figure positioning problem

[John W. Eaton]

The other day on #octave, I was mistaken about the changeset that
caused the figure positioning problem for me.  It is actually

  # HG changeset patch
  # User Michael Goffioul <address@hidden>
  # Date 1318884362 -3600
  # Node ID d99aa455296e0ece82bddca44859d62344470dd4
  # Parent  f80473f245532a328f7c925a8c4845581c56d978
  Add outerposition property to figure objects.

I don't know how I got the wrong one before.  I was certain that I had
tested it.

In any case, the trouble seems to be that
figure::properties::get_outerposition is returning the initial value
of [-1 -1 -1 -1], so the function
figure::properties::get_boundingbox does not produce a reasonable
value.  Oddly, doing set (gcf, 'outerposition', [300, 300, 700, 500])
at the command line does not change the appearance of the plot, but
forcing that value inside get_boundingbox does.

Is the value of the outerposition property supposed to be set to
something other than [-1 -1 -1 -1] somewhere?

As I said on #octave, I only see the strange behavior with a single
figure, and unfortunately it is created in such a complicated way by
some code that I did not write and I can't see an easy way to extract
a simpler example to demonstrate the problem.  So I'm hopeful that the
above information is enough to see what the problem is.

jwe