Subject: Re: kernel compile report
To: None <port-pmax@NetBSD.ORG>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
Date: 08/29/1995 04:22:10
Thanks to Arne for reporting the problems he did.
I fixed the rcons.h problem by un-commenting-out the "rcons"
lines from conf/files.pmax; "config XXXCONF" will now produce
an rcons.h that says no rcons devices are present.
I committed a (still-untidy) version of pmax/asic.h, which
I've tried to brin closer to the Alpha port's asic.h.
That fixes the DIAGNOSTIC problems Arne describes, and it
also provides the symbolic definition that are needed by the new
Ethernet driver, which uses the machine-independent LANCE driver
in sys/dev/ic/am7990.c. This LANCE eliminates some bugs in the old
LANCE driver, particulary spurious reports of LANCE errors while
I've also committed changes to the cfb driver that make it work
on a 3MIN (with thanks to Yoseff Francus for reporting the problem
and testing fixes). These should all show up via sup tomorrow;
they were committed far too late to make today's.
The new-config kernel and drivers have now been booted on all
DECstation models except the MAXINE (5000/xx). What's more,
they should now build!
There are still several warnings in the kernel about prototype
mismatches on interrupt handlers. That's because the pmax
drivers are in transition from the old convention of passing a
unit number to a driver (which uses a static softc array),
to the new convention of passing a pointer to the softc structure
itself. It really is safe to ignore those warnings, until
all drivers and the prototypes are converted to pass pointers,
and return an int status (used to indicate spurious interrupts.)
I'm out of touch from *now* until early next week...