On Monday, 18 July 2016 14:43:03 CEST Richard W.M. Jones wrote: > > > - let validate_key key = > > > - if String.length key == 0 || key == "-" || key == "--" > > > - || key. != '-' then > > > - invalid_arg (sprintf "invalid option key: '%s'" key) > > > + let validate_key = function > > > + | L"" -> invalid_arg "Getopt spec: invalid empty long option" > > > + | L"help" -> invalid_arg "Getopt spec: should not have L\"help\"" > > > > Theoretically both Arg and the current Getopt allow applications to > > provide an own handler for --help, instead of the built-in one. > > Sure, but I don't think that's a good idea :-) Well, I could have avoided that when proposing the final Getopt patch... Also, there should be the same checks for M"something" of the L"something" ones. > > IMHO it'd be better to sort the specs at this point, like done before; > > otherwise, --help (and potentially any non-hidden built-in option added > > here) will be shown only at the end of the other specs. > > At the beginning of the list, and it was deliberate. However it's > just a matter of preference, and the options could be sorted later. Well I'd prefer it sorted, since it won't create arbitrary lines of --help with different order, which look a bit odd. Also, making everything ordered was basically the reason for the helper type t, and the mutable state of its specs field. -- Pino Toscano
Description: This is a digitally signed message part.