On Sun, 16 Jan 2005 21:17:52 -0500, Kumba <[EMAIL PROTECTED]> wrote:
I've always grasped the concept of keywording multiple packages in rapid fashion.
What you fail to grasp is proper QA. As a developer, one does not just go keyword happy and commit to CVS - One has to test *each* and *every* package, otherwise, bad things can happen.
You dont do that now for every folder in kdebase, do you? -- thats up to KDE upstream. No reason why you should do that for kdebase-meta either.
It is up to kde upstream to test everything on every arch? Maybe in an ideal world, but this world is far from ideal.
With monolithic KDE, we can test & keyword something like kdenetwork. That covers a number of packages in one sweep. With the split ebuilds, we now have to test and keyword each ebuild *individually* -- this consumes more time, but is necessary to stick to QA guidelines.
Ofcourse you dont. You test & keyword something like kdenetwork-meta. That covers a number of packages in one sweep...
Ummm...no.
Sure, it's the same XX ebuilds that we have to test between the monolithic and
split forms of KDE, and yes, chances are if a package in the split form
breaks, it'll also break in the monolithic form, so one would think that what
I argue is a moot point.
Bingo!
In a perfect world, it is a moot point, but alas, in a perfect world, many things would be different and/or better. QA is there for a reason, and if a dev skimps on it, we generally find the RepoMan and Mr. Bones waiting outside our front door the following morning.
Making sure things have correct dependencies and the kde-meta packages work is for the KDE team (what a hard job they are doing) -- all thats needed off the arch teams is to do exactly what they do with the monolithic ebuilds except add meta to what they typically emerge and run a simple script to keyword things properly (or leave that to the KDE team).
Roman
Again, no. Keywording a single meta-ebuild does not magically keyword all the individual sub-packages the meta-build pulls in. You are sorely mistaken here.
Steve
-- [email protected] mailing list
