3.2.2 General Coverage Guidelines

As noted in Section 1.2, the ACATS strives for complete coverage of the Ada standard. However, complete coverage does not mean that every sentence in the standard has an associated test objective and test. Not all rules are testable; one important way to determine testability of a rule is to check its category. The next clause (3.2.3) will discuss how the category of a rule affects its mapping to test objectives.

There are also additional considerations that don't apply to specific rule categories.

Text sometimes contains definitions (especially in Static Semantics rules). Definitions are hard to handle, because they usually are not testable by themselves. The definition has to be used in another rule to make them visible. For instance, a categorization rule such as "something is either this, that, or fuzzy" cannot be tested by itself. "Something" has to be used in some other rule in order for the categorization to matter. That being the case, it usually makes sense to test the definition as part of the other rule(s). Such rules are marked as having a kind of "Widely Used" or "Subpart" in the coverage documentation. "Subpart" rules are tested as part of specific other objectives, while "Widely Used" rules are thought to be tested by many tests indirectly (and no attempt is made to verify that). In some cases, especially those that otherwise would have a combinatorial explosion, it may make more sense to test the definition as directly as possible (in which case objectives will be assigned to the definition).

Text sometimes includes sentences marked as redundant by the AARM. Such sentences should be given normatively elsewhere, and the testing should be done in the place where they are normatively given. The AARM often will indicate this place.

Combinational explosion is always a problem for visibility rules. Rules that define the scope or visibility of something should always be tested in place (even though other rules will need to be used to accomplish that).