5:12 p.m.
configure.ac contains a check which fails if any other version
of autoconf than 2.59 is used:
dnl Due to a dependence on the implementation of certain internal autoconf
dnl macros, die if any version other than 2.59 is used.
AC_PREREQ(2.59)dnl
m4_if(m4_version_compare(m4_defn([m4_PACKAGE_VERSION]), [2.59]), 1,
[m4_fatal([This script can only be generated with autoconf 2.59])], [])dnl
To try whether using autoconf 2.60 works, I removed that check
and called
autoconf --force
Unfortunately, m4 runs into an endless loop then.
I have no idea yet why.
On an SXEmacs mailing list I found a patch related to autoconf 2.60:
http://www.sxemacs.org/list-archives/html/sxemacs-patches/2006-03/msg0004...
but this seems to fix an additional, it doesn't help for the endless
loop problem.
Do you have any ideas what the problem with autoconf 2.60 is?
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
5:53 a.m.
>>>> "Mike" == Mike FABIAN
<mfabian(a)suse.de> writes:
Mike> Do you have any ideas what the problem with autoconf 2.60
Mike> is?
No. To be honest, *personally* I don't care, and do not intend to
investigate it myself. We distribute configure itself; IMO developers
can reasonably be required to use autoconf 2.59.
That said, I'm quite embarrassed about the long delay between
submission of patches to make 2.5x work and our acceptance and release
of a configure.ac compatible with 2.5x. If you wish to find out
what's going on and submit a patch, *I* promise that this time I'll
make sure we give it timely attention.
Also, Malcolm Purvis and Jerry James might be more interested, but
Malcolm hasn't been terribly active in recent weeks, and Jerry has
declared that his time is limited. So maybe you'll get a nibble from
one of them, but for now at least I wanted to tell you that a 3rd
party patch will get prompt attention.
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
4:43 p.m.
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
That said, I'm quite embarrassed about the long delay between
submission of patches to make 2.5x work and our acceptance and release
of a configure.ac compatible with 2.5x. If you wish to find out
what's going on and submit a patch, *I* promise that this time I'll
make sure we give it timely attention.
Also, Malcolm Purvis and Jerry James might be more interested, but
Malcolm hasn't been terribly active in recent weeks, and Jerry has
declared that his time is limited. So maybe you'll get a nibble from
one of them, but for now at least I wanted to tell you that a 3rd
party patch will get prompt attention.
Andreas Schwab <schwab(a)suse.de> made a patch to make it work with
autoconf 2.60. It is attached.
-------------------------------------------------------------------
Tue Jul 18 19:08:53 CEST 2006 - schwab(a)suse.de
- Fix broken use of autoconf internal macros.
-------------------------------------------------------------------
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
3:03 a.m.
>>>> "Mike" == Mike FABIAN
<mfabian(a)suse.de> writes:
Mike> Andreas Schwab <schwab(a)suse.de> made a patch to make it work
Mike> with autoconf 2.60. It is attached.
Great! Thanks! And pass on our thanks to Andreas.
I'm tied up this weekend, but will be working on some autoconf issues
early next week, and will test this at the same time.
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
2:35 a.m.
New subject: autoconf 2.59 [was: 2.60] runs into an endless loop
>>>> "sjt" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
>>>> "Mike" == Mike FABIAN
<mfabian(a)suse.de> writes:
Mike> Andreas Schwab <schwab(a)suse.de> made a patch to make it work
Mike> with autoconf 2.60. It is attached.
sjt> Great! Thanks! And pass on our thanks to Andreas.
Unfortunately, Andreas's patch breaks 2.59 in the same way. Given the
way that loop termination test is written (not to mention autotools
history), it's unsurprising. I don't think you can win in both versions.
I spent a few minutes trying to find a way to conditionalize, but it's
quite unclear how to do so portably (seems at least in the documents
available to me m4_version_compare is undocumented, which means that
semantics may change at any time), so I'm punting.
My feeling is that we should avoid using a language that can't even
implement a fundamental primitive with the same semantics from one
minor version to the next, and go back to programming this kind of
thing in shell. Malcolm?
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
10:15 a.m.
New subject: autoconf 2.59 [was: 2.60] runs into an endless loop
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
>>>>> "sjt" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
>>>>> "Mike" == Mike FABIAN <mfabian(a)suse.de> writes:
Mike> Andreas Schwab <schwab(a)suse.de> made a patch to make it work
Mike> with autoconf 2.60. It is attached.
sjt> Great! Thanks! And pass on our thanks to Andreas.
Unfortunately, Andreas's patch breaks 2.59 in the same way.
I'm adding Andreas to the CC:. Maybe he can give us some insight.
Given the way that loop termination test is written (not to mention
autotools history), it's unsurprising. I don't think you can win in
both versions.
Strange, XEmacs builds here with Andreas's patch for older
versions of SUSE Linux as well, for example it builds for
SUSE Linux 10.0 which still has autoconf 2.59:
http://software.opensuse.org/download/M17N/SUSE_Linux_10.0/i586/xemacs-21...
http://software.opensuse.org/download/M17N/SUSE_Linux_10.0/i586/xemacs-el...
http://software.opensuse.org/download/M17N/SUSE_Linux_10.0/i586/xemacs-in...
http://software.opensuse.org/download/M17N/SUSE_Linux_10.0/src/xemacs-21....
There are also xemacs packages for SUSE Linux 10.1 in
http://software.opensuse.org/download/M17N/SUSE_Linux_10.1/ which have
been build with autoconf 2.59 as well.
And the packages in
http://software.opensuse.org/download/M17N/SUSE_Factory/
are for the next release of SUSE Linux (which will be called openSUSE
10.2). These are built with autoconf 2.60.
The sources and patches for all of these rpms are identical.
I spent a few minutes trying to find a way to conditionalize, but
it's
quite unclear how to do so portably (seems at least in the documents
available to me m4_version_compare is undocumented, which means that
semantics may change at any time), so I'm punting.
My feeling is that we should avoid using a language that can't even
implement a fundamental primitive with the same semantics from one
minor version to the next, and go back to programming this kind of
thing in shell. Malcolm?
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
3:02 p.m.
New subject: autoconf 2.59
>>>> "Mike" == Mike FABIAN
<mfabian(a)suse.de> writes:
> Given the way that loop termination test is written (not to
> mention autotools history), it's unsurprising. I don't think
> you can win in both versions.
Mike> Strange, XEmacs builds here with Andreas's patch for older
Mike> versions of SUSE Linux as well, for example it builds for
Mike> SUSE Linux 10.0 which still has autoconf 2.59:
Did you actually run autoconf? Did it take longer than usual?
Normally on my 1GHz PPC Mac "Panther" it takes about 20 seconds to run
autoconf 2.59; with Andreas' patch I killed the process after 15 min,
about 5 min CPU.
Mike> The sources and patches for all of these rpms are identical.
You're sure that configure gets rebuilt for every build?
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
3:59 p.m.
New subject: autoconf 2.59
>>>> "sjt" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
sjt> Did you actually run autoconf? Did it take longer than
sjt> usual? Normally on my 1GHz PPC Mac "Panther" it takes about
sjt> 20 seconds to run autoconf 2.59; with Andreas' patch I killed
sjt> the process after 15 min, about 5 min CPU.
Confirmed on Debian GNU/Linux 'sid', Pentium II 450MHz. Clock time
41s, CPU 34s for a successful run without the patch, killed without
finishing after 2m 17s, 1m 47s CPU, with the patch. Debian package
autoconf_2.59a-3_all.deb.
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
10:55 a.m.
New subject: autoconf 2.59
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
>>>>> "Mike" == Mike FABIAN
<mfabian(a)suse.de> writes:
>> Given the way that loop termination test is written (not to
>> mention autotools history), it's unsurprising. I don't think
>> you can win in both versions.
Mike> Strange, XEmacs builds here with Andreas's patch for older
Mike> versions of SUSE Linux as well, for example it builds for
Mike> SUSE Linux 10.0 which still has autoconf 2.59:
Did you actually run autoconf?
Definitely yes. I can see that in the build log (I attach the
build.log for SUSE Linux 10.0 here).
Extract from the build log:
[...]
installing autoconf-2.59-81
[...]
+ echo 'Patch #1 (xemacs-autoconf.patch):' <-- Andreas's
patch
Patch #1 (xemacs-autoconf.patch):
[...]
Executing(%build): /bin/sh -e /var/tmp/rpm-tmp.14003
+ umask 022
+ cd /usr/src/packages/BUILD
+ /bin/rm -rf /var/tmp/xemacs-21.5.27.20060705-build
++ dirname /var/tmp/xemacs-21.5.27.20060705-build
+ /bin/mkdir -p /var/tmp
+ /bin/mkdir /var/tmp/xemacs-21.5.27.20060705-build
+ cd xemacs-21.5.27.20060705
+ AUTOMAKE_DIR=/usr/share/automake
+ '[' -d /usr/share/automake ']'
+ AUTOMAKE_DIR='/usr/share/automake*'
+ '[' -d /usr/share/automake-1.9 ']'
+ for d in .
+ for f in config.sub config.guess
+ test -f ./config.sub -a '!' ./config.sub -ef
/usr/share/automake-1.9/config.sub
+ cp -f /usr/share/automake-1.9/config.sub ./config.sub
+ test -d . -a '!' -f ./depcomp -a -f /usr/share/automake-1.9/depcomp
+ cp -f /usr/share/automake-1.9/depcomp ./depcomp
+ echo 'please add depcomp to sources for new automake!'
please add depcomp to sources for new automake!
+ test -f ./missing -a '!' ./missing -ef /usr/share/automake-1.9/missing
+ for f in config.sub config.guess
+ test -f ./config.guess -a '!' ./config.guess -ef
/usr/share/automake-1.9/config.guess
+ cp -f /usr/share/automake-1.9/config.guess ./config.guess
+ test -d . -a '!' -f ./depcomp -a -f /usr/share/automake-1.9/depcomp
+ test -f ./missing -a '!' ./missing -ef /usr/share/automake-1.9/missing
+ for f in ltconfig ltmain.sh
+ test -f ./ltconfig
+ for f in ltconfig ltmain.sh
+ test -f ./ltmain.sh
+ autoconf --force
autoconf: warning: both `configure.ac' and `configure.in' are present.
autoconf: warning: proceeding with `configure.ac'.
[...]
Did it take longer than usual?
It did not take very long. This is on SUSE Linux 10.1-x86_64:
magellan:/usr/src/packages/BUILD/xemacs-21.5.27.20060705# uname -a
Linux magellan 2.6.16 #1 Wed May 31 06:52:41 UTC 2006 x86_64 x86_64 x86_64 GNU/Linux
magellan:/usr/src/packages/BUILD/xemacs-21.5.27.20060705# autoconf --version
autoconf (GNU Autoconf) 2.59
Written by David J. MacKenzie and Akim Demaille.
Copyright (C) 2003 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
magellan:/usr/src/packages/BUILD/xemacs-21.5.27.20060705# ll configure
-rwxr-xr-x 1 abuild abuild 1260079 2006-07-24 09:40 configure
magellan:/usr/src/packages/BUILD/xemacs-21.5.27.20060705# time autoconf --force
autoconf: warning: both `configure.ac' and `configure.in' are present.
autoconf: warning: proceeding with `configure.ac'.
real 0m7.489s
user 0m6.960s
sys 0m0.188s
magellan:/usr/src/packages/BUILD/xemacs-21.5.27.20060705# ll configure
-rwxr-xr-x 1 abuild abuild 1260079 2006-07-24 09:52 configure
magellan:/usr/src/packages/BUILD/xemacs-21.5.27.20060705#
Normally on my 1GHz PPC Mac "Panther" it takes about 20
seconds to run
autoconf 2.59; with Andreas' patch I killed the process after 15 min,
about 5 min CPU.
Mike> The sources and patches for all of these rpms are identical.
You're sure that configure gets rebuilt for every build?
Yes.
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
12:06 p.m.
New subject: autoconf 2.59
"Stephen J. Turnbull" <stephen(a)xemacs.org> さんは書きました:
>>>>> "sjt" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
sjt> Did you actually run autoconf? Did it take longer than
sjt> usual? Normally on my 1GHz PPC Mac "Panther" it takes about
sjt> 20 seconds to run autoconf 2.59; with Andreas' patch I killed
sjt> the process after 15 min, about 5 min CPU.
Confirmed on Debian GNU/Linux 'sid', Pentium II 450MHz. Clock time
41s, CPU 34s for a successful run without the patch, killed without
finishing after 2m 17s, 1m 47s CPU, with the patch. Debian package
autoconf_2.59a-3_all.deb.
It builds fine here for all platforms (i386, x86_64, ppc, ppc64)
for SUSE Linux 10.0 and SUSE Linux 10.1 (which both have autoconf
2.59).
RPMS are here:
ftp://ftp.suse.com/pub/projects/m17n/{10.0,10.1}/RPMS/
--
Mike FABIAN <mfabian(a)suse.de> http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。
4:38 p.m.
New subject: autoconf 2.59
Mike FABIAN <mfabian(a)suse.de> wrote:
It builds fine here for all platforms (i386, x86_64, ppc, ppc64)
for SUSE Linux 10.0 and SUSE Linux 10.1 (which both have autoconf
2.59).
I had no problems with this patch + autoconf 2.59 on my Fedora Core 5
machines, nor on a Ubuntu Dapper Drake box with autoconf 2.59. Has the
Mac version of autoconf been patched in any way from the upstream
sources?
--
Jerry James, Assistant Professor james(a)xemacs.org
Computer Science Department http://www.cs.usu.edu/~jerry/
Utah State University
4:05 a.m.
New subject: autoconf 2.59 [was: 2.60] runs into an endless loop
>>>> "Stephen" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
Stephen> Unfortunately, Andreas's patch breaks 2.59 in the same way. [...]
Stephen> Malcolm?
The ultimate cause of the problem is that 2.60 fixes a bug in the internal
macro m4_cdr, resulting in different return value when passed a one-member
list. As Andreas found out it now returns '[]' rather than '[[]]'.
The solution to the compatibility question is to make our own copy of the 2.60
m4_cdr so that we can use it with both versions of autoconf. Attached is a
patch that does this. When run through 2.59 it produces a copy of configure
that is identical to what is in CVS.
Looking through the release notes of 2.60 I see that the definitions (but not
the values) of datadir, infodir and mandir have changed. This will have to be
looked into.
Malcolm
--
Malcolm Purvis <malcolmp(a)xemacs.org>
6478
days inactive
6497
days old
11 comments
4 participants
participants (4)
-
Jerry James -
Malcolm Purvis -
Mike FABIAN -
Stephen J. Turnbull