Re: Fix a couple of links in standards.texi

    It's preferable to link to up to date sources.  

At one time I was under the impression that the CVS repos mirrored the
git repos and were up to date, just read only.  I'm still not sure if
that's never the case, sometimes the case, or always the case.

Anyway, whatever.

Thanks,
k

Re: [PATCH] PR bootstrap/35169: HP-UX/IA awk bug

On Mon, Mar 31, 2008 at 11:18:15PM +0200, Ralf Wildenhues wrote:
> GCC PR 35169 uncovers an awk printf %u bug on ia64-hp-hpux11.23 not
> present on HP-UX/PA.  I confirmed that the patch against GCC produces
> a correct options.c with its awk, and of course with GNU awk.
> OK for trunk, 4.3, 4.2, and 4.1?

Hi Ralf,

Thanks for looking into this, we were going to get to it later, but
once we found a workaround (using GNU awk), it was hardly a priority.

> Below is also a patch for the portability section of the Autoconf
> manual.  OK to apply, and who should be credited?

We can't approve. As for credit, while we figured out that it was an awk
bug, we did not go any further than that, feel free to take credit,
you went the extra mile and discovered the real problem.

> 
> Can the bug reporter please file a trouble ticket with HP so they can
> fix this?  Also, could you bootstrap GCC on ia64-hp-hpux11.23 with the
> patch applied?

I'll ask Albert to file the bug. Bootstrap is underway.

Thanks,
Peter 

Re: [PATCH] PR bootstrap/35169: HP-UX/IA awk bug

According to Ralf Wildenhues on 3/31/2008 3:18 PM:
| Below is also a patch for the portability section of the Autoconf
| manual.  OK to apply, and who should be credited?

Please commit the autoconf portion of the patch, although it would be nice
if we knew who to credit first.

| 	* doc/autoconf.texi (Limitations of Usual Tools): Mention awk %u
| 	bug on HP-UX/IA.
| 	Report by ...
|

--
Don't work too hard, make some time for fun as well!

Eric Blake             ebb9 <at> byu.net

Re: Stack-like AC_LANG_WERROR

Hello Ralf,

On Wed, Mar 26, 2008 at 12:00:49AM +0100, Ralf Wildenhues wrote:
> 
> I would probably like a shell function version better.  But more
> importantly, this patch needs a testsuite addition to exercise the new
> (and the old! claiming compatibility) behavior.  And a NEWS entry.

[skip]

> I wonder a bit whether the functionality should be in two new macros
>   AC_LANG_WERROR_PUSH(on/off)
>   AC_LANG_WERROR_POP([on/off])
> 
> where the optional argument to POP would allow for a consistency check
> with the one passed to the previous PUSH.  This would be nicely similar
> to AC_LANG_{PUSH,POP}.

This is another backward compatible modification for AC_LANG_WERROR:

1. AC_LANG_WERROR([on/off])
  a) no arguments -- similarly to AC_LANG_WERROR([on]);
  b) `on'  -- treat warnings as fatal errors;
  c) `off' -- do not treat warnings as fatal errors.

2. AC_LANG_WERROR_PUSH(on/off)
  a) `on'  -- remember current settings + treat warnings as fatal errors;
  b) `off' -- remember current settings + do no treat warnings as fatal
     errors.

Recommend M4 1.4.11

The release of M4 1.4.11 was one of the last show-stoppers for Autoconf 2.62.  
I'm committing this, then starting my efforts on preparing the release; 
hopefully I will have the official tarball uploaded within the next 48 hours or 
so.  The release will be GPLv2+ plus exception (ie. unchanged from 2.61); 
hopefully, the FSF lawyers will resolve the holdups that prevented this release 
from being GPLv3+ plus exception in time for the next release.

As a reminder, Autoconf 2.62 will work out of the box with the planned M4 1.6, 
but unpatched Autoconf 2.59 will not, due to a planned change in semantics of 
the m4wrap builtin in M4 1.6 to comply with POSIX.

From: Eric Blake <ebb9 <at> byu.net>
Date: Wed, 2 Apr 2008 13:21:33 -0600
Subject: [PATCH] Recommend the just-released M4 1.4.11.

* NEWS: Update recommendation.
* README: Likewise.
* doc/autoconf.texi (Introduction): Likewise.
* m4/m4.m4 (AC_PROG_GNU_M4): Likewise.

Signed-off-by: Eric Blake <ebb9 <at> byu.net>
---
 ChangeLog         |    8 ++++++++
 NEWS              |    2 +-
 README            |    4 ++--
 doc/autoconf.texi |    2 +-
 m4/m4.m4          |    6 +++---
 5 files changed, 15 insertions(+), 7 deletions(-)

diff --git a/ChangeLog b/ChangeLog

upstream file sync

There still isn't a nice rule to automatically update files borrowed from
other locations (ie. 'make update' is currently broken), but I figured a
manual file sync prior to the release was better than nothing.  Coreutils'
maint.mk is still not fully synced, but is closer.  In the process of
syncing, I fixed a few things to once again pass 'make syntax-check'.  I'm
installing this big patch:

--
Don't work too hard, make some time for fun as well!

Eric Blake             ebb9 <at> byu.net

From 991e8cc0ca82045e91a4a639bfa259907912bcc8 Mon Sep 17 00:00:00 2001
From: Eric Blake <ebb9 <at> byu.net>
Date: Thu, 3 Apr 2008 06:22:48 -0600
Subject: [PATCH] Sync files from upstream, and pass 'make syntax-check'.

* config/announce-gen: Move...
* build-aux/announce-gen: ...here, and sync from gnulib.
* Makefile.am (EXTRA_DIST): Adjust accordingly.
* cfg.mk (announce_gen): Likewise.
(prev_version_file): Delete, relying on default in maint.mk.
(gpg_key_ID): New macro.
(url_dir_list): Rewrite to match coreutils.
* config/prev-version.txt: Move...
* .prev-version: ...here, and adjust to 2.61.
* build-aux/vc-list-files: Sync from coreutils.
* maint.mk: Resynchronize with coreutils, where possible.

Re: upstream file sync

Eric Blake <ebb9 <at> byu.net> writes:

> There still isn't a nice rule to automatically update files borrowed from
> other locations (ie. 'make update' is currently broken), but I figured a
> manual file sync prior to the release was better than nothing.  Coreutils'
> maint.mk is still not fully synced, but is closer.  In the process of
> syncing, I fixed a few things to once again pass 'make syntax-check'.

And another round of syncs, this time so that a VPATH 'make syntax-check' also 
passes.

From: Eric Blake <ebb9 <at> byu.net>
Date: Thu, 3 Apr 2008 08:31:37 -0600
Subject: [PATCH] Fix VPATH 'make syntax-check'.

* maint.mk (VC_LIST, VC_LIST_EXCEPT, sc_changelog)
(sc_prohibit_jm_in_m4, makefile-check): Support VPATH.
* build-aux/vc-list-files: Sync from upstream again.
* build-aux/texinfo.tex: Likewise.

Signed-off-by: Eric Blake <ebb9 <at> byu.net>
---
 ChangeLog               |    7 +++++
 build-aux/texinfo.tex   |   16 ++++++------
 build-aux/vc-list-files |   63 ++++++++++++++++++++++++++++++++++++++++------
 maint.mk                |   20 +++++++++-----
 4 files changed, 82 insertions(+), 24 deletions(-)

diff --git a/ChangeLog b/ChangeLog
index 8629f93..76fa890 100644

Re: upstream file sync

Eric Blake <ebb9 <at> byu.net> writes:

> And another round of syncs, this time so that a VPATH 'make syntax-check' 
also 
> passes.

More updates, so that in-place 'make maintainer-distcheck' passes.  (Boy, vc-
list-files is getting a lot of churn today.  I'm thinking that post-release, it 
would be nice to start using gnulib-tool as a prerequisite strap to 
bootstrapping autoconf; by removing upstream files from autoconf.git, it cuts 
down on the churn).

The package.m4 change is unique to autoconf - even though our manual documents 
that autotest users should distribute both `testsuite' and `package.m4', that 
is because users don't necessarily have m4 installed, and they can't regenerate 
`testsuite' if the package omits their distribution.  But autoconf users DO 
have m4 installed, and distributing files that can be rebuilt during 'make' is 
pointless.

From: Eric Blake <ebb9 <at> byu.net>
Date: Thu, 3 Apr 2008 11:02:50 -0600
Subject: [PATCH] More maintainer tweaks: pass 'make maintainer-distcheck'.

* GNUmakefile (_is-dist-target): Sync from upstream.
* build-aux/vc-list-files: Sync from upstream, yet again.
* tests/atlocal.in (unsupported_fs_chars): Always remove tdir.
* tests/Makefile.am (EXTRA_DIST): Don't distribute the built
package.m4.
* lib/autoconf/general.m4 (_AC_INIT_COPYRIGHT): Bump year.

Re: upstream file sync

Eric Blake <ebb9 <at> byu.net> writes:

> > And another round of syncs, this time so that a VPATH 'make syntax-check' 
> also 
> > passes.
> 

Next round.  This time, so that the man pages have a consistent version string 
in them.

From 1fda5799d3763286e53673bb1a0174c6c3518a6c Mon Sep 17 00:00:00 2001
From: Eric Blake <ebb9 <at> byu.net>
Date: Thu, 3 Apr 2008 16:45:11 -0600
Subject: [PATCH] Fix version number generation in man pages.

* Makefile.am (EXTRA_DIST): Distribute .version.
(.version): New rule.
* man/Makefile.am (common_dep): Depend on .version, not
configure.ac.
(.x.1): Use package name for version string.
* GNUmakefile [!_have-Makefile]: Sync from upstream, again.
* build-aux/git-version-gen: Sync from upstream.

Signed-off-by: Eric Blake <ebb9 <at> byu.net>
---
 ChangeLog                 |    9 ++++++++
 GNUmakefile               |    8 ++++++-
 Makefile.am               |   20 ++++++++++++++++++-
 build-aux/git-version-gen |   46 +++++++++++++++++++++++++++++++++++++++++++-
 man/Makefile.am           |    7 +++--

carriage return line endings vs. literal ^M in status.m4

Hi,

I've been working through some odd problems related to a user's
attempt to build recent coreutils snapshots on a Maemo/OS2008 system.
So far I've hit two problems:
The first affected coreutils only (configure.in):
The second (discovered minutes ago) affects the very latest autoconf:

For context, here's the coreutils patch:

	Accommodate building on Maemo
	* configure.in: Also filter out carriage returns from the value
	of $no_install_progs_default.  Reported by Elbert Pol.

diff --git a/configure.ac b/configure.ac
index 0e101a1..36fc0cf 100644
--- a/configure.ac
+++ b/configure.ac
 <at>  <at>  -268,7 +268,7  <at>  <at>  done
 v=no_install__progs
 t=`sed -n '/^'$v' =/,/[[^\]]$/p' $mk \
     | sed 's/^  *//;/^\$.*/d;/^'$v' =/d' \
-    | tr -s '\\012\\\\' '  '`
+    | tr -s '\\015\\012\\\\' '   '`
 # Remove any trailing space.
 no_install_progs_default=`echo "$t"|sed 's/ $//'`

--
1.5.5.rc3.1.gaece