StringSwitchEcjFilter fails to recognize switch where default is first

[Godin]

	private static void example(Object s) {
		switch (String.valueOf(s)) { // assertFullyCovered(0, 2)
			default:
				nop("default");
				break;
			case "a":
				nop("case a");
				break;
		}
	}

in this case ECJ omits last goto expected by filter -

jacoco/org.jacoco.core/src/org/jacoco/core/internal/analysis/filter/StringSwitchEcjFilter.java

Lines 83 to 86 in 68ab19f

	if (cursor.getNext().getOpcode() == Opcodes.GOTO) {
	// end of comparisons for same hashCode
	// jump to default
	nextIs(Opcodes.GOTO);

Looks like that this is due to optimization in algorithm that generates goto instructions - https://github.com/eclipse/eclipse.jdt.core/blob/97b95fa05a64615fc7296744b69bd741cc4af2c7/org.eclipse.jdt.core/compiler/org/eclipse/jdt/internal/compiler/codegen/CodeStream.java#L3544

which is used by algorithm that generates code for switch on String - https://github.com/eclipse/eclipse.jdt.core/blob/97b95fa05a64615fc7296744b69bd741cc4af2c7/org.eclipse.jdt.core/compiler/org/eclipse/jdt/internal/compiler/ast/SwitchStatement.java#L245-L269

[marchof]

Wow, how did you even discover this? I've never seen someone putting default first.

[Godin]

@marchof found this during check that #735 covers all cases from #496 , which mentions this case in comments 😉 don't remember how/why comment was originally added into #496 , however can admit that in C/C++ default as first case with early-return seems quite handy to me in some cases, especially in big switches