Re: Some warnings of type `unsupported' don't cause "automake -Werror" to fail

[Ralf Wildenhues]

[ let's drop bug-automake ]

* Stefano Lattarini wrote on Tue, Apr 20, 2010 at 12:55:36PM CEST:
> At Tuesday 20 April 2010, Ralf Wildenhues wrote:
> > ....
> > I'm applying this fix.
> > ...
> >     * tests/configure.test: New test.
> > ...
> OK, I'm a dope...  I found the bug while trying to write a test script 
> for the copresence of configure.{in,ac}, but then I didn't include that 
> script in the bug report -- which would have made the report better, 
> and saved yourself the work of writing a new testcase from scratch.
> Sigh.
> 
> Oh well, attached here is my original test script (which is IMHO 
> slighty stricter and more comprehensive than yours in some respects).  
> Maybe you can find it useful.

AFAICS, your test doesn't ensure that without -Werror turned on,
automake actually suceeds when both configure.ac and configure.in are
present.

OTOH, actually ensuring that configure.ac is used seems like a good
idea.  I'm amending the test in your name as follows, hope you don't
mind.

> Thanks for your fix (and sorry for my sloppiness),

No worries.  I didn't put you in as co-author of the original patch as
preemptive punishment.  ;-)

Cheers,
Ralf

2010-04-20  Stefano Lattarini  <address@hidden>

        Make test for configure.in vs. configure.ac stricter.
        * tests/configure.test: Use a configure.in file that provokes
        an automake error, to ensure configure.ac is preferred.

diff --git a/tests/configure.test b/tests/configure.test
index 4b0f28d..50a034f 100755
--- a/tests/configure.test
+++ b/tests/configure.test
@@ -28,6 +28,13 @@ AC_CONFIG_FILES([Makefile])
 AC_OUTPUT
 EOF
 
+cat >configure.in <<EOF
+AC_INIT([configure], [1.0])
+AM_INIT_AUTOMAKE([an-invalid-automake-option])
+AC_CONFIG_FILES([Makefile])
+AC_OUTPUT
+EOF
+
 : >Makefile.am
 
 $ACLOCAL 2>stderr && { cat stderr >&2; Exit 1; }