Invalid (?) assumption in freeinverse.gi that an Iterator for a free group has !.word bound

[wilfwilson]

Currently, in the stable-4.10 branch of GAP, the tst/standard/freeinverse.tst file fails when all packages are loaded. This is because (due to very long-standing a bug in GAP) a method for Iterator in RCWA that applies to certain groups ends up having higher rank than the special GAP library Iterator method for free groups.

The code in Iterator for a free inverse semigroup assumes that Iterator for a free group has a component !.word, and this presumably records something. I don't see this documented anywhere, so we probably shouldn't rely on this behaviour to exist forever. The test file currently fails because the Iterator for a free group returned by the RCWA method does not have a !.word component.

In the first instance, I will attempt to fix this by adding SUM_FLAGS to the GAP library method for Iterator, even though this won't be required any more when the master branch of GAP ends up being the version required by Semigroups.

In the second instance, we should try to make this code more robust.

[james-d-mitchell]

Just to mention that the most important point here is that we should not depend on the implementational details of free groups but rather on their documented public methods.

[wilfwilson]

Resolved by #540.

[james-d-mitchell]

Resolved in v3.0.20.