> I wouldn't know if this statement is true, but to say that it's
> better to use another controller for C-based programs is totally wrong.
> There are C-compilers available for PICmicros that are "bug-free".


I have been following this thread with considerable interest. The
compiler developers are faced with several major issues related to
bugs, features and optimization.  When we are aggressive with our
optimization then the likelihood of bugs increases. The same is
true with features. Bugs are induced in the compilers when we
change implementation strategies on new processors or processors
where there is not a large body of big completed applications to
understand the subtle issues. The early MPC development saw
a lot of this until experience with the PIC processor families
stabilized the compiler. Interestedly the current support for the
16C66, '67, '76, '77 required the knowledge gained from the first
half dozen or so applications now going into production.

Testing, as several have pointed out, determines only what doesn't
work. It is a very tough problem. The tests we use come from four
sources. We buy commercial test suites, two separate groups in
Byte Craft write test suites (developers and customer support
people), We contract out the writing of specific test suites to outside
companies specializing in this type of work and we use the publicly
available test suites and example code.

Our tests come from