Re: [PATCH 3/4] Use consistent extensions for generated header files.

[Akim Demaille]

Le 2 nov. 08 à 20:16, Akim Demaille a écrit :

> Le 2 nov. 08 à 19:06, Joel E. Denny a écrit :
>
> > Thanks.  Sorry to be so contradictory, but I'm really ready to be  
> > done
> > with 2.4.
>
> Maybe we should decide soon enough what 2.5 is about, and maybe even  
> schedule a 2.6 already.  Using branches it should be easy enough.
>
> As far as I'm concerned I wish to see the following changes in 2.5:
> - overhaul of lalr1.cc including variants and fusion of the three  
> stack into one
> - make_SYMBOL in all the skeletons
> - propagate some simplification pioneered by lalrl1.cc into the  
> other skeletons (e.g., macros to define the tables, macros for  
> symbol_value, removal of a useless table etc.)
> - %define token.prefix in all the skeletons (whatever the name we  
> finally choose)
>
> Later (2.6?) I would like to play with
> - fusion of the three stacks in the remaining skeletons (especially  
> yacc.c)

To make my feeling clearer, I would really like to see really  
experimental stuff *not* go into master until it's quasi ready for  
distribution.  Bison has suffered many times from infrequent releases  
because of unfinished features, which resulted in mature feature being  
held way too long.

I was clearly part of this.

But now that we have Git, it should be much easier to have candidate  
branches that we merge when ready.  The fact that they are not part of  
master will certainly help their authors focus on the last bits (say,  
documentation, tests and NEWS :).

But then (how unpleasant...) I would really like to be able to flush  
my patch asap...