1.2RC1 test_identity fails on arm64, powerpc, ppc64, ppc64el, s390x

[olebole]

This is an issue in the same function as #3486; however with a different output. It was hidden for a while since I had to skip wcs doctests due to #3380.

_________________ [doctest] astropy.wcs.wcs.WCS.all_world2pix __________________
2102         ...   print("Best solution:\\n{{0}}".format(e.best_solution))
2103         ...   print("Achieved accuracy:\\n{{0}}".format(e.accuracy))
2104         Indices of diverging points: [1]
2105         Indices of poorly converging points: None
2106         Best solution:
2107         [[  1.00000238e+00   9.99999965e-01]
2108          [ -1.99441636e+06   1.44309097e+06]
2109          [  3.00000236e+00   9.99999966e-01]]
2110         Achieved accuracy:
2111         [[  6.13968380e-05   8.59638593e-07]
Differences (unified diff with -expected +actual):
    @@ -6,5 +6,5 @@
      [  3.00000236e+00   9.99999966e-01]]
     Achieved accuracy:
    -[[  6.13968380e-05   8.59638593e-07]
    +[[  6.13968380e-05   8.59638595e-07]
      [  8.59526812e+11   6.61713548e+11]
    - [  6.09398446e-05   8.38759724e-07]]
    + [  6.09398446e-05   8.38759723e-07]]

astropy/wcs/wcs.py:2111: DocTestFailure

[astrofrog]

This should be fixed in #5080 - I'll mark this as 'Close?' to remind us it's probably solved.

[astrofrog]

@olebole - can you confirm this is fixed in the latest stable release?

[olebole]

Yes, this does not appear anymore and can be closed.

[astrofrog]

Thanks!