Re: programmed gc mark and free sweeps...

[Michael Livshin]

[ sorry for taking so long to answer ]

"Lars J. Aas" <address@hidden> writes:

> On Thu, Sep 14, 2000 at 04:49:08PM +0300, Michael Livshin wrote:
> : 
> : what kind of debugging do you have in mind?  it could probably be
> : achieved through more regular means.
> 
> I was just thinking of checking that scm_gc_mark on some widget smobs recursed
> correctly through smobs attached to widgets and subwidgets.  I figured if I 
> do a
> "global (gc)", the debug output might become too large if the system is large
> (which is not the case yet), while isolating and invoking mark on only one 
> smob
> would give me only the info I need.

I see.

so what would be really nice to have is a way to traverse arbitrary
data graphs beginning from some object.

the same functionality is needed for implementing persistence and
other similar things.

I can see no nice way to implement it with the current smob facility.
we should move away from smobs at some point, though...

[ on the other tentacle, you probably have a limited number of smob
  types defined.  so debug their mark functions once and forget about
  the whole thing ;) ]

-- 
I think we might have been better off with a slide rule.
                -- Zaphod Beeblebrox