On Mon, Feb 27, 2006 at 12:39:11AM +0100, Quentin Garnier wrote:
> On Sun, Feb 26, 2006 at 11:28:21PM +0100, Peter Postma wrote:
> > On Thu, Feb 23, 2006 at 01:49:52AM +0100, Quentin Garnier wrote:
> > 
> > I had to make a small change to hpcapm.c, removing the include
> > machine/autoconf.h. I don't think that will affect the other hpc ports
> > but I'll cross-compile them to be sure.
> 
> I think it's a left over;  it probably was needed for struct
> mainbus_attach_args, which of course is not used anymore.
> 
> > And there's a minor compatibility problem... the apmdev device uses
> > a different major number from the apm device, so users will need to
> > create a new device node. I've created a new major file
> 
> Hum, no.  Just use the same major as before.
> 

Uh, ok.

> > (hpcarm/conf/majors.hpcarm) for this (arm/conf/majors.arm32 didn't feel
> > right). A new apm binary is also required to query the status.
> 
> I'd rather have a MI definition for hpcapm.  But we'll still need a
> definition for compat_30 anyway.
> 

Ok, but then I don't see how you want to keep compatibility, we still
need to change the no. for the MI definition.

> Why would a new binary be required?  AFAICT, the ABI doesn't change.
> 

There's a small difference in the APM_IOC_GETPOWER ioctl: the old apm
uses "_IOR" and the new uses "_IOWR".

> > So maybe this should be noted in src/UPDATING and/or -current-users.
> 
> I don't think we have a compatibility problem.  I'll have to think more
> about it though.
> 

There might only be a ABI compatibility problem, but we'll have to live
with that...

-- 
Peter Postma