[Matplotlib-users] Incompatibility with axvspan and legend on a semilog plot

Hi all,
I am having a problem having both a vspan and a legend in a figure plotted
on a semilog axis. A simple code that gives the following error is shown in
red:

ERROR

    return self.frozen().__array__()
  File "/usr/lib/pymodules/python2.6/matplotlib/transforms.py", line 1051,
in __array__
    return self.frozen().__array__()
  File "/usr/lib/pymodules/python2.6/matplotlib/transforms.py", line 1706,
in frozen
    return blended_transform_factory(self._x.frozen(), self._y.frozen())
  File "/usr/lib/pymodules/python2.6/matplotlib/transforms.py", line 1875,
in frozen
    frozen = composite_transform_factory(self._a.frozen(), self._b.frozen())
  File "/usr/lib/pymodules/python2.6/matplotlib/transforms.py", line 1367,
in frozen
    return Affine2D(self.get_matrix().copy())
  File "/usr/lib/pymodules/python2.6/matplotlib/transforms.py", line 1451,
in __init__
    Affine2DBase.__init__(self)
  File "/usr/lib/pymodules/python2.6/matplotlib/transforms.py", line 1304,
in __init__
    Transform.__init__(self)
  File "/usr/lib/pymodules/python2.6/matplotlib/transforms.py", line 87, in
__init__
    self._parents = WeakKeyDictionary()
  File "/usr/lib/python2.6/weakref.py", line 232, in __init__
    def remove(k, selfref=ref(self)):
RuntimeError: maximum recursion depth exceeded while calling a Python object

CODE

x = numpy.arange(10,1e8,1000)
y = numpy.arange(0,100,0.001)

pylab.semilogx(x,y,'bo', label="Plot 1")

pylab.legend(loc='best')
pylab.axvspan(20,2e3,facecolor='k',alpha0.2)