Port complete. All tests pass on Jenkins on Python 2 (including lsst_ci). Python 3 passes all tests except for a test in random.py where setState crashes because the bytes from getState are being interpreted as a unicode string.

Can you please review this (you seem to already be looking at the pull request)?

I have a proposed fix for the Random::getState/setState issue on u/jbosch/DM-7152. I have not had a chance to test it, and likely will not before next week, but it should at least provide a skeleton for a solution. One thing that definitely needs to be tested: what happens when something other than bytes is passed to setState; the Python C API code I've written will correctly raise TypeError, but I'm not sure if Swig will call PyErr_Occurred and return null to finish the error propagation when that happens.

Jim Bosch works great and tests pass. If I try to pass in a string instead of bytes you guessed the outcome:

TypeError: expected bytes, str found

During handling of the above exception, another exception occurred:

Traceback (most recent call last):

  File "tests/random1.py", line 121, in testState

    self.rand.setState("hello")

  File "/Users/timj/work/lsstsw3/build/afw/python/lsst/afw/math/mathLib.py", line 8425, in setState

    return _mathLib.Random_setState(self, state)

SystemError: <built-in function Random_setState> returned a result with an error set

Less than ideal but I doubt it matters much in the short term.

Looks good. Some minor issues on the pull request.

Thanks. I've pushed some fixes for your comments. I'm not going to attempt to fix the _future_ issues in this commit. I am running a Jenkins build now. I probably should squash some of the fixes that I have just done and then I'll try to merge tomorrow before the afternoon session.

Tim Jenness discovered an intermittent problem in tests/testCameraGeom.py in the testTransformDet method: it would fail 1 in 5 or 10 times due to the order of dict keys being returned. The obvious fix that Tim Jenness suggested is to test all detectors instead of a subset. I adopted that fix and added some comments to explain more of what is going on. I think the test could use additional improvements (e.g. a more deterministic decision if a point off one detector should be found on another detector), but I think the current test is adequate (with this fix).

Thanks to everyone for helping me out on this. Everything passes. Merged!