Re: [PATCH - policyrep] add objpool to libsepol

[James Antill <jantill@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1591 bytes --]

On Thu, 2007-04-19 at 12:11 -0400, Karl MacMillan wrote:
> On Thu, 2007-04-19 at 11:57 -0400, James Antill wrote:

> >  I guess I just wonder why have the s->obj_free member at all, esp.
> > given the examples used with strdup() ... I kind of assumed you expected
> > input from strdup(), given the examples, and so was confused about the
> > non-useage of free().
> 
> Ahh - but the objects can be arbitrary, so they might be structs that
> need something more than free.
> 
> >  It also just seems wrong to have the member free function used for
> > input ... but if you have a use case feel free to ignore me :).
> 
> Well, the model I was thinking was that the caller was giving ownership
> of the object to the pool. The advantage is that the caller simply calls
> add and doesn't have to check return codes to see if their copy of the
> object should be freed.

 Well it needs to check for errors, and do something different then.

>  Otherwise the caller would conditionally free
> the object depending on whether another copy already existed. Object
> deletion would have the same problem.

 Yeh, I understand ... it feels kind of wrong to be mixing layers. Ie.
conceptually you seem to have:

struct higher_layer {
  void *(*make)(void); /* allocate */
  struct lower_layer {
    void (*free)(void *); /* deallocate */
  } lower;
};

...which I don't see often, and so assumed was wrong ... but it works,
you obviously meant to do it, and I can't see a quick way to split the
layers without downsides.

-- 
James Antill <jantill@redhat.com>

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]