Re: A lesson for compiler warning writers

tmb@arolla.idiap.ch (Thomas M. Breuel) writes:
> [ re using lint programs]
> That's not really a "much easier [...] choice". Lint programs tend to run
> quite slowly. That means that either your edit-compile-test cycle will
> become much slower, or that you won't run lint on every compilation.


Slowness is a problem, but correctness is far more serious. There are
many instances of lints having different grammars than their associated
compilers. Running a lint program is not always a good option. I have
worked in some environments (such as S.C.O. Xenix) where the compiler and
lint program accepted different grammars (compiler was sort of Ansi
Compliant, lint was K and R, in this case lint could not handle
prototypes).


You have to be sure that the tool that does the checking is 100 %
compliant with the compiler. I think the best way to do that is to make
the compiler rigorous in its checking. This also serves the purpose to
make sure that the compiler user knows that there is something about the
code that the compiler does not like.


Bill
--
| maniattb@cs.rpi.edu - in real life Bill Maniatty
--

Return to the comp.compilers page.
Search the comp.compilers archives again.