On Tue, 2002-09-17 at 13:54, Norman Walsh wrote: > / Norman Walsh <[EMAIL PROTECTED]> was heard to say: > | The TC proposes to solve RFE #582822 by changing the funcprototype > | in the following backwards-compatible way: > | > | funcprototype ::= > | (funcdef, > | (void|(paramdef*, varargs?))) > > Make that > > funcprototype ::= > (funcdef, > (void|varargs|(paramdef+, varargs?))) > > To avoid a funcprototype containing only a funcdef. >
<snip> Since I am the submitter of the RFE, I suppose I ought to at least be publicly on the record as supporting this change. Currently, I am documenting a rather large "C" api and am unable to represent the concept of repetitive optional arguments in a calling sequence. It seems unnatural to use &hellip inside <paramdef> to get the printed output correct when with the proper enhancement to <funcdef>, one could specify <varargs>. Additionally, <varargs> needs to appear at any point in the <funcdef> as the argument list can be in any arbitrary sequence according to the needs of the api. Thanks. John
