Re: [Chicken-users] valgrind - more details

[Jörg F . Wittenberger]

On Oct 6 2011, Peter Bex wrote:

> > Furthermore: 
> > Also I changed my mind: I'm afraid numbers in Scheme are complex
> > enough to use a real parser to read them.
>
> I agree.  There's pure Scheme code for parsing numbers syntax in trunk
> of the numbers egg (and the latest release has it too).  It's not very
> simple either, but it doesn't have any dependencies.
>
> When I posted my patches to fix some of the more esoteric syntax details
> in Chicken, I initially also tried my Scheme code, but it is about twice
> as slow as this C code, that's why I decided against switching.

This sound like a good case to try
http://people.csail.mit.edu/jaffer/Docupage/schlep.html

> Also, this code needs to be callable from C because decode_literal()
> relies on it. (but that could maybe be worked around)

Did I admit in that other mail that I did only compile, but not test
this patch?  I'd better had done so!

In fact the first valgrind complain is gone. "42" without the quotes
is now a valid Scheme program wrt. the valgrind report.

> > However let me reiterate: it does not fix valgrind's compaint.

Wrong I was.

> I think the patch is a good improvement in readability of the core code,
> so why not put it in? (of course after fixing the while/if mistake)

So let's go for it.

The attached diff is against git master as of 5p.m. Berlin.

However the only thing, which this patch does wrt. program logic
is to avoid running afoul of a possibly missing end marker in the
strlen call.  This cast some doubt that the code might still need
revision.

/Jörg

cleanup.diff
Description: cleanup.diff