Throw exceptions on early CCDB error

[sdobbs]

Certain problems with the CCDB in DANA programs, like if an invalid variation is specified, can be difficult to diagnose. Such errors are printed to the screen with useful error messages, but are often lost because execution continues until some code segfaults later on and a long, mostly-useless stack trace is printed out.

This commit attempts to provide an early exit (and clearer representation of the error) in the case of these problems by raising exceptions in the HDDM event sources, which are called early on, if the CCDB tables can not be loaded.

[gluex]

Build status for this pull request: SUCCESS

Build log: /work/halld/pull_request_test/sim-recon^early_exit_on_ccdb_problem/make_early_exit_on_ccdb_problem.log
Build report: /work/halld/pull_request_test/sim-recon^early_exit_on_ccdb_problem/report_early_exit_on_ccdb_problem.txt
Location of build: /work/halld/pull_request_test/sim-recon^early_exit_on_ccdb_problem

[markito3]

Is runtime_error the standard thing we throw? Wait...do we have a standard thing to throw?

[sdobbs]

Oh yeah, I should probably be using a JException, shouldn't I?

[pmattjlab]

Well, I think there's probably two different things you would want to do: print a huge warning/error message, and then either default to some other variation (e.g. ""), or just abort() the program. This may be better than throwing an exception, because you can just handle it right there.

[pmattjlab]

Nevermind, this looks fine.

[sdobbs]

I won't post the response I was in the middle of writing then =)

Honestly, more calibration table calls should probably fail harder (and I prefer throwing exceptions in libraries, since they can be handled downstream), but figuring out all of the different use cases is a lot of work - this is just getting some low-hanging fruit.