> For me, this is even less obvious than all the other scenarios you've > mentioned. And sometimes, it is a problem to have i386 packages installed > on x86_64, which the excludes in mock illustrate perfectly. Compiling packages with the 32bit packages installed has its own problems that we just try to avoid. > I'd really prefer having a behavior closer to what Christopher was trying > to do, as the reason for having i386 packages on x86_64 is mostly because > of broken software (OpenOffice) or proprietary software (web browser > plugins), which aren't good reasons in the first place. So keeping them > out except when explicitly requested (i.e. *.i386) is what I'd prefer. Agreed, but it is not that easy to make the right decisions for this. The current solution should get things running for more setups by installing all available packages, but getting the dependencies right on how to resolve i386 deps and if that should pull in again i386 packages or maybe x86_64 packages and the details on this is not very easy. greetings, Florian La Roche
