Currently, there are quite a few tests where a single test string actually contains:

• One or more commands that generate intermediate data prior to running the test
  (partly to reduce storage space of the test data, partly to make the nature of the test data self-explanatory)
• The command invocation itself
• Evaluation of the outputs from the command, typically by comparing to a pre-generated output from an earlier version

Questions primarily for @daljit46:

1. Can ctest "classify" lines based on which of the three categories each belongs to?
2. Is there an "industry accepted" way of arranging such?