Re: m4-1.4.10 breaking autoconf?

[Thomas Klausner]

On Fri, Jul 20, 2007 at 09:38:00PM -0600, Eric Blake wrote:
> Thanks for the report.  I have re-bootstrapped m4 1.4.10 with itself, and
> didn't have any problems on my cygwin platform with the resulting
> configure script, so either the m4 configure.ac is simpler than mng's, or
> m4 1.4.10 is tickling a platform-specific bug.  Hopefully you can help us
> narrow in on the problem.

I can reproduce this problem with m4's configure script.
The file created by running 'autoconf' is just 35354 bytes long, while
the one distributed with the m4-1.4.10 tarball is 647263.
I'll attach the generated configure file (m4-configure.bz2).

> I'm assuming that the only variable you changed was whether you used m4
> 1.4.9 vs. 1.4.10?  Could you also attach a diff between the configure
> scripts generated by the two versions, so that we can narrow down on what
> in the source to the configure script may have tickled a difference?

Yes, the change is using 1.4.9 (with a bug fix applied that's included
in 1.4.10) vs. 1.4.10.
The diff is too big, I'll just attach the two versions.

> Also, what platform is this on (both OS and processor type)?  Could you
> also post the results of running 'make -k check' on the m4 1.4.10
> testsuite? (Even if this was the NetBSD-4.99.23/amd64 of your previous
> message, at least rerun with make -k so that we can see the results of the
> rest of the testsuite.)

The testsuite works fine, but skips a few tests:

Skipped checks were:
  ./094.changeword ./095.changeword ./096.changeword ./097.changeword
  ./098.changeword ./099.changeword

Because of:
Checking ./099.changeword
 skipping: no changeword support

Hope this helps,
 Thomas

configure-1.4.9.bz2
Description: Binary data

configure-1.4.10.bz2
Description: Binary data

m4-configure.bz2
Description: Binary data