[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: New stable version after Sarge



Someone should patch Thunderbird so it handles M-F-T:. Grump.

Wouter Verhelst wrote:
Packages qualify to be enter prestable after residing in testing for
ten days and having NO RC BUGS. The idea is to keep prestable in a
highly stable state at all times, a rolling stable if you will.
That's how testing started off. We stopped doing this because

Err, no, we didn't stop doing anything of the sort.

a) it at one point stalled glibc; as a result, nothing moved to testing
   anymore, and when it finally did, the changes were so dramatic that
   testing was broken for quite some time.

Everytime glibc gets RC bugs, glibc gets stalled. It was stalled pretty continuously for something like 8-10 months (iirc) in late 2002, early-mid 2003. That is to say that at that point we couldn't produce a releasable gibc in less than eight months. At the point that got fixed, something was broken about at least KDE, and a whole range of other interdependencies had come up. Outside of KDE, I don't recall any overly major issues though.

Fixing bugs early and often while still doing development is a tricky task. In particular it involves both an attitude change from the way many people tend to work (viz, "I'll fix all my RC bugs now, not wait for someone to tell me to, or after I've implemented these few features") and a different set of skills and techniques to actually achieve (such as better pre-release testing so you don't release some features that turn out to be horribly buggy and unfixable, but also so useful that you don't want to rip them out from under people either).

The glibc team took a while to get a handle on those issues when they took over the package; hopefully that's not an issue that'll come up again. The KDE and Gnome folks are getting the hang of it now too, which is why you can see major new versions of KDE and Gnome being accepted now, and not causing significant disruptions to the release schedule (such as it is).

OTOH, over half the open RC bugs were filed over three months ago, so it's not like this is a solved problem. But we are getting somewhere.

b) Some RC bugs are only discovered when they migrate to testing. If the
   promise of 'prestable' would be that it would contain NO RC BUGS,
   then we would have to throw those out. That would likely result in
   prestable being a very incomplete system.

If the promise were "NO BUGS" in any shape or form, we'd never be able to include any software. A new stable version gets released when we've got some new software that essentially has no RC bugs that we know about, not when it has no bugs at all. Trying to hold testing or some 'prestable' or whatever else to a higher standard isn't reasonable, let alone feasible.

Also, adding yet another distribution that is even harder to update than
testing is doesn't seem like a good idea to me. We're already having
issues with testing and security; $DEITY forbid that we would make it
worse.

One of the reasons presented for not having security support for testing is that it changes too much. I suspect that argument would still be made for anything that changes more often than stable though.

Cheers,
aj



Reply to: