autoconf 2.59 [was: 2.60] runs into an endless loop

[Mike FABIAN]

"Stephen J. Turnbull" <stephen at xemacs.org> さんは書きました:

>>>>>> "sjt" == Stephen J Turnbull <stephen at xemacs.org> writes:
>
>>>>>> "Mike" == Mike FABIAN <mfabian at suse.de> writes:
>
>     Mike> Andreas Schwab <schwab at 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-21.5.27.20060705-3.1.i586.rpm
http://software.opensuse.org/download/M17N/SUSE_Linux_10.0/i586/xemacs-el-21.5.27.20060705-3.1.i586.rpm
http://software.opensuse.org/download/M17N/SUSE_Linux_10.0/i586/xemacs-info-21.5.27.20060705-3.1.i586.rpm
http://software.opensuse.org/download/M17N/SUSE_Linux_10.0/src/xemacs-21.5.27.20060705-3.1.src.rpm

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 at suse.de>   http://www.suse.de/~mfabian
睡眠不足はいい仕事の敵だ。