BUG: integrate: Exception raised by odeint when t=[] and full_output is True.

[WarrenWeckesser]

An invariant of a call of odeint(func, y0, t) is that the output has shape (n, d) where n is len(t) and d is len(y0). This includes the edge case where t = []:

In [41]: odeint(lambda x, t: -x, [3], [])                                                                               
Out[41]: array([], shape=(0, 1), dtype=float64)

However, if full_ouput is True in this case, an exception is raised:

In [42]: odeint(lambda x, t: -x, [3], [], full_output=True)                                                             
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-42-46d3f4c22fe0> in <module>
----> 1 odeint(lambda x, t: -x, [3], [], full_output=True)

~/mc37allpip/lib/python3.7/site-packages/scipy-1.7.0.dev0+624fd76-py3.7-macosx-10.9-x86_64.egg/scipy/integrate/odepack.py in odeint(func, y0, t, args, Dfun, col_deriv, full_output, ml, mu, rtol, atol, tcrit, h0, hmax, hmin, ixpr, mxstep, mxhnil, mxordn, mxords, printmessg, tfirst)
    255                              full_output, rtol, atol, tcrit, h0, hmax, hmin,
    256                              ixpr, mxstep, mxhnil, mxordn, mxords,
--> 257                              int(bool(tfirst)))
    258     if output[-1] < 0:
    259         warning_msg = _msgs[output[-1]] + " Run with full_output = 1 to get quantitative information."

ValueError: negative dimensions are not allowed

The problem is that when full_output is True, a dictionary containing information about the behavior of the solver is created, and some of the values in this dictionary are arrays with length n - 1. When t = [], n is 0 and the attempt to create the arrays for the dictionary fail.