Re: OpenBSD/amd64 / pkgtools/pkg_install - pkg_add compatible_platform() check oddness

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

On Mon, Jun 01, 2015 at 01:54:06AM +0100, Sevan / Venture37 wrote:
 > Ignoring the rest of the infrastructure and focusing on pkgtools/pkg_install
 > I went about hardcoding MACHINE_ARCH to see what it takes to pass the
 > check_platform() test, hardcoding MACHINE_ARCH (changes below) had no
 > effect despite it being set.
 > Setting effective_arch = "x86_64"; in add/perform.c allowed thing to pass.
 > I'm a little confused by how MACHINE_ARCH chould be overwritten or
 > does bmake's variable take priority (if so, these statements are
 > redundant are they not?)

It is a bloody mess. If you could sort it out once and for all we'd
all greatly appreciate it :-(

Because bmake and pkg_install can both be compiled outside of
bootstrap, they both need to be able to generate the right results
based on their own configury. The bootstrap script should not be doing
any more than crosschecking that they get the same values. Meanwhile,
at runtime the values generated by pkg_install and bmake should always
be the same; if they aren't, something's wrong.

What we currently have is rather a long way from that ideal...

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