Re: Dubious features

[Akim Demaille]

>>> "Joel" == Joel E Denny <address@hidden> writes:

 > On Wed, 7 Jun 2006, Akim Demaille wrote:
 >> In the course of splitting the scanner, so that we have a more
 >> state-less scanner and therefore more freedom in the parser, there are
 >> two "features" that I would like to discuss, because I don't think we
 >> should keep them (as are).
 >> 
 >> - Adding a trailing semi-colon
 >> - The "arbitrary C-code after %union".

 > I don't understand the need for either feature in any form.  I don't use 
 > them, but I have no feel for what others do.

I'm responsible for the second one, because that's how I used to trick
Bison to produce OK C++ code.  Now that we have decent C++ skeletons,
there is no point for it.  What annoys me is that it is ugly from the
Bison grammar point of view, and results in a nasty scanner.  I have
always used this "feature" (which at the time was rather an exploit
from a loophole in the parser) with a single id: the name I wanted to
give to the union.

As for the first point, I had received complaints from users who
disliked the fact that they faced errors with yacc: they wanted bison
to be more yackish.  Which it does today with -y, but still, I think
this is a misfeature.  It's probably going to be even worse when we
add more output languages, where adding a `;' might be wrong.