Suite-part list is wrong

[dustinswales]

Description

The "suite part list" order (e.g. group order) in the host API are sorted alphabetically, not as written into the Suite Definition File.

Steps to Reproduce

Using the capgen_test, change physics1 in temp_suite.xml to radiation, run the tests and look at the suite_part_list in the host Cap:

The order is incorrect

[gold2718]

I'm not sure "incorrect" is accurate here. There was no requirement for this before I wrote it (did prebuild even have this feature before?).
Inconvenient might be a better term.

[climbfuji]

Is this a problem? For example, when running through all groups (when the ccpp_physics_run call is executed without group=...), does it follow the order in the list? If that were the case, then it would be a problem.

prebuild doesn't use a list, but it always follows the order of the groups and schemes in groups.

[gold2718]

Is this a problem?

Absolutely not. It was not important before but if someone is going to rely on an ordering, we just need to implement the desired ordering, document it and test it (which I guess #633 is doing). Using the SDF ordering seems sensible to me.

[climbfuji]

I think it is imperative that when running through all groups in one go, the order in the suite is maintained. In any and all of the GFS suites, radiation runs first, then the "other physics". There are strict dependencies between these groups (radiation does certain things that the other physics rely on). That applies to all phases, too (init, timestep_init, ...).

[dustinswales]

Ditto what Dom said. This and #629 are critical to ensure the SDF matches the suite cap