On Thu, May 20, 2004 at 12:09:54PM +0100, Colin Watson wrote: > On Thu, May 20, 2004 at 01:04:37PM +0200, Sven Luther wrote: > > On Thu, May 20, 2004 at 11:48:01AM +0100, Colin Watson wrote: > > > That's actually exactly the current state. However, it's more difficult > > > to get user testing of t-p-u uploads before they get into testing, so > > > it's not really something we want to rely on too much. > > > > Ah, last we tried this in february or so, testing-proposed-update was > > not being autobuilt, and the result was that it was not usable for > > debian-installer, but then maybe it has changed since then ? > > t-p-u has been autobuilt for years. I certainly have a package which was > uploaded to t-p-u last September and autobuilt.
Well, it certainly did not when we tried to use it for debian-installer shortly after beta2 out, if i remember well, which is way we are not using it for d-i purpose. But then it may have been another problem, not sure. > > If so, then this is the ideal method for solving the current problem, > > and we could quite well upload gnome 2.6 to experimental and use t-p-u > > for RC bug fixes in testing should they show up, > > Since the autobuilders build against unstable, uploading GNOME 2.6 to > unstable affects lots of things; it's big enough that, at this stage, it > pretty much totally commits us to having GNOME 2.6 in sarge, even with > t-p-u. As such, it is very much the release team's responsibility to > ensure that the risk is as close to zero as can feasibly be managed. Yep, i understand that. That said, given the timeframe involved, not moving 2.6 to unstable, pretty much commits us to _not_ having GNOME 2.6 in sarge one it is released, so, it is indeed a choice to make. > Please, it will be a lot less work to demonstrate that GNOME 2.6 is > ready than to try to find hacky workarounds. As far as i can tell, t-p-u has never worked as it should, and using a single repository experimental which is not autobuilt is no user friendly way of getting testing done. It is a real pain to use experimental in the current state of things, so we will get little testing done, especially on the non mainstream architectures. Friendly, Sven Luther
