Fix compile error with final field assignment in constructor with prologue in local class in constructor

[coehlrich]

What it does

Fixes #4700

How to test

Attempt to compile

public class Test {
	private final Object a;
	public Test() {
		a = new Object();
		class Test2 {
			private final Object b;
			public Test2() {
				System.out.println();
				super();
				b = new Object();
			}
		}
	}
}

Author checklist

• I have thoroughly tested my changes
• The change is following the coding conventions
• I have signed the Eclipse Contributor Agreement (ECA)

[iloveeclipse]

Two previous Jenkins build silently did nothing...

I've manually triggered Jenkins build https://ci.eclipse.org/jdt/job/eclipse.jdt.core-Github/job/PR-4701/3/console

If the problem persists, please file a ticket at https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/issues/ or ping me here.

[stephan-herrmann]

Cool, this effectively fixes the issue, but something smells strangely here: the flow info that produced the bogus error concerned the field a of the outer class, i.e., we were actually misinterpreting the field index 0.

@coehlrich do you think we can streamline this code section wrt field of the enclosing class? See this comment few lines above:

// discards info about fields of inclosing classes

Given that no analysis within the current type should bother about fields of the enclosing class (?), could we find a single location for calling unconditionalFieldLessCopy() once and for all?

[coehlrich]

Currently internalAnalyseCode uses flowInfo without calling unconditionalFieldLessCopy in 5 locations:

• 2 times as arguments for InitializationFlowContext which only uses it in getInitsForFinalBlankInitializationCheck which also looks for the correct parent before returning the corresponding FlowInfo
• 2 times for calls to reachMode which is copied when calling unconditionalFieldLessCopy
• 1 time as an argument when calling analyseCode on all methods that aren't for initialization

None of the regression tests failed when adding flowInfo = flowInfo.unconditionalFieldLessCopy() to the top of internalAnalyseCode

While looking into this I did also notice that constructors with prologues aren't checked to see if all of the blank final fields have been initialized

[coehlrich]

all uses of the methods relating to fields in FlowInfo in ast either only look at fields in the same class or have a check for if the current class is the same as the one containing the field with some also only calling it on the output of getInitsForFinalBlankInitializationCheck. for subclasses of FlowContext I couldn't get it to recognize final field assignments for the outer class in finally blocks or in the condition of a while loop (which still did give a compile error).

[coehlrich]

Jenkins had another issue with building this PR. For a while after the builds failed the https://ci.eclipse.org/jdt/ was timing out while https://ci.eclipse.org was fine.