Test for use of keep_lattice in find_orbit6().

[willrogers]

This is how I expect to use the keep_lattice keyword argument, and it appears to be failing in pyat-0.2.0.

The second call (with keep_lattice=True) is returning nan in each element of the array.

Is this related to #299 ?

cc @MJGaughran

[willrogers]

I see that this test starts failing in 4f21954.

Is this a valid test and something needs fixing? Or am I misunderstanding something?

[swhite2401]

I think this is a problem that needs fixing, @lfarv, any idea?

[lfarv]

@willrogers: what you try is exactly how it is supposed to work.

I see the problem, and strangely find_orbit6 is the only affect one: find_orbit4, find_m44, find_m66 work as expected.
So I am looking at the problem ! Thanks for the hint on the origin of the problem.

[willrogers]

I have been debugging this morning.

The difference is calling get_timelag_fromU0() with ELossMethod.TRACKING instead of ELossMethod.INTEGRAL.

The problem is the call in energy_loss.py line 66:

        ringtmp = ring.radiation_off(dipole_pass=None,
                                     quadrupole_pass=None,
                                     wiggler_pass=None,
                                     sextupole_pass=None,
                                     octupole_pass=None,
                                     copy=True)

I think this call must be modifying the original lattice somehow.

[lfarv]

@willrogers: well spotted, thanks. get_energy_loss does not change the original lattice, but it does track with a different one. So if it is used in find_orbit6 (case of no initial guess and method ELossMethod.TRACKING, I'll have to force keep_lattice to False. Not a big deal, but then you test is irrelevant. The best would the to change the test to use find_orbit4, where it makes sense, while I'll fix find_orbit6.

[willrogers]

I added a warning if you pass keep_lattice when it is not valid. That might be a bit noisy so I can remove that commit if you like.

I update the find_orbit6() test and added one for find_orbit4().

[lfarv]

I don't think the message is useful: keep_lattice is a hint given to try and save some time (not a lot, in fact…). How this is honoured in the function does not really matter since it does not impact the result at all. So I would remove the warning.