On 8 Jun 00, at 9:22, Gilbert Carl Herschberger II wrote:
> Yes. Same here. Couldn't each of these classes could be written out "long
> hand" and use inheritence rather than templates? Or, should I wait for a
> new compiler?
Not having much to say about the code itself since I haven't written
any of it (and so don't know what design decisions went into
certain choices) I would shy away from templates in native code for
JOS just for portability sake. In fact, I wouldn't go much past
EC++ (embedded C++) standard stuff (which I believe nixes
templates). I'd also avoid using C++ exceptions because of its
mixed implementation platform to platform.
I hold these opinions because of two factors that have been
gnawing at the back of my mind for a while. First, we want to
maximize portability. The simpler the language you use, the better
the chance you'll find a compliant compiler on a platform. Second,
although its not mentioned much at all, it would be best if at some
point JOS were self hosting. I see no reason to port the entire gcc
to java. It would seem more logical to use C or even a subset of C
since the native code should be very limited in scope (at least as
far as our ideal JOS design is concerned). That way, we could
build a small C compiler (probably based on the old classic
SmallC) and a small assembler. Not something we'll be doing
soon, but something I'd like us to keep in the back of our minds...
-iain
_______________________________________________
Kernel maillist - [EMAIL PROTECTED]
http://jos.org/mailman/listinfo/kernel
