Thinking some more about it, I think there's going to be some other
refactor.
First, I'm unhappy with how pkg_create's error handling would look, and
also, I see very similar code in make update-plist.
So, I'm probably going to move the subst stuff into OpenBSD::Subst (or
something) first, then have CMD_SUBST be a very small script under
${PORTSDIR}/infrastructure, as I figure the end user won't want it.
I'm also thinking of having make update-plist registering VAR_SUBST in
the PLIST `before' substitution (so only update-plist and pkg_create
would need to deal with that). The idea would be that update-plist would
remember what variables have already been dealt with, and not reintroduce
back-substitutions if variables are too common. This is probably utterly
unclear to almost everyone, except porters who have had update-plist be
ways too smart in its substitution patterns. ;-)
