, the old compiler allowed
> it.