Thanks, yes, the SWI version I tried was Fabrizio Riguzzi's one. It's possible the use of false/0 was indeed the problem because IIRC I just kept getting silent failures.

It's funny there's both fail/0 and false/0 in ISO (though not sure which version) but only true/0 and no succeed/0. My guess is that an "I-mean-false" predicate is useful (e.g. for failure-driven loops) but "fail" strikes some folks as too procedural so we also got "false". But I'm only guessing.

I'd like to blame Aleph for using "false:-" for negative examples, rather than ISO. I suspect it's a hack to make it easier to cause negative examples to fail, somewhere. Normally ":- a" should be immediately recognised as a goal with a negative literal, but since that's the syntax used in directives it's probably not immediately obvious to most that "a" is negated. It wasn't to me for the first 10 years or so.