Script parser: No error reported for unknown labels

[mattulbrich]

If a label is unknown, then the script silently ignores the code and does not do anything.

Load the main example, first PVC and use script

cases {
	case match "this_case_does_not_exist!": {
		thisCommandDoesNotExist ;
	}
}

No error is reported. Also not for the non-existing command.

While this could be named a feature, I'd rather say it is a bug.

Refers to 39a8d27.

[sgrebing]

This is correct behaviour according to our definition of the language. the interpreter interprets the string literal as string expression and tries to find it in the label data of a goal. If no goal with such an expression exists, the case is dismissed.

[mattulbrich]

OK. This definition is theoretically sound and reasonable.

However, from a user experience point of view, this has undesired consequences: A typo in a case condition remains totally undetected.

I would suggest that either the interpreter issues a warning or the code highlighting takes according actions such that such a situation of dead proof code can easily be interpreted.

[sgrebing]

Yeah, I was thinking about building a special type of exception that can be rendered as a warning in the script code. This would however mean a major rebuild of the cases handling in the interpreter to be able to capture and bookkeep which cases have been matched. This issue is then closely related to our enhancement about non-used cases.

[mattulbrich]

While you are at it:

It would be very valuable to continue executing the script on other branches if one branch fails.
Hence having a list of exceptions would be better. That list could contain warnings (probably recognisable by the class of exception) and errors. Alternatively, there could be two lists for warnings and errors.

(After vstte)