coffee@aero.ARPA (Peter C. Coffee) (05/06/87)
munchmunchmunchmunchmunch......... Can anyone suggest a reason why the Forgy Common Lisp OPS5, with gaps in implementation filled so that it will run under Gold Hill's GC 286 Developer on an AT-class machine, should load and compile correctly on Gold Hill's 80386-based "Hummingboard" (with accompanying "GC 386" Lisp) but should give incorrect results? Specifically, the commands for inspecting memory show the rules -- apparently -- loading correctly, but pattern matching isn't happening. The Monkeys-and-Bananas benchmark, for example, runs the initializing T1 rule correctly, then says that the least specific matching rule -- MB19, "There is an active goal," which is _never_ supposed to fire -- is the only rule that matches. Leaving MB19 out of the benchmark source results in the system saying that no rules match. This occurs with both interpreted and compiled code, both of which work correctly in every respect on the 286. Suggestions welcome -- Gold Hill, are you listening? Cheers, Pc^2