Re: bug: inlining changes start state

[Adrian Quark <adrian-sablecc@xxxxxxxxxxxxxxxxxx>]

On Wed, 9 Jul 2008, Etienne M. Gagnon wrote:
> This is effectively quite an annoying bug. Have you tried, in the short term, 
> to reorder your productions.

I just tried modifying the inliner to not remove productions after inlining 
them, and that apparently resolved the problem.  It's a one-line change and I 
trust this won't hurt anything, since if the inlined production turns out to 
not be reachable from the start state, it won't appear in the generated parser.

Fortunately or unfortunately, it turns out that inlining wasn't helping my 
grammar after all.  By changing the initial state, the inliner was just masking 
ambiguities caused by the original start state, not resolving them.  If I 
modify my grammar to fix those ambiguities, it doesn't even need inlining any 
more.

> Instead, I'll try to get version 4 up and running ASAP. I know that it's not 
> visible on the web site, but I am actively (and almost exclusively) working 
> on it these days.

That sounds great. Will v4 offer directives for resolving shift/reduce 
conflicts? I am working on a language similar to Lua in that it allows 
expressions to appear in sequence with no separator, and I don't relish the 
prospect of maintaining two versions of the expression grammar when the 
conflict could be resolved by a standard "prefer shift" policy. I really like 
SableCC's architecture but this one issue is making me consider changing to 
CUP.

> PS: Could you (re-)enter your bug report using your existing sablecc.org 
> account on http://sablecc3.sablecc.org

Done. I closed the old report too.

Thanks for your assistance,
Adrian Quark