Re: proposed patch for Bison core dumps when using lalr1.cc

[Akim Demaille]

 >> From: Akim Demaille <address@hidden>
 >> Date: Sat, 01 Feb 2003 17:11:24 +0100

 >> This patch is wrong: the header is not optional.  M4 should die if the
 >> value is not defined.  I'll do that when I have time.

 Paul> OK, thanks.  I should warn you that I installed the patch in 1.875a,
 Paul> but I guess the only downside my jumping the gun should be the
 Paul> substitution of one wrong behavior (whatever it is) for another (a
 Paul> core dump).

I know.

 Paul> However, I'm puzzled: why is the header required for C++?

Because the code is designed this way.  Actually, the real question is
why is it not in C.

 Paul> Isn't C++ just the same as C with respect to whether a header
 Paul> is needed?  Are you saying that a C++ user must invoke Bison
 Paul> with the -d option?

Nope.  IIRC that before some changes, the M4 macro defined the header
name was always defined, but only if Boolean flag was changed when -d
was used or not.  Lalr1.cc just didn't pay attention to -d.  But now
the code is different.