On Sat, 2008-04-19 at 23:11 +0200, Fabian Groffen wrote: > On 19-04-2008 16:07:19 -0500, matt hull wrote: > >> Due to numerous occasions of b0rkerage in the bootstrap snapshots, > >> bootstrapping has failed. A selection of issues that come up: > >> - XXXX/YYYYY-p.q is the latest version in the snapshot, but this version > >> has been removed, and its distfiles have become unavailable (e.g. > >> rsync-3.0.0_pre2) > >> - XXXX/YYYYY-p.q was added in the snapshot, but breaks several packages > >> (e.g. gcc-4.3.0) > > > > perhaps we need to keep older versions for bootstrap ? > > I don't think that's a problem. In general the highest version in the > tree is used at the moment due to our single keywordness. > > >> To aid the pre-sync stage, I am considering to switch to usign stable > >> keywords for the system packages *only*. That is, the bootstrap process > >> is done with stable keywords, all other packages remain ~arch and hence > >> a user has to add ~arch to ACCEPT_KEYWORDS after bootstrapping finishes. > > > > i would think it would be better to use stable and testing like normal. i > > would like to try to keep stable here. > > That's very unlikely to happen from my point of view. For that we > really need a bunch of devs that are going to maintain the stable > keywords like the arch teams currently do for gentoo-x86.
As we have approximately less than 1 dev per arch, IMO for the bootstrap _only_ it could be possible to have "stable and testing like normal", but we might need to reduce the definition of "normal" (drop stablereq bugs, ...), but then we cannot call it "normal" any more ;) /haubi/ -- Michael Haubenwallner Gentoo on a different level -- [email protected] mailing list
