>>>>> "Chris G D" == Chris G Demetriou <Chris_G_Demetriou@LAGAVULIN.PDL.CS.CMU.EDU> writes:

>> > Chris D> I've looked at the problem (twice in fact), and did not
>> find > Chris D> an 'easy' or 'nice' solution either time...  > >
>> How about generating makedev(x, y * MAXPARTITIONS + z) in
>> swapnetbsd.c > instead of precalculating the 2nd arg?  I mean
>> MAXPARTITIONS literally > here, not the value of the system
>> config.new is built upon.  Then when > compiling the kernel, the
>> target machine MAXPARTITIONS will be used, > just like it should
>> be.
>> 
>> This is from memory:
>> 
>> Basically the problem was that the numbers are calculated early and
>> then stored and manipulated that way all over the place.  For 1.0 I
>> basically took the number, un-"manipulated" it and then generated
>> something similar to the above.  It was a hack but we needed a
>> quick fix.

Chris G D> also, it's significantly complicated by the fact that one
Chris G D> can explicitly specify a major/minor number pair for the
Chris G D> device, rather than a name...

Pardon my ignorance (maybe it's just that I'm not a native English-
speaking person) but I'm still wondering about the magic '8' in sem.c.
Is it intentional that it is this way, or are my sources bad?  If it's
intentional, it's definitely bad style.  I understand to fix this
properly is hard, but to document current behaviour in the code (i.e
add a comment to the static function "resolve" in sem.c) wouldn't be
out of line, would it?  What if I state: "netbsd root on sd2j" how
will it be handled?  I'll try it myself when I find the time,  I
suspect it'll try to boot from sd3b or something, no?

Niklas