Matthew Garrett wrote:
Anthony Towns <aj@azure.humbug.org.au> wrote:If they can all satisfy the criteria, they're likely to be doing well enough that there's not much *point* to dropping them -- the reason 11 architectures are hard to manage is because they're not all being supported at an adequate level. The criteria listed try to give a good idea of what "an adequate level" is likely to look like.That doesn't seem to match up very well with: "there is a general feeling that twelve architectures is too many to try to keep in sync for a release without resulting in severe schedule slippage."
The *reason* there's that feeling is that our eleven architectures just aren't being maintained all that smoothly. Problems on the new architectures included in woody don't seem to be being attacked with the same admirably rabid viciousness now that they were then, and that really is a problem. It seems to me the same's true of most of the proposed non-release-candidate architectures too, though I'm not following any of this as closely as others.
I'm not even sure if that's not also true of the proposed release-candidate architectures; but if it is, at least it's at least easier to deal with those problems on a small number of reasonably popular architectures.
I agree with the idea that architectures that have a history of falling behind and having severe toolchain breakage shouldn't be allowed to hold up the other architectures from releasing,
The issue isn't "severe" toolchain breakage, it's one of frequent small inconveniences and breakages that end up compounding into major problems; and because they're locally seen as only minor problems, don't get fixed quickly enough to avoid compounding into greater problems.
but if it's the feeling of the release team that releasing that many architectures is basically impossible then we may have problems.
Please don't read too much subtext into these things; communication's a lot harder if you have to double check everything you say to make sure it won't be misinterpreted. TTBOMK, architectures which legitimately pass the listed requirements (ie, meet the spirit as much as the letter) won't have any problems being released.
If there are eight architectures who're going to try to meet them, then they'll probably be clarified further to include other specific problems that require attention (like kernel and toolchain issues), but that's with a view to ensuring that they're going to be maintained smoothly, not as some indirect way of limiting the number of architectures getting released.
Cheers, aj