On Tuesday 23 June 2009, Akim Demaille wrote: > I should have pointed out two changes that might help you: > > - %no-default-prec disables the automatic assignment of a default > precedence to rules. This is a nice feature, and I don't understand > today what I had cold feet about it. The feature is available in > shipped Bison, it is just not documented. See > http://lists.gnu.org/archive/html/bison-patches/2003-05/msg00014.html for > instance.
This would certainly be a partial solution to my problem. But it would make the operator precedence table harder to read and maintain. And since it's still not a complete solution, I don't think I'm going to use it for now. > - there is a new %precedence directive that allows completes %right, > %left and %nonassoc: it defines only a precedence level, and leaves > undecided associativity matters. This is not released, it will be in > Bison 2.6. See > http://lists.gnu.org/archive/html/bison-patches/2008-11/msg00125.html Kudos! It won't help with my current problem, but it's certainly a feature I will use. Bison 2.6 is gonna be great. I read that YYSTYPE will not be limited to integral/pointer types. That would greatly reduce complexity in my grammar, in which I now use C-pointers to Boost shared pointers. ;-) -- Michiel Helvensteijn m.helvenste...@gmail.com _______________________________________________ help-bison@gnu.org http://lists.gnu.org/mailman/listinfo/help-bison