tabulon
commented
5 years ago Can't agree more. Was thinking along the same lines, but would not dare do it, or even propose.
the "optspec()" part could also be divided into two sub-parts (2 & 3 below), resulting in 3 parts:
1) GoL non-regression, without optspec(...);
This is the first group in your comment.
2) GoL non-regression, with optspec(...);
Most of the optspec(...) test cases that exist today fall into this category
3) Any other kind of test for ensuring GLM is functioning correctly with regard to its own features, e.g. completion, HelpText, etc, ... This part is almost empty as of today
I'm thinking of splitting the tests into two parts.
The first part is anything that doesn't involve optspec(). In this part, we are just trying to make sure that GLM's behavior is identical with GL, fulfilling the doc's promise of being a "drop-in replacement".
The second part is anything that involves optspec().