On Wed, Jun 06, 2007 at 04:44:11PM -0700, Eric Anholt wrote: > On Wed, 2007-06-06 at 17:41 -0400, Kris Kennaway wrote: > > On Wed, Jun 06, 2007 at 11:24:44PM +0200, Pav Lucistnik wrote: > > > Kris Kennaway p??e v st 06. 06. 2007 v 16:04 -0400: > > > > On Wed, Jun 06, 2007 at 04:25:03PM +0000, Pav Lucistnik wrote: > > > > > pav 2007-06-06 16:25:03 UTC > > > > > > > > > > FreeBSD ports repository > > > > > > > > > > Modified files: > > > > > x11/libX11 Makefile distinfo manpages pkg-plist > > > > > Removed files: > > > > > x11/libX11/files patch-src_ImUtil.c > > > > > Log: > > > > > - Update to 1.1.2 > > > > > - Don't remove lib/X11, it's in mtree > > > > > > > > BTW, we had decided not to do minor updates to the X.org ports without > > > > a good reason. The disruption caused by frequent tiny updates is > > > > tremendous. > > > > > > Why not? There are no consequences in updating to dot-dot releases here. > > > I was told by flz and anholt that I'm free to help out keeping X ports > > > up to date. > > > > The main consequences are: > > > > * Forcing the package clusters to frequently rebuild all of X, slowing > > down all package builds and the rate at which we can do other testing > > on pointyhat. > > > > * The inevitable problems with commits to "deep" ports like this one > > which will take out most of the ports tree when they fail to build, > > leading to yet more delays in publication of usable package sets. > > > > What was decided was to only update in batches when x.org makes a new > > "kamikaze" release (or whatever they are calling it ;), unless there > > is a really good reason to update a port otherwise. > > That's an unfortunate policy, given that generally X developers can only > be bothered to make releases when there's an important fix to push > out(*). In this libX11 case, this point release included stability and > security fixes. I suggested to Pav that libs updates were a good thing > to do, since they're generally bugfixing and such, unlike drivers and > the server (and the protos they depend on) where more context is > sometimes needed. > > * The exception to this in the past has been at katamari time, when we > were rolling up any modules with trivial changes that developers hadn't > been bothered to release already. I think I've successfully killed that > part of the katamari process.
As I said, if there is a really good reason then we can include point updates, but both FreeBSD developers and X.org developers (I guess that means you) need to understand that there is a high cost for us to incorporate them into the ports tree. The cost becomes prohibitive if updates are happening somewhere in one of the core X.org ports on a weekly basis, because the package build clusters will spend most of their time continuously rebuilding everything that uses X instead of letting us do real work. Since you (X.org) have made the decision to effectively push out a large part of the X.org "product" release engineering (integration testing and decision-making about components) onto each individual distro, this is part of the cost of that decision that the X.org release engineers (you again, I guess) needs to be aware of. The FreeBSD project does not have the resources (or desire) to effectively do full-time incremental X.org release engineering because of X.org changes being continuously pushed into ports. The best we can do is periodic batch updates when X.org announces a significant accumulation of enhancements, with the occasional critical fix when necessary. I hope you will factor these constraints into your development process and work with us to achieve something we can both live with. Thanks, Kris
pgpOCfvSNX9yT.pgp
Description: PGP signature