autoreconf Best Practice?

[Michael Schwendt]

On Sat, 17 Jul 2004 08:53:18 -0700, Toshio Kuratomi wrote:

> Some months ago I asked whether patching or rerunning the auto* tools was
> the better method when modifying build files in a spec file.  This
> question is for those that thought that using the auto* tools was the right
> way to go.

Where size doesn't matter, I'm all for patching a pristine source
tarball with regenerated auto* files. Usually, however, such a patch
will be larger than 1 MiB uncompressed and duplicate the size of an
average rpm easily.
 
> I've recently run across autoreconf [1]_ in the autoconf toolset and
> wondered if the best way to rebuild build scripts would be something like::
> 
>   %setup
>   # Patch to configure.in, Makefile.am, etc
>   %patch0 -p1
>   %build
>   autoreconf
>   ...
> 
> Is this cleaner than using libtoolize --force or autogen.sh?

autogen.sh usually suffers from the drawback that it doesn't add
--force options. As a result, you end up with a mismatch between
autotools and libtool versions/files and e.g. get libraries without
the .so at the end and other defects.

Use the method that works for you. Both methods create an ugly
and sometimes troublesome dependence on the auto*/libtool packages,
often the versioned ones like automake16, autoconf253, ...

> If so, are there some options that would be appropriate to always pass to
> autoreconf?  --force --install seem like likely candidates.

Certainly --force --install.
 
-- 
Fedora Core release 2.90 (FC3 Test 1) - Linux 2.6.7-1.478