On Sun, Feb 26, 2012 at 12:55, David Walser <luigiwal...@yahoo.com> wrote: > This got me to thinking, however, that something like this should not have > broken the build system. There's no reason for iurt to be setting > up the chroot with packages from Cauldron, unless they are part of the > BuildRequires (including recursively) for the package being built. In > this situation, once a package needed in the base of the chroot is broken, > there is no way to fix it without sysadmin intervention. This > would be true for a broken package in Cauldron, or stable/updates_testing.
Cauldron chroot only uses cauldron packages. Stable chroot only uses stable packages. There is no reason to use non cauldron packages when building for cauldron, and this would likely break anyway. There is a bug opened to keep old chroot when we fail to generate new one. > I believe a better way for iurt to work would be this. It would set up the > initial chroot from the newest stable release of Mageia that is > the same version of Mageia the package is being built for, or older. So > right now, that would mean Mageia 1 in all cases. Once Mageia 2 is > out, it would mean Mageia 2 for packages being built for 2 or Cauldron, and > Mageia 1 for packages being built for 1. Also, when the initial > chroot was being installed, release and updates would be enabled only. Then, > when it gets to the BuildRequires stage, first enable either > updates_testing if it's being built for a stable relase, or Cauldron if it's > being built for that, and then install the BuildRequires and > build the package. This is wrong, config files, rpm-build, compiler, etc would be from stable. We want cauldron packages built on cauldron.
