t_test `can't test a constant constraint`

I'm trying to use t_test to get the covariance for a degenerate linear transformation, but patsy complains.

question: Should this be allowed?

Note: for a while now I have been using t_test to get the results for linear transformations and not just for "t-tests"

    tt = res3.t_test(transf3.transf_mat, transf3.constant.squeeze())
  File "e:\josef\eclipsegworkspace\statsmodels-git\statsmodels-all-new2_py27\statsmodels\statsmodels\base\model.py", line 1050, in t_test
    LC = DesignInfo(self.model.exog_names).linear_constraint(r_matrix)
  File "build\bdist.win32\egg\patsy\design_info.py", line 220, in linear_constraint
  File "build\bdist.win32\egg\patsy\constraint.py", line 392, in linear_constraint
  File "build\bdist.win32\egg\patsy\constraint.py", line 58, in __init__
ValueError: can't test a constant constraint

>>> transf3.transf_mat
array([[ 0.,  1.,  0.],
       [ 0.,  0.,  1.],
       [ 1.,  0.,  0.],
       [-0., -0., -0.],
       [-0., -0., -0.]])
>>> transf3.constant.squeeze()
array([ 0.,  0.,  0.,  0.,  1.])
>>> res3.model.exog.shape
(20, 3)

example is for constraint estimation, the covariance matrix in the constraint full model (parameterized for original exog) is degenerate or rank reduced.

>>> cov_params3 = transf3.transf_mat.dot(res3.cov_params()).dot(transf3.transf_mat.T)
>>> np.sqrt(np.diag(cov_params3))
array([ 0.59285072,  0.31334777,  0.30669526,  0.        ,  0.        ])

>>> print res3_ols.bse
[ 0.59285072  0.31334777  0.30669526]

(example based on "dirty" development script for estimation with linear constraints)

Start in statsmodels/base/model.py at Results.t_test, then follow its call to patsy's DesignInfo.linear_constraint. Reproduce the shown rank-reduced constraint with the supplied matrices and determine whether constant constraints should be accepted; done means the intended covariance and constraint results work without patsy rejecting the input.