Re: [PATCH] b4_args is more general than only C++.

[Akim Demaille]

>>> "EB" == Eric Blake <address@hidden> writes:

 >> +2008-11-25  Akim Demaille  <address@hidden>
 >> +
 >> +   b4_args is more general than only C++.
 >> +   * data/lalr1.cc (b4_args, _b4_args): Move to...
 >> +   * data/bison.m4: here.

 >> +m4_define([b4_args],
 >> +[m4_if([$#$1],
 >> +       [1], [],
 >> +       [m4_ifval([$1],
 >> +                 [$1[]_$0(m4_shift($@))],
 >> +                 [$0(m4_shift($@))])])])

 > This implementation is quadratic on M4 1.4.x.  Wouldn't you rather
 > use a linear implementation?  Recently, there has been a lot of
 > churn in autoconf.git's copy of m4sugar making it even faster;
 > it is probably time for me to post a resync patch again.  But one
 > of those improvements is the addition of m4_map_args_sep; the
 > following is a linear implementation (with less text, to boot):

 > m4_define([b4_args],
 > [m4_shift(m4_map_args_sep([_$0(], [)], [], $@))])

 > m4_define([_b4_args], [m4_ifval([$1], [, $1])])

There is no m4_map_args_sep in Bison.  I don't know what's the status
bw m4sugar in Autoconf and in Bison.  Did they finally converge?  In
which case we should probably add autoconf.git as a submodule and deal
with bootstrap to import the appropriate files in bison.  That would
make it easier to keep them synch'ed.