Issue with numba=0.43.1 with scalar maximisation

[mmcky]

It appears that scalar_maximisation.py fails in nopython mode for numba=0.43.1 which is the current version of numba in anaconda.

travis-ci installs via pip which is using 0.44 which doesn't seem to have this issue during the ci test.

Should we just wait for anaconda to update numba?

[mmcky]

numba.errors.LoweringError: Failed in nopython mode pipeline (step: nopython mode backend)
'NoneType' object has no attribute 'groups'

File "../../../../../../../../anaconda3/lib/python3.7/site-packages/quantecon/optimize/scalar_maximization.py", line 50:
    ... def f(x):
        <source elided>
    if not np.isfinite(a):
        raise ValueError("a must be finite.")
        ^

[1] During: lowering "raise <class 'ValueError'>('a must be finite.')" at /home/qebuild/anaconda3/lib/python3.7/site-packages/quantecon/optimize/scalar_maximization.py (50)
[2] During: resolving callee type: type(CPUDispatcher(<function brent_max at 0x7ff85282ce18>))
[3] During: typing of call at <ipython-input-5-3f7c7f7b0dfc> (30)

[mmcky]

This does also raise the question - perhaps we should swap travis over to using conda just so the two are aligned?

[mmcky]

this is no longer an issue in anaconda as the latest release has 0.45.1.

[mmcky]

closing this as travis is documented here