In message: <4B2C127A.8060509@FreeBSD.org>
Doug Barton <dougb@FreeBSD.org> writes:
: Dag-Erling Smrgrav wrote:
: > John Baldwin <firstname.lastname@example.org> writes:
: >> Warner Losh <email@example.com> writes:
: >>> Log:
: >>> Fix compiling FREEBSD_COMPAT[4,5,6] without FREEBSD_COMPAT7.
: >>> Note: Not sure this is the right way to do compat, but it makes the
: >>> headers consistent with the implementations.
: >> Please revert. I think this has already been reverted once before.
: > Yep, that was me... we concluded back then that this is a perfectly
: > reasonable dependency; perhaps we should simply add the appropriate
: > #ifndef / #error constructs so you can't define FREEBSD_COMPAT(n)
: > without FREEBSD_COMPAT(n+1).
: Trying to look at this from the user perspective, it's kind of silly
: that if what I want is COMPAT5 I also need to add COMPAT<somethingelse>.
: I don't know the issues in the code so I don't know what the solution
: should be on the back end, but on the front end what needs to work is
: that I specify the compat level I want in my kernel config, then the
: code does what's necessary to make that work.
The trouble is that we need to do this globally, or certain
definitions don't get defined. It is easier to just #error out than
hack config to do the right thing I think...