Re: [Bug-gnulib] AC_PROG_YACC

[Derek Robert Price]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Let me start over - I missed a mistake Bruno made in his last email:

Derek Robert Price wrote:

> Bruno Haible wrote:
>
> >Derek Robert Price wrote:
>
> >>It appears that some `make's (notably some BSDs) have a default
> >>setting for YFLAGS (YFLAGS = -d) that some of my users are not happy
> >>with.  It breaks builds, notably with the GNULIB getdate package,
> >>which comes with its own getdate.h header (which gets overwritten by
> >>calls to `bison -y -d getdate.y'...).  Is this something that
> >>AC_PROG_YACC (or gl_BISON for the GNULIB folks) should be checking for
> >>and dealing with?
> >>
> >>My workaround will be to just stick a `YFLAGS=' line in my
> >>Makefile.am, but I thought you might like the bug report.
>
>
> >The automake documentation says:
> > "When `yacc' is invoked, it is passed `YFLAGS' and `AM_YFLAGS'.  The
> >  former is a user variable and the latter is intended for the
> >  `Makefile.am' author."
>
> >So this means, YFLAGS is meant as a variable through which the user can
> >influence the compilation (like CFLAGS, CPPFLAGS, LDFLAGS etc.). By
> >setting it to empty, you give the user less freedom.


Not true.  I am _defaulting_ it to empty by setting it via configure.
The user can still override it as an argument to configure or make
(`./configure YFLAGS=-d 'or `make YFLAGS=-d').

> >I'd say, since option '-d' is not appropriate in all cases, it's a
>
> mistake
>
> >to use it by default. The bug is therefore in the BSD make.
>
> I'm afraid I agree with you.  It might still make sense to have an
> autoconf workaround for the bug, when detected, however.


I still agree with you but I still think a workaround in macros is
appropriate, both GNULIB & Autoconf.  My new solution just AC_ARG_VARs
both YACC & YFLAGS, as CC, CFLAGS, CPPFLAGS, etc. are AC_ARG_VARed by
other autoconf macros and require the user override them as arguments
to configure or make.  Nobody seems to have a problem with autoconf
doing that for CC, CFLAGS, CPPFLAGS, etc.

Revision 3 of patch for GNULIB attached.  I think similar AC_ARG_VAR
statements for YACC & YFLAGS for AC_PROG_YACC would also be appropriate.

Derek
- --
                *8^)

Email: address@hidden

Get CVS support at <http://ximbiot.com>!
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFAloiILD1OTBfyMaQRAnpdAJ9gYR3HaJ3/1xQShsRonjFA125aQACgmcvK
IMI6fua+PN5LP/gQv420nvI=
=Ljfg
-----END PGP SIGNATURE-----

Index: ChangeLog
===================================================================
RCS file: /cvsroot/gnulib/gnulib/ChangeLog,v
retrieving revision 1.148
diff -u -p -r1.148 ChangeLog
--- ChangeLog   27 Apr 2004 22:11:37 -0000      1.148
+++ ChangeLog   3 May 2004 17:52:45 -0000
@@ -1,3 +1,8 @@
+2004-05-03  Derek Price  <address@hidden>
+
+       * m4/bison.m4: Use AM_MISSING_PROG rather than a set/subst combination.
+       AC_ARG_VAR() YACC & YFLAGS.
+
 2004-04-27  Derek Price  <address@hidden>
 
        * m4/dos.m4 (gl_AC_DOS): Add Cygwin to list of Windows/DOS
Index: m4/bison.m4
===================================================================
RCS file: /cvsroot/gnulib/gnulib/m4/bison.m4,v
retrieving revision 1.3
diff -u -p -r1.3 bison.m4
--- m4/bison.m4 18 Apr 2004 18:12:50 -0000      1.3
+++ m4/bison.m4 3 May 2004 17:52:45 -0000
@@ -3,6 +3,12 @@
 AC_DEFUN([gl_BISON],
 [
   # getdate.y works with bison only.
-  : ${YACC='bison -y'}
-  AC_SUBST(YACC)
+  AM_MISSING_PROG(YACC, bison -y)
+  AC_ARG_VAR(YACC,
+[The `Yet Another C Compiler' implementation to use.  Defaults to `bison -y'.
+Values other than `bison -y' will most likely break on most systems.])
+  AC_ARG_VAR(YFLAGS,
+[YFLAGS contains the list arguments that will be passed by default to Bison.
+This script will default YFLAGS to the empty string to avoid a default value of
+`-d' given by some make applications.])
 ])