Re: [Chicken-users] scrutiny

[Peter Bex]

On Fri, May 29, 2009 at 04:18:55PM +0200, felix winkelmann wrote:
> Hello!
> 
> I have merged the recent work on the "scrutiny" branch into trunk (r14828).
> "Scrutiny" is a lightweight local conservative (read: simple) flow analysis
> pass over the internal representation of the code and performs a check on
> argument-count/-types and results of calls to core library procedures.
> Additionally, the number of result values is checked in various places.

This is really great!

> It is likely that the type-database still contains some bugs, so if you
> try it out and get strange warnings, please mail them to me.

The one thing that I keep getting when I compile
is warnings that procedures return a different number of arguments in
branches like this:

(if (number? foo)
    (values foo (add1 foo))
    (error "foo is not a number"))

I think procedures like ERROR, RAISE etc that can invoke different
continuations should probably be treated specially.

Also, it gives errors for each branch of a COND:

(cond ((number? foo) (values foo (add1 foo)))
      ((string? foo) (values foo (string-append foo " etc")))
      (else (error "blah")))

This is a bit annoying in 9p for example, where there's a procedure
with 7 cases and one error in the ELSE branch, it gives 7 warnings :)

> Usage is simple: just pass "-scrutinize" (or "-S" in csc).

Why is it not enabled by default?  It only issues warnings, so even if
it's wrong, it'll still compile the code properly...

Cheers,
Peter
-- 
http://sjamaan.ath.cx
--
"The process of preparing programs for a digital computer
 is especially attractive, not only because it can be economically
 and scientifically rewarding, but also because it can be an aesthetic
 experience much like composing poetry or music."
                                                        -- Donald Knuth

pgphoubievnet.pgp
Description: PGP signature