But what about kernel printfs?

[Greg Limes]

In article <4659 at ihlpf.ATT.COM> nevin1 at ihlpf.UUCP (00704a-Liber,N.J.) writes:
>In article <12360 at tut.cis.ohio-state.edu> lvc at tut.cis.ohio-state.edu (Lawrence V. Cipriani) writes:
>>Even if one argues that an implementation of the UNIX
>>kernel is not portable to other machines it should be
>>at least portable to other compilers for the same cpu.
>
>Unfortunately, the implementation of the kernal (or is it kernel? :-)) is
>not only CPU dependent but very machine-dependent.  I agree that an
>implementation of the kernal should be written to avoid as many of these
>dependencies as possible, but the dependencies on all of the supporting
>hardware is still there.  For example: I would NOT expect A/UX to run on
>any 680x0 machine except the Mac-II (and its' decendents).

Don't be too sure. Remember that portable sections of the kernel can be
used to provide similar services on very different machines; that makes
porting to, say, a new product line easier. While I was not part of the
porting team, I can imagine how portable (and nonportable) code
impacted the jobs of the people setting up SunOS on the Sun-4 and the
Sun-386i. I *do* know that most of the source for SunOS 3.x is the same
between the Sun-3 and Sun-4 versions. Writing archetecture, machine, or
compiler dependent code without good reason may be asking for trouble
down the line.
-- 
   Greg Limes [limes at sun.com]			Illigitimi Non Carborundum