Richard Shaw wrote:
Could you build the package twice using mock (x86_64 and i686) and
run
rpmdiff on them to see if there's anything significant?
FYI, if you use an arch-specific dummy main package and a noarch -devel
subpackage, that is actually automatically done by Koji at every build. So
if there is arch-specific content of any kind (arch-specific paths, multilib
conflicts (arch-specific file contents), etc.), it will fail the build.
Of course, if you use a noarch main package, you're on your own (Koji will
only build it once on a random architecture), but that is frowned upon
anyway, because tests (%check) will also only be run on one architecture.
Kevin Kofler