On Mon, 2 Oct 2006 10:33:11 +0200 (CEST), Nicolas Mailhot wrote:
Le Lun 2 octobre 2006 09:50, Michael Schwendt a écrit :
> rpmbuild fails if in the build root no filenames match the %exclude
> pattern, which can happen when using a single spec file for multiple
> targets. Hence "rm -f" is safer, because it works fine also with
> non-existant files.
replace "safer" with "less work for the lazy packager" there
If you expect to remove some files but they're not present *something*
changed upstream. Blindly continuing without any packager review certainly
isn't the "safe" choice.
Or something changed in the development tool-chain.
Well, I disagree anyway, since semantically it might be "if the file is
there, remove it as I do not want it, ever, as it would break my
package". The removal is intentional. The removed file might break the
build or create file-based conflicts or introduce precompiled binaries
(depending on the build host arch) and bad deps -- whatever. You don't
want extra burden for packagers, that they need to check with every
[perhaps tiny] update whether any "rm -f" commands are still needed for
every target platform. As with "rm -f %{_infodir}/dir" or things like
that.