2012/4/11 Ken Dreyer <[email protected]>: > On Wed, Apr 11, 2012 at 7:48 AM, Paul Howarth <[email protected]> wrote: >> Latest stable version for F-15 is 0.99, which I think should be OK, but >> there's a broken 1.01 in updates-testing. The builders aren't pulling >> that one in, are they? > > They do pull from updates-testing. RPM Fusion (kwizart) frequently > rebuilds all the kmods against each of the new kernels that land in > updates-testing, so if I were to guess, that's probably why the This is wrong, the kernel is fetched from a dedicated override repo. But indeed, updates-testing is enabled by default on the builder to save maintenance when a new RPM Fusion package need to be rebuilt against a new fedora package (thinking about all the freeworld variant and etc). This has to stay like that until there is a per-users way to handle overrides (using bodhi). I can only agree this is not ideal.
> builders are set up the way they are. One of the confusing factors in > this situation is that the local mock-rpmfusion-free/nonfree packages > do not include updates-testing, so it wasn't obvious how to duplicate > this error locally with mock.[1] You make the assumption that mock-rpmfusion-free is only aimed to reproduce RPM Fusion builder behaviour. This is wrong, one can expect to build only with fully validated packages from both fedora and rpmfusion. Reproducing the buildsys behaviour is only one way to use the mock cfg and is not the default. So I'm closing the bug. > Paul, do you think it's feasible that we can fix Params::Validate for > F15? Check out Ralf's comment in Bodhi[2]. I'm trying to figure out > what breaks with 1.02. If one package is broken in f15 updates-testing it can also be withdrawn. Nicolas (kwizart)
