[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [csmith-dev] The answer to some overdue requests

On Sun, 31 Mar 2019 at 19:48, Arthur O'Dwyer <arthur.j.odwyer@gmail.com> wrote:
> On Sun, Mar 31, 2019 at 10:04 PM John Regehr <regehr@cs.utah.edu> wrote:
> > On 3/31/19 7:04 PM, Eitan Adler wrote:
> > >
> > > It would be interesting to be able to request cyclomatic complexity or
> > > other such measures. I say this not because it would be useful for
> > > finding C compiler bugs itself - but might be interesting to validate
> > > (or counter the usefulness) of such measures in the future.
> >
> > Cyclomatic complexity and friends seem very 70s. Does anyone use or
> > believe in that kind of thing now?
> At least I think I'm confused about what Eitan means by "usefulness." Is the implication that some codebase might institute a limit on cyclomatic complexity in the (IMHO mistaken) belief that programs with low cyclomatic complexity are less likely to trigger compiler bugs?  I don't think such a belief is either justified or widespread.

I think I sent a short note and in doing so didn't actually write what
I meant.  Rewriting my original comment:

It would be interesting to think about what other measures of software
complexity (of which CC is but one ancient one) are either
- predictive of code that finds compiler bugs
- limits the chance of production running into existing bugs

Other possible metrics include the originally mentioned - number of
AST nodes, lines of code, or others which I have no idea about.

Eitan Adler