Re: [Sbcl-devel] Re: Bug#273606: sbcl: (proclaim '(optimize (debug 3))) and aserve compilation

[Nikodemus S. <tsi...@cc...>]

On Tue, 28 Sep 2004, Kevin Rosenberg wrote:

> I've reproduced your results. I'm not convinced that SBCL is operating
> correctly when DEBUG is set to 3:

I believe this is STEP's fault. Unless someone gets there first, I'll 
investigate this later this week / early next week. "Should be simple."

>> ; compiling DEFUN READ-HTTP-REQUEST:
>> debugger invoked on a SB-INT:BUG in thread 18142:
>>     full call to SB-KERNEL:DATA-VECTOR-SET

...but this actually looks related to bug #302. I wonder if this is a case 
of two separate bugs, the new #302-lookalike being triggered by the 
stepper bug.

> call to SB-KERNEL:DATA-VECTOR-SET when (OPTIMIZE (DEBUG 3)). Also,
> hopefully someone on that list knows about the general
> suitability/caveats of using DEBUG 3.

DEBUG 3 when SIZE & SPEED are less then DEBUG introduces a significant 
performance penalty and code bloat these days (say a factor of x5 compared 
to what it used to be?), due to the new instrumentation based single 
stepper.  Bugs there are just that --bugs-- and hopefully will be weeded 
out soonish -- but the performance penalty is likely to stay. Then again, 
as long as the performance is not a problem, I see no reason to avoid 
DEBUG 3.

Cheers,

  -- Nikodemus              Schemer: "Buddha is small, clean, and serious."
                   Lispnik: "Buddha is big, has hairy armpits, and laughs."