Re: [Chicken-hackers] [PATCH] catch exceptions in finalizers, remove dynamic resizing of finalizer vector

[Felix]

>> This means the number of finalizers is limited to (currently) 4096. Code
>> that produces many finalizers must make sure they are triggered (and
>> thus un-registered) fast enough.
> 
> Ick, ick, ick.  This is a horrible solution.  The main use of finalizers
> that I've ever had is when dealing with objects allocated by C libraries;
> a pointer object wrapped in a finalizer insures that when the foreign
> object is no longer interesting to the Scheme side, it is properly freed,
> which may or may not involve calling free().

It probably isn't even a solution. I don't disagree with you here.

> 
> Having a fixed low limit on the number of finalizers means that this
> sort of design is not practical: you wind up being better off with
> larger-grain foreign objects and more C code to access their insides.
> That's un-Schemey.

It probably is.

> 
>> Note that the "-:f" runtime option can be used to change the number
>> of available finalizers.
> 
> This helps a little, but not much, because it's basically guesswork what
> to set the option to, and it may vary greatly depending on the input,
> leading to programs that work fine and then suddenly crash.

Yep. Just like running out of memory. There always will be some hard
limits.

> 
> Why does there have to a be a persistent array of all finalizers anyway?
> The garbage collector itself can find them during the mark phase.

It is much more complicated than that (see my reply to Peter).


cheers,
felix