Am Do., 19. Sept. 2024 um 21:32 Uhr schrieb Wolfgang Corcoran-Mathe < [email protected]>:
> Unless someone objects, I’m going to remove with-abort-restarter. The > semantics aren’t right for something named “abort”, since it installs a > restarter that invokes the continuation of the w-a-r form rather than > completely aborting the computation (which is the proposed meaning of > an abort restarter). This can be very confusing, since the restarter is > only an “abort” if the w-a-r form encompasses the entire computation. > "Abort" does not necessarily have to mean "abort to the REPL or the OS". The terminology is also used when an abort to some continuation prompt happens, which is directly comparable to the case here. Maybe some extra description should be given to the form so that it can clearly prompt to which context the computation is aborted. I agree that "Abort!" without any further info is misleading. > On the other hand, with-abort-restarter could be respecified to install > a “true” abort restarter, i.e. one which returns to the REPL (or > something similar). The semantics of this would be strange, since the > restarter would abort not only the evaluation of *thunk*, but of the > entire computation containing the w-a-r form. This is a bit > “anti-lexical”, to my mind. > We are talking about a dynamic property, so I am not sure whether the term "lexical" fits here. But in any case, such a thing is not needed (and cannot be specified in a portable anyway). If an exception is not caught, one usually ends up in a debugger. Marc
