At 2:17 PM -0700 4/25/02, Steve Fink wrote:
>Thank you!
No, thank you, for a rather entertaining and depressingly dead-on message. :)
>I think I more or less understand things now. Once you make
>your changes and I'm convinced that I really do know what's going on,
>I'll try to find time to write things up properly.
I should have mark capability for PMCs in the repository in a little
while, when the tests finish running. I'm yanking a couple of vtable
entries while I'm at it, as the GC entries make no sense. (I tossed
collect and move_to, as we never actually process the PMCs when we do
a GC run, so these are meaningless. I may toss real_size as well)
Allocation of tracked objects (subclass of Buffer) will be next, but
I want to take things in pieces so I can break things properly. ;-P
>But it does seem like there's something missing: general fixed size
>tracked objects. They would only have a 'flags' field in common, I
>guess, but there could be a general setup_pool for a given size. Then
>the whole hierarchy would look like:
>
>TrackedObject
> PMC
> Buffer
> STRING
The hierarchy should be:
PMC
Buffer
TrackedObject
String
Since things are either PMCish or Bufferish. Or so my thinking goes,
but we've seen the result of that...
>I'll probably have more questions once I really start using this
>stuff properly.
Cool. If it turns out some of the design decisions are, well, stupid,
let me know. I'll fix 'em.
--
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
