[Openmcl-devel] GUI element needs

[Hamilton Link]

The evaluator might be born portable, but I think the compiled route 
would be more desirable since it doesn't affect the performance of 
Everything Else.  But the "better" the compiler (as far as inlining, 
constant-expression folding, expression rewriting, instruction 
shuffling, multiple architectures, etc.) the harder the problem, 
although I suppose under sufficient-for-stepping debug settings a lot 
of the code transformations could be disabled through compiler advice.

I've always been a print-statement man (coupled with backtrace for 
errors and the inspector for data structure checking), in part because 
of lack of stepping/break-point tools and in part because in most 
systems you can step forever and never get through the reams of code 
that precede the bug you're looking for (not to mention the 
difficulties of stepping through code in one thread while other threads 
are running full speed).  But I did use the MCL stepper from time to 
time.

On Aug 8, 2005, at 5:35 PM, Gary Byers wrote:

>
>
> On Mon, 8 Aug 2005, Hamilton Link wrote:
>
>> Also maybe GB can at least remind us what the state of the stepper is 
>> in openmcl (even if "absent").
>
> It's absent, though it used to be present (and there was even a GTK
> frontend for it, which almost certainly wouldn't work anymore.)
>
> The old stepper depended on the old evaluator, which was suffering
> severe bitrot.  I don't remember what the final straw was (maybe
> some MOP things), but it didn't seem like it'd be maintainable any
> longer.   In order to use the stepper, you had to either (a) turn
> off CCL:*COMPILE-DEFINITIONS* before defining anything that you
> wanted to step through of (b) save the lambda expressions along
> with compiled functions and hope that macros/constants that were
> in effect at compile-time were still around at "run evaluated under
> the stepper" time.
>
> The handwavy plan (for the last 15 years or so) has been to instrument
> compiled code with
>
>    (trap-if [stepping-enabled-in-current-thread])
>
> sequences at strategic points and under either default or
> slightly-oriented-towards-debugging optimization settings.  That's
> a fair amount of work; you'd also want to map chunks of the source
> lambda to those "strategic points" - I think that Alan Ruttenberg's
> done some work on this in the past - and there are other issues.
> I thought - and still think - that it'd be better to pursue the
> idea of enabling stepping through compiled code than it would be
> to try to keep a (rarely used and not always practical to use)
> evaluator around.
>
> _______________________________________________
> Openmcl-devel mailing list
> Openmcl-devel at clozure.com
> http://clozure.com/mailman/listinfo/openmcl-devel
>