Re: incorrect yychar for unambiguous GLR

[Joel E. Denny]

On Wed, 11 Jan 2006, Paul Hilfinger wrote:

> > I agree about yylookaheadStatuses.  However, yylookaheadAvailable sounds 
> > like it means yychar != YYEMPTY.  The polarity of yylookaheadUnused is 
> > confusing, and it sounds like it might indicate whether the lookahead has 
> > been shifted.  How about yylookaheadNeeds, which follows from my comments 
> > above its declaration and maintains the plural?  I'm open to other 
> > suggestions of course.
> 
> That one's fine.  

OK, I'll rename it.

> > I believe gcc likes the extra parentheses when you have an assignment used 
> > as a boolean expression.
> 
> ... and in this particular assignment, you didn't have that situation.

Sorry, the parentheses I thought were redundant were not.  I should have 
tested it before responding.

So, I have no idea why the truly redundant parentheses were there:

  (yystackp->yytops.yycapacity)

> Anyway, following your suggestion, I have submitted the following:

Thanks.

> > I believe the push to satisfy lint started with a user request.  It's all 
> > buried somewhere in the mailing lists, but Paul Eggert probably knows the 
> > history best.
> 
> Well, the GDB people have eliminated all uses at this point.  Forgive
> my blasphemy, but we don't have to respond to all requests.  Reporting
> unused function arguments is a dubious check at best, because unused
> arguments are common when a function header is dictated by interface
> requirements, in contrast to unused locals, which are harder to
> justify.  It's plausible to complain in C++, where argument names are
> optional for precisely this reason.

Personally, I've never cared about this issue except when YYUSE broke my 
code.  I'm happy to leave the debate to people who do, and I'll continue 
to mindlessly follow their lead when I contribute code to bison.  Paul 
Eggert and Akim Demaille seem to care more.

Joel