> > I would think that the bucket memory allocator and the bucket memory
> > deallocator would have to be consistent.  Put it in the pool.
> 
> Huh?  What pool?

The pool that would need to be in the bucket structure if you want to support
morphing of buckets.  Never mind.

The short answer is: don't morph buckets.  The long answer is that free is
a function of how the bucket content was allocated and not a function of
its behavior as a bucket type.  So either move free out of apr_bucket_type_t
or compound the number of bucket types by the number of potential free
mechanisms for those types.

....Roy

Reply via email to