NEWS entry order

Is NEWS sorted alphabetically or chronologically, inside one version
section?

Anyways, the AH_HEADER was misappropriated to pre-2.59b instead of post.

Cheers,
Ralf

        * NEWS: Move AH_HEADER mention to right place.

Index: NEWS
===================================================================
RCS file: /cvsroot/autoconf/autoconf/NEWS,v
retrieving revision 1.348
diff -u -r1.348 NEWS
--- NEWS	20 Oct 2005 06:13:34 -0000	1.348
+++ NEWS	28 Oct 2005 13:04:10 -0000
 <at>  <at>  -60,7 +60,11  <at>  <at> 
 ** AT_COPYRIGHT
   New macro for copyright notices in testsuite files.

+** AH_HEADER
+  New macro which is defined to the name of the first declared config header
+  or undefined if no config headers have been declared yet.
+
 ** ALLOCA, LIBOBJS, LTLIBOBJS
   Object names added to these variables are now prefixed with `${LIBOBJDIR}',
   as in `${LIBOBJDIR}alloca.o'.  LIBOBJDIR is meant to be defined from
   `Makefile.in' in case the object files lie in a different directory.
 <at>  <at>  -97,10 +104,6  <at>  <at> 

Re: NEWS entry order

Hello,

On Tue, Nov 01, 2005 at 07:17:11AM +0100, Ralf Wildenhues wrote:
>         * NEWS: Move AH_HEADER mention to right place.

comitted, thanks for your patch.

> Is NEWS sorted alphabetically or chronologically, inside one version
> section?

It seems that it is sorted alphabetically.

But the 2.59b section doesn't seem to be sorted.
AC_PROG_GREP is before AC_PROG_EGREP because the latter mentions the former.
Similarly, it seems logical to place *_TARGET_TOOL macros after *_TOOL.

Have a nice day,
	Stepan Kasal

Re: AC_PROG_CC_C_O doesn't work with VC++

Hello,

On Thu, Oct 27, 2005 at 12:09:22PM -0700, Paul Eggert wrote:
> If memory serves the name change from conftst2.c to conftest2.c means
> we don't have to worry about cleaning the files up separately, since
> the already-existing "rm -f conftest*" will clean them up.

yes, the previous patch had to change ac_clean_files.

> [...] we shouldn't clutter Autoconf up with comments about 8+3, [...]

OK, deleted that comment, see the attached patch.

> While we're on that subject, there may still be reasons to live within
> the 14-byte file name length limit of original Unix, [...]

Does this limitation affect the Autoconf somehow?

A quick grep revealed that the 14 byte limit was the reson for
AM_MAKEINFOFLAGS = --no-split
in doc/Makefile.am.  The there is AC_SYS_LONG_FILE_NAMES, of course.

I don't see any need for any change.

Stepan

2005-11-01  Stepan Kasal  <kasal <at> ucw.cz>

	* lib/autoconf/c.m4 (AC_PROG_CC_C_O): Remove the comment about 8+3

AC_CHECK_TOOL AC_SUBSTs more than necessary

Hi there,

The AC_CHECK_TOOL/AC_PATH_TOOL and similar macros cause variables of the
form
  ac_ct_*
  ac_pt_*
to be AC_SUBSTed.  This looks unnecessary to me: the variables are not
documented, neither are they used outside configuration; they are merely
helper variables to avoid conflicts with the cache variables.

If you agree with me, I suggest the following patch to fix this.

Cheers,
Ralf

	* lib/autoconf/programs.m4 (AC_CHECK_PROG, AC_PATH_PROG):
	Factor functionality to..
	(_AC_CHECK_PROG, _AC_PATH_PROG): these new macros, but only
	`AC_SUBST($1)' in the public version.
	(AC_CHECK_TOOL, AC_PATH_TOOL, AC_PATH_TARGET_TOOL)
	(AC_CHECK_TARGET_TOOL): Use internal versions for ac_ct_* and
	ac_pt_* variables.

Index: lib/autoconf/programs.m4
===================================================================
RCS file: /cvsroot/autoconf/autoconf/lib/autoconf/programs.m4,v
retrieving revision 1.46
diff -u -r1.46 programs.m4
--- lib/autoconf/programs.m4	20 Jun 2005 14:48:14 -0000	1.46
+++ lib/autoconf/programs.m4	5 Nov 2005 06:25:22 -0000

Re: AC_CHECK_TOOL AC_SUBSTs more than necessary

That change looks good to me.  Can you please install it?
(I've added you to the autoconf group, if you don't mind.)

Re: AC_CHECK_TOOL AC_SUBSTs more than necessary

Hi Paul,

* Paul Eggert wrote on Sun, Nov 06, 2005 at 09:52:19AM CET:
> That change looks good to me.  Can you please install it?
> (I've added you to the autoconf group, if you don't mind.)

Thanks, I feel honored.  

However, there seems some permission bit missing still:

| $ cvs -t -z3 -d rwild <at> savannah.gnu.org:/cvsroot/autoconf co autoconf
|   -> parse_cvsroot ( rwild <at> savannah.gnu.org:/cvsroot/autoconf )
|   -> main loop with CVSROOT=rwild <at> savannah.gnu.org:/cvsroot/autoconf
|   -> safe_location( where=(null) )
|  -> Starting server: ssh -l rwild savannah.gnu.org cvs server
| access denied
| cvs [checkout aborted]: end of file from server (consult above messages if any)

But ssh access is ok (above with s/autoconf/libtool/g works, too):

| $ ssh -v rwild <at> cvs.savannah.gnu.org
[...]
| Last login: ...
[...]
| Direct shell access is not possible for security reason.
| CVS commands are allowed.

Note the hourly cron job to update account data should've propagated
by now.

Re: conftest.c should #include confdefs.h

On Wed, Sep 07, 2005 at 03:34:04PM +0200, Stepan Kasal wrote:
> 2005-09-07  Stepan Kasal  <kasal <at> ucw.cz>
> 
> 	* lib/autoconf/general.m4 (_AC_MSG_LOG_CONFTEST): Move...
> 	* lib/autoconf/lang.m4 (_AC_MSG_LOG_CONFTEST): ...here. If a language
> 	  specific definition is available, call it.  Otherwise use the
> 	  previous code.
> 	* lib/autoconf/c.m4 (_AC_MSG_LOG_CONFTEST(C), _AC_MSG_LOG_CONFTEST(C++)):
> 	  New macro, add confdefs.h to the log.
> 	(AC_LANG_SOURCE(C)): Do not copy confdefs.h, include it.

I do not see this patch in CVS.  It does look good to me.

Re: AC_CHECK_TOOL AC_SUBSTs more than necessary

* Ralf Wildenhues wrote on Sun, Nov 06, 2005 at 12:15:39PM CET:
> * Paul Eggert wrote on Sun, Nov 06, 2005 at 09:52:19AM CET:
> > That change looks good to me.  Can you please install it?
> > (I've added you to the autoconf group, if you don't mind.)
> 
> However, there seems some permission bit missing still:

Works now.  I've committed the patch.

Cheers,
Ralf

Re: conftest.c should #include confdefs.h

Noah Misch <noah <at> cs.caltech.edu> writes:

> On Wed, Sep 07, 2005 at 03:34:04PM +0200, Stepan Kasal wrote:
>> 2005-09-07  Stepan Kasal  <kasal <at> ucw.cz>
>> 
>> 	* lib/autoconf/general.m4 (_AC_MSG_LOG_CONFTEST): Move...
>> 	* lib/autoconf/lang.m4 (_AC_MSG_LOG_CONFTEST): ...here. If a language
>> 	  specific definition is available, call it.  Otherwise use the
>> 	  previous code.
>> 	* lib/autoconf/c.m4 (_AC_MSG_LOG_CONFTEST(C), _AC_MSG_LOG_CONFTEST(C++)):
>> 	  New macro, add confdefs.h to the log.
>> 	(AC_LANG_SOURCE(C)): Do not copy confdefs.h, include it.
>
> I do not see this patch in CVS.  It does look good to me.

Thanks for looking at it.  Stepan, can you please install it?

Re: uninitialized variables

[ Moving from libtool-patches ]

* Ralf Wildenhues wrote on Sun, Nov 06, 2005 at 07:37:48PM CET:
> 
> I went through the fun of
>   sh -u configure
> adding environment variables until I hit bugs.

1) Nasty race condition: Let's assume I have $tmp set in my environment,
or in configure.ac; the status.m4 code uses it at the end of configure
to name a directory which is removed later by a trap.  However, the trap
is installed before $tmp is initialized.  If a signal arrives too early,
it might remove rather arbitrary files.

        * lib/autoconf/status.m4 (_AC_OUTPUT_MAIN_LOOP): Initialize
        `tmp' to avoid file removal race.

Index: lib/autoconf/status.m4
===================================================================
RCS file: /cvsroot/autoconf/autoconf/lib/autoconf/status.m4,v
retrieving revision 1.77
diff -u -r1.77 status.m4
--- lib/autoconf/status.m4	21 Oct 2005 07:36:32 -0000	1.77
+++ lib/autoconf/status.m4	6 Nov 2005 19:41:17 -0000
 <at>  <at>  -1319,6 +1319,7  <at>  <at> 
 # Hook for its removal unless debugging.
 $debug ||
 {
+  tmp=
   trap 'exit_status=$?; rm -fr "$tmp" && exit $exit_status' 0