PKG_SKIP/FAIL_REASON

[David Holland <dholland-pkgtech%netbsd.org@localhost>]

After some discussion a couple weeks ago, I'm working on adding
BROKEN_ON_PLATFORM (and BROKEN_ON_COMPILER) to supplement
NOT_FOR_PLATFORM and NOT_FOR_COMPILER. The idea is that we restrict
NOT_FOR to cases where it doesn't make sense to build the package
(e.g. netbsd kernel modules on linux) and use BROKEN_ON where the
package does make sense but doesn't work, like most of the non-64-
bit-clean packages.

This seems like a useful distinction, and we could teach pbulk to
recognize the difference and both shut up about the packages that
don't make sense and more visibly report the ones that don't work.

However.

After looking into implementing it, it seems that the sensible way to
implement this is to have BROKEN_ON set PKG_FAIL_REASON and NOT_FOR
set PKG_SKIP_REASON. After all, skipping vs. failing is exactly the
distinction I'm trying to make, these variables already exist, the
proposed usage is consistent with what documentation of them I've
found, etc.

The problem is: currently most of the usage in the tree doesn't really
distinguish PKG_FAIL_REASON and PKG_SKIP_REASON. Some cases set one,
some the other, mostly PKG_FAIL_REASON but without any clear pattern.

Would anyone object to me cleaning this up along similar lines? (That
is, PKG_FAIL_REASON for failures that should be reported,
PKG_SKIP_REASON for cases where trying to build wasn't sensible.)

It also seems pbulk doesn't distinguish these at all, but that can be
tackled later on.

-- 
David A. Holland
dholland%netbsd.org@localhost