From: billtj@... (Bill Tj)
Date: 2002-08-13T01:28:01+09:00
Subject: Re: Are All C Destructors Guaranteed to Be Called?

Hi,

Yes, after performing repetitive test (i.e., creating the objects and
"resetting" them many-many times), I could see some memory leaks. 
After the leaks were fixed, then when I tried the simple test case, I
could see that all the object "free" functions were indeed called.

Thanks for your assertion.  Now I conclude that if a C extension is
implemented correctly, all the object free functions should be called
before Ruby exits.

Regards,

Bill

=============================================================================
Paul Brannan <pbrannan@atdesk.com> wrote in message news:<20020809123333.M11837@atdesk.com>...
> On Fri, Aug 09, 2002 at 11:24:27PM +0900, William Djaja Tjokroaminata wrote:
> > Are all C destructors, i.e., the "free" function that is specified in
> > Data_Make_Struct (VALUE, c-type, mark, "free", c-type*), guaranteed to be
> > called when the Ruby script/process exits?
> 
> They get called when the object needs to be destroyed.  I think all
> objects are supposed to be destroyed at exit.
> 
> > I put "printf" on all the "free" functions, and one of them did not get
> > called.  Is it a sign that something is still not clean?  How about
> > circular reference (I think I have one in my code)?  Will circular
> > reference cause the free function not called at the end of the
> > script/process?
> 
> 1) Can you reproduce this behavior in a small test case?
> 2) Ruby uses mark/sweep instead of reference counting, so circular
>    references won't cause an object to not get GC'd.  What *can* happen
>    is that an object doesn't get destroyed because someone is keeping a
>    reference who is not supposed to be keeping a reference, but both
>    objects should be destroyed at exit.
> 
> Paul