[11:19] <infinity> cjwatson: http://paste.ubuntu.com/5867830/
[11:19] <infinity> cjwatson: http://paste.ubuntu.com/5867831/ <-- The product of the above.
[11:20] <infinity> cjwatson: My gut feeling is that we might also want those deps versioned, so we can force ourselves to rebuild against any new d-i component to keep sources in sync, but that seemed draconian for a first cut.
[11:21] <infinity> (And this matches what britney was already doing)
[11:24] <cjwatson> Certainly looks good for now.  Thanks!
[11:24] <cjwatson> Go for it.
[11:24] <infinity> Kay, will upload after I've tested that d-i also builds and does sane things with the linux-ppc 3.10 bump.
[11:24] <cjwatson> I suspect we'll find this is a "wow, we should have done this years ago".
[11:25] <infinity> (And possibly some lunch :P)
[11:27] <infinity> The big bonus here for removing the irriating seed skew and fiddling is that I'll be happier with the idea of rtg and apw doing ABI bump uploads of d-i, since it's just one package instead of some delicate "watch for migration and change seeds in the same publisher cycle" madness.
[11:27] <cjwatson> Indeed.
[11:40]  * apw looks up, ahh yes indee
[19:07] <infinity> Dangit my d-i change is going to make me slip off the "top contributor" list to ubuntu-seeds.
[19:07] <infinity> My precious, precious karma!
[19:07]  * infinity uploads a revert.
[20:27] <GrueMaster> Hmm.  Karma vs Quality.
[20:28] <infinity> Quality, schmality.  It's all about pointless metrics and popularity contests, right?
[21:22] <GrueMaster> THAT's why I'm no longer there!  I knew I was doing something wrong.
[22:16] <infinity> cjwatson: Bah, germinate's Kernel-Version pruning routine has thwarted me.  Updating seeds for now, but we might want to make germinate less clever.
[22:16] <cjwatson> You can't just delete the Kernel-Version seed entries?
[22:16] <infinity>     def _is_pruned(self, di_kernel_versions, pkg):
[22:16] <infinity>         """Test whether pkg is for a forbidden d-i kernel version."""
[22:16] <infinity>         if not di_kernel_versions or not di_kernel_versions.kernel_versions:
[22:17] <infinity>             return False
[22:17] <infinity>         kernvers = di_kernel_versions.kernel_versions
[22:17] <infinity>         kernver = self._packages[pkg]["Kernel-Version"]
[22:17] <infinity>         if kernver != "" and kernver not in kernvers:
[22:17] <infinity>             return True
[22:17] <infinity> I assume that's pruning anything not specified as a current/allowed kernel-version.
[22:17] <infinity>         return False
[22:18] <cjwatson> not di_kernel_versions => not pruned
[22:18] <cjwatson> first test
[22:18] <cjwatson> So if the seeds simply don't contain Kernel-Version at all ...
[22:18] <infinity> Oh, so if we specify none... Kay.
[22:18] <infinity> S'pose that's worth a test.
[22:19] <cjwatson> At least that's how it's meant to work.
[22:21] <infinity> I wonder if we'll find a bunch of modules udebs that don't show up in debian-installer-udebs and need seeding anyway.  Hrm.
[22:23] <infinity> I'm going to look at this tomorrow when I get in to London, I have an early flight and should nap instead of pondering this.
[22:25] <cjwatson> Probably.  Perhaps "Extra-Include: *-modules-di *-firmware-di" would deal with that.
[22:25] <cjwatson> Or similar.
[22:25] <infinity> Or just seed the kernel source packages wholesale.
[22:25] <cjwatson> (i.e. include all modules from the same source)
[22:25] <infinity> But I'll play around with some germinate runs and output tomorrow.  Sleep now.
[22:26] <cjwatson> There have historically been some cases where we've wanted to seed only some flavours but not others, so I'd like to make sure we still have that flexibility.
[22:26] <cjwatson> Sleep well.