Pandas 32bit with recent matplotlib 2.1.1 MemoryError: In draw_markers: Out of memory

Code Sample, a copy-pastable example if possible

While building pandas packages for Debians, where we set matplotlib's backend to Agg, I ran into this failure:

=================================== FAILURES ===================================
____________________ TestSeriesPlots.test_valid_object_plot ____________________

self = <pandas.tests.plotting.test_series.TestSeriesPlots object at 0xdd879eac>

    @pytest.mark.slow
    def test_valid_object_plot(self):
        s = Series(lrange(10), dtype=object)
        for kind in plotting._core._common_kinds:
            if not _ok_for_gaussian_kde(kind):
                continue
>           _check_plot_works(s.plot, kind=kind)

../debian/tmp/usr/lib/python2.7/dist-packages/pandas/tests/plotting/test_series.py:726: 
...
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <matplotlib.backends.backend_agg.RendererAgg object at 0xde04a76c>
kl = (<matplotlib.backend_bases.GraphicsContextBase object at 0xe02e854c>, Path(array([[ 0.,  0.],
       [ 0.,  0.],
     ...(array([[ 0.,  0.]]), None), IdentityTransform(), (0.12156862745098039, 0.46666666666666667, 0.70588235294117652, 1.0))
kw = {}

    def draw_markers(self, *kl, **kw):
>       return self._renderer.draw_markers(*kl, **kw)
E       MemoryError: In draw_markers: Out of memory

/usr/lib/python2.7/dist-packages/matplotlib/backends/backend_agg.py:117: MemoryError
=============================== warnings summary ===============================

happens only on recent debian/ubuntus and only with 32bit userland build

neurodebian@smaug ~/deb/builds/pandas/0.22.0-1 % grep 'In draw_markers: Out of memory' *build             
pandas_0.22.0-1~nd100+1_i386.build:E       MemoryError: In draw_markers: Out of memory
pandas_0.22.0-1~nd17.10+1_i386.build:E       MemoryError: In draw_markers: Out of memory
pandas_0.22.0-1~nd+1_i386.build:E       MemoryError: In draw_markers: Out of memory

I guess I will just skip this test for now, but wanted to let you know about it

Comment From: gfyoung

cc @TomAugspurger

Comment From: mroeschke

Thanks for the report but appears this is with a fairly old version of pandas and matpliotlib. We can reopen if it this is reproducible with a new versions of those libraries