This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Bug: gcc/cp/parse.y & gcc/cp/decl.h: Conflict on TYPENAME


>>>>> "Zack" == Zack Weinberg <zack@codesourcery.com> writes:

Zack> Have you made any test releases of Bison that we could get and
Zack> experiment with?  

Nope.  I can provide an alpha if you want, but a lot of work is still
needed.

http://www.lrde.epita.fr/~akim/download/bison-1.49b.tar.gz

Zack> Incidentally, we have a couple of problems with defining a
Zack> custom yyoverflow hook with bison 1.35.  One is an unavoidable
Zack> warning:

Zack> bison.simple: In function `yyparse': bison.simple:924: warning:
Zack> label `yyoverflowlab' defined but not used

OK, I'll try to address this issue.  Do you thing 1.36 is needed?  I'm
fine with that if you think that can help.  But it will not include
the stuff that 1.49b caught in GCC.

Zack> The other is that YYSTYPE is not available in the %{ ... %}
Zack> region, so we cannot declare a variable with its proper type.
Zack> (See comments in gcc/c-parse.in for what I'm talking about.)

YYSTYPE is available *after* %union.  This is because some people have
a %{ %} before %union that define types in %union.  If you find some
more elegant scheme, I buy it!

BTW, do you mean pre 1.35 was better in this respect?

Zack> It would be nice if these could be addressed in a newer release.

Not soon, unfortunately.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]