Bug 235781

Summary:

[compiler] difference to javac in definite unassignment analysis involving an exception within a constructor

Product:

[Eclipse Project] JDT

Reporter:

Maxime Daniel <maxime_daniel>

Component:

Core

Assignee:

Srikanth Sankaran <srikanth_sankaran>

Status:

VERIFIED FIXED

QA Contact:

Severity:

normal

Priority:

CC:

Olivier_Thomann

Version:

3.4

Target Milestone:

3.6 M5

Hardware:

OS:

All

Whiteboard:

Attachments:

Description	Flags
Proposed patch	none

Description Maxime Daniel

2008-06-05 05:31:41 EDT

I20080604-2000

On the following test case, Eclipse complains that i may remain uninitialized, while javac is silent:
public class X {
  final int i;
  X() {
	if (true) {
	  throw new NullPointerException();
	}
  }
}

Note that both keep quiet on the following:

public class X {
  X() {
	final int i;
	if (true) {
	  throw new NullPointerException();
	}
	System.out.println(i);
  }
}

Added FlowAnalysisTest#50 & 51.

Comment 1 Srikanth Sankaran

2009-12-15 07:49:23 EST

class Y {
	Y(Object o) {
	}
}
public class X extends Y {
	final int blank;
 {
    if (true)
        throw new NullPointerException();
 }
 X() {
    super(new Object() {});     
 }
}

elicits an error from us, but not from javac (5/6/7)

This behavior is codified by us in org.eclipse.jdt.core.tests.compiler.regression.FlowAnalysisTest.test031()

Comment 2 Srikanth Sankaran

2009-12-15 23:16:46 EST

Created attachment 154536 [details]
Proposed patch

Proposed fix and tests

Comment 3 Srikanth Sankaran

2009-12-16 02:15:54 EST

Released in HEAD for 3.6M5

Comment 4 Olivier Thomann

2010-01-25 14:14:33 EST

Verified for 3.6M5 using I20100125-0800