[Cgl] cgl, cpx, mir, error

[Lou Hafer]

Matt,

	You're touching on a point where behaviour varies from one OsiXXX to
the next.  You can set the primal and dual variables. Some OSI's will observe
these when returning row activity or reduced costs. Some will not. There's no
easy work-around that I'm aware of except to patch the OSI. I added a set of
tests to the base Osi unit test about two weeks ago (branches/devel; came up
while I was bashing OsiGlpk into better shape) but prior to that it was up to
the implementer.

	More generally, I guess you could say that people are starting to really 
push the OSIs, and that's exposing the places where the early implementers 
differed in their interpretation of the interface spec. There's more room for 
ambiguity than you might think, even after a critical read.

							Lou