Spehro wrote:

>This situation illustrates why some of us were beating Lawrence L.
>over the head with ANSI compatibility issues. If you write most of
>your code in conforming C, then you should be able to port it to
>another compiler (or even another processor) with minimal effort.
>(you will have to do the usual regression testing, of course)
>
>Unfortunately, most things that do something directly useful
>are not part of standard C, almost by definition. Extensions
>that tend to vary include methods of mixing ASM with C, ways
>of dealing with interrupts, directives to control memory space
>for given variables, higher level I/O functions, functions such
>as delay() etc. And even on the best microcontroller compilers
>you'll likely find a certain number of problems, unfortunately,
>if you do any serious programming.


That's the nice thing about writing an RTOS (or ours, at least). It
doesn't do anything outside of ANSI C except for a  wee bit in the
context switcher and accommodating interrupt code. So it's eminently
portable...

But, as we've found out on many occasions, regression testing is
absolutely necessary to uncover various problems relating to the
compiler under test. With a good compiler vendor who's interested in
improving the quality of their product, the discover bug - report to
vendor - bug is fixed cycle isn't too painful.

Regards,
--

  ______________________________________
   Andrew E. Kalman, Ph.D.   aek@pumpkininc.com

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.