[Top][All Lists]

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

Re: [Chicken-users] "picky" scrutinizer mode

From: Alan Post
Subject: Re: [Chicken-users] "picky" scrutinizer mode
Date: Fri, 10 Dec 2010 05:02:42 -0700

On Fri, Dec 10, 2010 at 03:22:20AM -0500, Felix wrote:
> From: Alan Post <address@hidden>
> Subject: Re: [Chicken-users] "picky" scrutinizer mode
> Date: Thu, 9 Dec 2010 20:11:28 -0700
> > On Thu, Dec 09, 2010 at 05:35:22AM -0500, Felix wrote:
> >> Hi!
> >> 
> >> I have added "-picky" to the experimental branch (I haven't created a
> >> change request, since the impact is minimal and this feature is purely
> >> optional). When given, the scrutinizer warns about undefined branches
> >> in conditionals in tail-position of global procedures that do not
> >> perform a self-call.  The scrutinizer does no inter-procedural
> >> flow-analysis, so there is no warning for this situation for local
> >> procedures which are called in tail-position (which, if I understood
> >> correctly, is what Joerg would have liked most).
> >> 
> > 
> > Is there a (declare ...) syntax to mimic this command-line option?
> No, not yet. I don't think using declarations for this is overly
> practical, to be honest, since this should be independent of the
> code.

Ah, interesting.  I keep all of my optimization settings in a
special file using (declare ...).  I've been meaning to ask how well these
correspond to the command-line options, whether there are (declare ...)
optimizations that aren't available from the command-line or command-line
optimizations not avaliable from (declare ...).

The counter-point to your assertion is that the code isn't the stuff you
test and run.  It is the compiled form of a program, on a particular
platform with a particular set of optimization settings that passes
the regression tests.  Since you've tested that combination, (the
optimization settings that improve the runtime or size of the code, and
don't make a metric you care about worse) they don't wind up being
independent of the released product.

.i ko djuno fi le do sevzi

reply via email to

[Prev in Thread] Current Thread [Next in Thread]