On 2007-May-19 14:46:56 -0700, Garrett Cooper <[EMAIL PROTECTED]> wrote: > Doug Barton wrote: >> Garrett Cooper wrote: >>> Wouldn't it be sufficient to force major component testers (in this >>> case Xorg 7.2) to use periodic snapshots of the ports tree (possibly CVS >>> branching), while allowing continued development in the ports tree?
My understanding is that most of the testing was done in a separate repository. But at some stage, it needs to be committed - the approach used ensured that the changes could be tested by a larger group of people before being finally released. This particular change affected 6168 (more than 35% of) existing ports. A portupgrade of this magnitude can easily take days unless you are game to enable BATCH mode. It is simply too difficult to allow ports to be randomly changed in the midst of this upheaval. > True, but I wonder how end users are going to take to the fact that a lot of > ports were removed or changed in the 7.2 integration period. I've noticed a > lot of 'deletes' for ports when running csup today. Looking at the commit message, no ports were removed. There were 15 removed files. > Perhaps, but I think that some of these things maybe could have been handled > differently ( / better?) if source branching was in place, both for devs and > for end-users in the X.org 7.2 evaluation phase. Maybe it would have been useful to have a 'pre 7.2 import' tag but (as is regularly pointed out) actually branching the ports tree is impractical. The ports team already has to try and keep >17,000 ports working across 3 FreeBSD branches, each with about 7 architectures. Attempting to have 'stable' and 'development' ports trees would just double the work involved and computing power necessary to do the builds. -- Peter Jeremy
pgpOTLtQCystR.pgp
Description: PGP signature
