Hi Paul, On Tue, 25 Jan 2011, Paul Eggert wrote:
> On 01/24/2011 07:47 PM, Joel E. Denny wrote: > > Named references are experimental, and > > omitting semicolons is deprecated, so I don't see any pressing need to > > address this now. > > OK, but this would mean that we would allow grammars that have > named references but omit semicolons, and this would mean > trouble later if we start rejecting such grammars. I disagree. Named references are labeled experimental. That means "use at your own risk" until we remove the label. Thus, forward compatibility is not a major concern if at all. On top of that, omitting semicolons is deprecated. Users who combine the two should definitely not expect forward compatibility. > Hmm, how about if we work around the problem for now, > by refusing to accept such grammars? That is, we have one > variable keep track whether named references are used, and another > to keep track whether any semicolons are omitted, and exit > with nonzero status if both variables are set. That would > be a small change to what we have now, and would allow us > freedom of implementation later. As long as it's simple and doesn't make this area even more difficult to read, I suppose there's no harm in something like that. Even so, would you mind waiting until we've resolved the previous issues? I realize Bison development is slow right now. It just so happens that everyone who was involved in the development of 2.5 is very busy at the moment.