[geeks] And The Linux Weenies Wonder Why They Aren't Mainstream...`

[Joshua Boyd]

On Thu, Mar 02, 2006 at 11:26:20AM -0500, der Mouse wrote:
> > Just using two architectures will shake out most problems.  Say a
> > Lintel machine and a Sparc system with the Sun compilers.
> 
> Well...if the SPARC is a sparc64.  But otherwise, you've got two 32-bit
> machines there, with nothing to smoke out word-size dependencies.

It could be a 32bit SPARC and a 64bit Lintel box.  But, I wasn't really
considering 64bit issues.  I guess that two different 32bit boxs might
only shake out 90% of the issues rather than 95%.
 
> Of course, if you really want to do a good job of portability testing,
> throw in some even less "standard" machines, like, say, a PDP-10 (or
> more likely a simulator) and a Lisp Machine.  (I've been told C
> compilers exist for Lisp Machines, and they are great at finding
> assumptions like "all the world's a byte-addressed flat address space",
> since their pointers really do take advantage of the freedom the spec
> offers about how pointers work.)

Well, that would be optimal, but I'm not aware of anyone even desiring
to port current software to PDP-10s or C code to lisp machines.  It
certainly would be educational to try though.  I keep thwacking at a guy
here for making assumptions he can always freely convert pointers to
ints.  

-- 
Joshua D. Boyd
jdboyd at jdboyd.net
http://www.jdboyd.net/
http://www.joshuaboyd.org/