[PATCH] Explain a bit of how loadparm works

[Michael Adam]

On 2013-10-16 at 10:27 +0200, Michael Adam wrote:
> On 2013-10-16 at 09:00 +0200, Volker Lendecke wrote:
> >
> > One thing sticks out
> > immediately though is the requirement to use talloc_unlink.
> > This needs to go ASAP. This is not acceptable. Sorry to be
> > so strict, but talloc_reference/talloc_unlink is just a
> > complete no-go in central code. Please focus on removing
> > that very, very soon.
> 
> I absolutely agree generally, but I think talloc_unlink was
> mentioned here just by habit instead of talloc_free:
> Looking at the implementation of loadparm_init_s3():
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> struct loadparm_context *loadparm_init_s3(TALLOC_CTX *mem_ctx,
> 					  const struct loadparm_s3_helpers *s3_fns)
> {
> 	struct loadparm_context *loadparm_context = talloc_zero(mem_ctx, struct loadparm_context);
> 	if (!loadparm_context) {
> 		return NULL;
> 	}
> 	loadparm_context->s3_fns = s3_fns;
> 	return loadparm_context;
> }
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> it seems that talloc_free() woule be safe on the return value, right?

Oops, apparently not right:

in lpcfg_gensec_settings(), a
settings->lp_ctx = talloc_reference(settings, lp_ctx);
is done...

We should really get rid of this!

Cheers - Michael

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 215 bytes
Desc: Digital signature
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20131016/e2872bdd/attachment.pgp>