yacc/bison

[Markus Weissmann]

On 8 Apr 2009, at 06:45, Ryan Schmidt wrote:

> On Apr 7, 2009, at 09:59, Jeremy Lavergne wrote:
>
>> I have a port that is using yacc.  Like g++/gcc, it can use the  
>> system ones without a problem.  Should this not be marked as a  
>> dependency then?
>
> I suggest declaring a dependency on the bison port. I do this in my  
> pure/pure-devel ports which need bison. This way, you won't be  
> confused later when differences arise between the system bison and  
> the MacPorts bison; you can be assured all users are using the  
> MacPorts bison.
>
>> If not, what ports provide bin/yacc besides bison +yacc?  It is my  
>> understanding we can't dictate variants yet, so I'd like a better  
>> solution than printing a ui_msg.
>
> True. But:
>
> The +yacc variant doesn't make bison take longer to build [1]. It  
> doesn't add any dependencies. All it does is install a static  
> library liby.a and install a wrapper script "yacc" which calls  
> "bison -y".
>

How about setting YACC='bison -y' ? Or does the port require the  
library as well?


> The +yacc variant was added to the bison port in r10216 when it was  
> updated to version 2.0. No explanation why it was a variant instead  
> of always on.
>

If in doubt, I try to mimic the FreeBSD ports -- thats tried and  
tested already for the "closest relative" OS. I recall there were some  
ports that failed due to expecting "bison1" behind 'yacc'..


> This sounds like a great example of where we should remove the  
> variant and have bison always include yacc support.
>

Thinking about enabling 'yacc' by default, there are (at least) 3  
candidates to claim the 'yacc' executable: bison1, bison (version 2),  
byaccj and bisoncpp [1];

Another solution would be to install the yacc/bison2 executable as  
'yacc2' or as 'yacc' in $prefix/lib/bison/bin -- thought?


Regards,

-Markus


[1] http://bisoncpp.sourceforge.net/

--
Markus W. Weissmann, M. Sc.
http://www.mweissmann.de/