Re: [Condor-devel] Condor disk size

[Peter Keller <psilord@xxxxxxxxxxx>]

On Thu, May 26, 2011 at 04:45:12PM -0500, Jaime Frey wrote:
> On May 20, 2011, at 8:43 AM, Timothy St. Clair wrote:
> 
> > On Fri, 2011-05-20 at 07:44 -0400, Matthew Farrellee wrote:
> >> On 05/20/2011 07:26 AM, Brian Bockelman wrote:
> >>> 
> >>> On May 19, 2011, at 10:08 PM, Matthew Farrellee wrote:
> >>> 
> >>>> Ok to assume elimination of --allow-shlib-undefined is left as an exercise for the hopelessly heroic?
> >>>> 
> >>> 
> >>> Hopelessly heroic.
> >>> 
> >>> Actually, it'd just take an expert with a scalpel, some time, and permission to cause a temporary mess.  A lot of the issues looked relatively mild.
> >> 
> >> If you're bored some Saturday and want to take a stab at it, let me know.
> > 
> > As someone who has "taken a stab at it", and is still mildly sheepish
> > given all the knife wounds :-/ , I may be able to offer some breadcrumbs
> > on where I left off.  The last thing I was considering was to dissect
> > the headers in condor_includes and parcel them out into their
> > appropriate library.  Given the nature of some libraries you may even
> > want to push the deps into yet another library for isolation (Dan & I
> > did this with std:u as not to pollute condor proper).  I was trying to
> > hut for a tool helped to isolate circular deps, but then I decided to
> > just focus on build transformation 1st.  
> > 
> > With that last bit, I will back away slowly from this topic, turn tail,
> > and run like the wind. 
> 
> 
> I also took a stab at it a couple months back. I was doing pretty well until I found that utils depends on sysapi, privsep, procd_client, and procapi, but those four use dprintf() and param(), which is in utils. At that point, I gave up. I still have the branch sitting around, if anyone is interested.

The deal is that config/dprintf need to be its own lowest level library
that all others build upon.

-pete