svn commit: r383191 - head/Mk

[Alexey Dokuchaev]

On Thu, Apr 09, 2015 at 10:15:17AM -0500, Bryan Drewery wrote:
> *That* made sense. Frankly I think all dev warnings should be disabled
> in that case though. Setting up ports like this may work but it is not a
> good way to test a port before committing or submitting it. Many ports
> and their framework files reference "CURDIR/../..". In r327743 I
> modified www/linux-seamonkey to use PORTSDIR here and I believe I broke
> the workflow for the maintainers testing. The ports tree really needs to
> be tested as a whole. This is why portshaker exists, so you can bring in
> your partial tree into a full tree. Unionfs in theory is for this too.

Bryan, I'm on Dmitry's side here.  Building some "foo" port from outside
/usr/ports (or a properly checked out tree) is very handy feature, and it
had worked for many years; and while perhaps never officially advertised,
silently breaking it is a huge POLA violation for power users.  I would
really appreciate if we can have this behavior back.

On a related noted, I'm generally getting more and more concerned that
the tree becomes primarily chroot/jail building-oriented.  Example:

I'm personally heavily relying on ability to set OSVERSION to arbitrary
value on per-port basis; it is very helpful to test the logic of some
complicated port (e.g. x11/nvidia-driver) to make sure you got all these
checks right.  Now since r369644 (also by you) I have to comment that
harness code from Mk/bsd.port.mk out.

I was actually surprised that I have to do it, since log message talks
only about "OSVERSION or UNAME_r is improperly set when building in a
jail/chroot", but in reality I cannot do a "make OSVERSION=xxx" even when
building locally.  I'd also appreciate if you could change the code so
it works as described (i.e. for jail/chroot-environments, bulk package
builders, tinderbox/p*re, et al.), but not for /usr/ports, thank you.

./danfe