Small fixes for test cases that fail under gfortran

[sit23]

As described in #221, some of the test cases are failing when run under gfortran. I've identified and fixed two of these 3 issues, and am working on the 3rd. The changes are as follows:

• The axisymmetric case was failing because we were supplying dt_rad as a float when it should have been an integer. Intel didn’t mind this, but gfortran objects.
• The Top down test seg-faulted because one of the loops over the y dimension was looping over the length of the x dimension instead
• I’ve had a look at the giant-planet test case to see if I can see anything obvious. I’ve made a few changes that might help, but I’d need to get a working example of what the problem is before I can make too much progress.

I'm running the trip tests on these now to see what progress I might have made, but as they all failed previously, this is more just to check that they now run.

[RuthG]

Nice! Looks good at a glance, happy to review next week once the trip tests confirm it's all working :)

[sit23]

I've finally got back to running the trip tests properly for this p/r. I've run the tests for an intel compiler and the gfortran compiler. The results for the intel compiler are as follows:

Results for all of the test cases ran comparing 403f230 and 809b9a3 are as follows...
axisymmetric : pass
bucket_model : pass
frierson : pass
giant_planet : pass
held_suarez : pass
MiMA : pass
realistic_continents_fixed_sst : pass
realistic_continents_variable_qflux : pass
socrates_aquaplanet : pass
socrates_aquaplanet_cloud : pass
top_down_test : pass
variable_co2_grey : pass
variable_co2_rrtm : pass
ape_aquaplanet : pass
Congratulations, all tests have passed

All the tests have passed, so these modifications make no difference to the running of the code under intel compilation.

[sit23]

I also ran the tests under gfortran compilation:

Results for all of the test cases ran comparing 403f230 and 809b9a3 are as follows...
axisymmetric : pass
bucket_model : pass
frierson : pass
giant_planet : pass
held_suarez : pass
MiMA : pass
realistic_continents_fixed_sst : pass
realistic_continents_variable_qflux : pass
socrates_aquaplanet : fail
socrates_aquaplanet_cloud : fail
top_down_test : fail
variable_co2_grey : pass
variable_co2_rrtm : pass
ape_aquaplanet : fail
Nightmare, some tests have failed

There were 4 fails, all of which are because the current version of the master (403f230) fails to run those test cases with gfortran. But all of these problems should be fixed by this p/r. The other tests passing is further confirmation that these modifications work alright. I'm therefore happy that the trip tests are clean, and it'd be great if these changes could be merged in soon.

[sit23]

@RuthG - now that I've put up the trip test results, would you be happy to review this for me?

[RuthG]

Looks good to me!

[sit23]

Thanks @RuthG - would you mind doing a review and approving it so that I can merge it in? Thanks!

[RuthG]

Changes are minor bug fixes, confirmed not to influence trip tests on intel compilers. I am happy for these to be merged in, thanks for spotting it all Stephen! :)

[sit23]

Thanks @RuthG