Bug 367203

Summary: [compiler][null] detect assigning null to nonnull argument
Product: [Eclipse Project] JDT Reporter: Stephan Herrmann <stephan.herrmann>
Component: CoreAssignee: Stephan Herrmann <stephan.herrmann>
Status: VERIFIED FIXED QA Contact:
Severity: normal    
Priority: P3 CC: amj87.iitr, srikanth_sankaran
Version: 3.8   
Target Milestone: 3.8 M5   
Hardware: Other   
OS: Linux   
Whiteboard:
Attachments:
Description Flags
tests & fix none

Description Stephan Herrmann CLA 2011-12-20 10:52:54 EST 
From bug 367154 comment 1:
> Is this case already captured in a bug?
> 
> @NonNullByDefault
> public class ShowNPE2 {     
>      public Object foo(Object o1, final boolean b) {
>          o1 = null;   // MISSING NPE warning
>         System.out.println(o1.toString());   
>         return null;  // Correct warning
> 
>     }
> }

This test reveals an obvious omission inside 
AbstractMethodDeclaration.addParameternonNullAnnotation
which can be easily fixed: update the tagBits of argument.binding.

I should also add a test for the inverse:
arg is nonnull from default, passed to a method requiring nonnull.
Expecting this to fail currently, too.
Comment 1 Stephan Herrmann CLA 2011-12-20 13:03:25 EST 
Created attachment 208637 [details]
tests & fix

(In reply to comment #0)
> This test reveals an obvious omission inside 
> AbstractMethodDeclaration.addParameternonNullAnnotation
> which can be easily fixed: update the tagBits of argument.binding.

The patch does this at the call site in MethodBinding.fillInDefaultNonNullness
 
> I should also add a test for the inverse:
> arg is nonnull from default, passed to a method requiring nonnull.
> Expecting this to fail currently, too.

Test has been added, my prediction was wrong, though: this case was
already handled via AbstractMethodDeclaration.analyseArguments()
(which uses MethodBinding.parameterNonNullness instead of tagBits).

Aggressive tests are currently running.
Comment 2 Ayushman Jain CLA 2011-12-20 14:35:48 EST 
(In reply to comment #1)
> > I should also add a test for the inverse:
> > arg is nonnull from default, passed to a method requiring nonnull.
> > Expecting this to fail currently, too.
> 
> Test has been added, my prediction was wrong, though: this case was
> already handled via AbstractMethodDeclaration.analyseArguments()
> (which uses MethodBinding.parameterNonNullness instead of tagBits).

Yes, we'd already tested this particular case.
Comment 3 Stephan Herrmann CLA 2011-12-20 17:03:59 EST 
Released for 3.8 M5 via commit 3a3faf3d0aa82cf2ed2b8375e651ef0b429b1314
Comment 4 Srikanth Sankaran CLA 2012-01-23 03:23:38 EST 
Verified for 3.8 M5 using build id: I20120122-2000