Re: [updated PATCH] %language declaration

[Joel E. Denny]

On Fri, 8 Dec 2006, Joel E. Denny wrote:

> Paolo, I get the impression we've reached an understanding about %language 
> and why it isn't harmful.  In that case, I will soon try to find time to 
> review your patch in detail.

A few comments on the patch I found here:

  http://lists.gnu.org/archive/html/bison-patches/2006-11/msg00034.html

1. When there is a command-line flag and an equivalent grammar directive, 
the description of one in the manual usually cross-references the other.  
-L should probably reference %language, and the same for -S and %skeleton.  
Also, I noticed a typo in your documentation: "langauge".

2. The version number 2.3a+ is a temporary version number indicating 
collectively all stages of development between 2.3a and whatever we decide 
to call the next release (2.3b, 2.4, or 3.0).  We shouldn't reference 
2.3a+ in the manual because no version named 2.3a+ will ever be released, 
and I'm afraid we'd likely forget to change it before the next release.  
I think 2.3b is probably a safe bet.  You also mention 2.4 in the manual, 
and I wonder if that should be 2.3b as well.

3. I think we should document this %language feature as experimental in 
the manual and in NEWS.  I have no plans to change %language in a future 
release, but we should reserve the right for a little while just in case 
we find problems.

4. In language_argmatch, I believe your `if (!language)' is a fallacy 
since language is initialized upon allocation.  Also, why is `args' plural 
here and not in skeleton_arg?

5. In output.c, why do you have DEFINE (program_name)?  If this is 
intentional, please explain it in ChangeLog.

6. I think your new m4_fatal messages should follow the format of the 
existing ones in glr.cc and lalr1.cc.  By the way, can 
internationalization work with m4_fatal?

7. I seem to recall that Akim dislikes the ".tab" that appears in output 
file names.  I don't care for it, and Automake prefers to remove it.  Of 
course, we hang on to it anyway for backward-compatibility.  Since 
%language is new, is this perhaps a good opportunity to get rid of ".tab" 
for all languages?  It kinda makes sense: if you specify the target 
language, Bison will choose your output files' default names using a 
convention appropriate for the target language rather than using 
traditional Yacc/Bison conventions.  We probably need to check that this 
doesn't cause problems with Automake though.

Thanks.