Re: [updated PATCH] %language declaration

[Joel E. Denny]

On Sun, 17 Dec 2006, Paolo Bonzini wrote:

> > They're not yet relying on that behavior when using %language.
> 
> It is "as if" they were, if we want users to use %language "C++" instead of
> %skeleton "lalr1.cc" without additional headaches.

To me, it seems too minor to be viewed as a headache.  My guess is that 
current users of "lalr1.cc" will see it as a step forward, or they'll see 
%language as a bad move regardless of file name conventions.

Then again, that's just a guess.  I could be very wrong.  When it comes to 
something as subjective as this, I sometimes wonder if we should just take 
a poll on help-bison... maybe after a few more developers give their 
opinions first.

> > > if Bison
> > > started removing ".tab" *on its own*, Automake would break.
> > 
> > Automake's Yacc rules depend on Yacc's file name conventions not Bison's.
> > Thus, if you use Bison in Automake in this way, you specify -y.  So, -y
> > should probably override %language for the output file names.  Thanks for
> > mentioning this.
> 
> This starts getting too complicated to be "right".

It doesn't seem complicated to me.  The sole original documented purpose 
of -y was to create output files named y.tab.h and y.tab.c.  It makes 
sense that it should override the default file names regardless of what 
they are.

How many people are going to combine -y and %language anyway?  I'm 
thinking only Automake users will, and they'll get exactly the ".tab"-less 
file names promoted by %language without even needing to recognize that, 
in this case, it's actually Automake not %language that's doing it.  
It'll just work right.