On Wed, Mar 05, 2008 at 01:59:53PM +0000, Julian Pullen wrote:
> 
> 
> Nicolas Williams wrote:
> >On Tue, Mar 04, 2008 at 11:51:52AM -0600, Nicolas Williams wrote:
> >>More later.
> >
> > - dbutils.c:3826-3864
> >
> >   I can't figure out what this is for, nor why it's needed in the first
> >   pass instead of the second pass.  But if it's really needed in the
> >   first pass then the corresponding code to call ns_lookup_byname() in
> >   pid2sid_second_pass() needs to go, and anyways 3826-3864 should move
> >   to at least 3856.
> 
> I added this as I found that we were storing mappings in the cache that
> did not have a UID or GID. I did it before lookup_wksids_pid2sid() because
> this requires a PID to work.
> 
> It was caused by "idmap show" using names.

Ah.  OK, that makes sense.

> >
> >
> > - dbutils.c:4193-4195,4327-4329
> >
> >   Could we avoid copying idres.info given that we're about to free it?
> >
> We are copying from &idres.info to &mapping->info and freeing &idres.info
> 
> I could do a move rather than a copy?

That's what I'm suggesting.

> > - Sepaking of failure observability, what changes might we need to the
> >   protocol and schema?  Can we make those changes now so we don't have
> >   to make them again later?
> The protocol is complete. There will be no changes for phase 2.

Do we not need a method-specific error code?

Reply via email to