On Tue, Aug 24, 2010 at 04:12:54PM -0700, Shawn Walker wrote:
> On 08/24/10 03:46 PM, Edward Pilatowicz wrote:
> >On Tue, Aug 24, 2010 at 03:32:40PM -0700, Shawn Walker wrote:
> >>On 08/24/10 03:27 PM, Edward Pilatowicz wrote:
> >>>On Mon, Aug 23, 2010 at 10:21:22PM -0700, [email protected] wrote:
> >>>>Folks,
> >>>>Over time the pkg(5) source code has sprouted a number of different,
> >>>>independent file locking routines.  This leads to a maintenance hassle
> >>>>for the code, as any change to the locking code generates updates in
> >>>>various places, if any bugs are found.  This change makes the code
> >>>>common, but customizable, so differnet modules can implement different
> >>>>behaviors.  I fixed a couple of additional locking bugs that were
> >>>>discovered along the way.
> >>>>
> >>>>  http://cr.opensolaris.org/~johansen/webrev-14615/
> >>>>
> >>>>Thanks,
> >>>>
> >>>
> >>>for lock file contents we have:
> >>>- image.py: pid, pidname, and hostname.
> >>>- indexer.py: pid, hostname
> >>>- repository.py: pid, hostname
> >>>
> >>>why not have all the callers above save pid, pidname, and hostname and
> >>>then put just one copy of the get/set lock string callback functions in
> >>>lockfile.py.
> >>
> >>Image wants to save client name in addition to all of the above.
> >>
> >
> >sure.  but it seems like the others could save that as well and reduce
> >duplication.
> 
> The client name doesn't apply to the other classes.

What about a compromise?  I can create client_[get|set]_lock_str and
generic_[get|set]_lock_str.  Indexer and Repository can use generic,
image can use client, and cfgfiles will use None.

-j
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to