Bug 162991 - [1.5][compiler] Name clash reported for (correct) eclipse-generated code
Summary: [1.5][compiler] Name clash reported for (correct) eclipse-generated code
Status: VERIFIED FIXED
Alias: None
Product: JDT
Classification: Eclipse Project
Component: Core (show other bugs)
Version: 3.2.1   Edit
Hardware: PC Windows XP
: P3 blocker (vote)
Target Milestone: 3.2.2   Edit
Assignee: Philipe Mulet CLA
QA Contact:
URL:
Whiteboard:
Keywords:
: 167268 (view as bug list)
Depends on:
Blocks:
 
Reported: 2006-10-31 16:15 EST by Iain Matcham CLA
Modified: 2007-01-15 12:57 EST (History)
4 users (show)

See Also:

Attachments
Proposed patch for 3.2.2 (954 bytes, patch)
2007-01-08 09:32 EST, Philipe Mulet CLA 		no flags Details | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Iain Matcham CLA 2006-10-31 16:15:50 EST 
Eclipse build M20060921-0945

For the following three sources:
----Something.java----
public interface Something {
	
}
----Doing.java----
public interface Doing {
	public <S extends Something, T extends S> T get(Class<S> clazz);
}
----DoingImpl.java----
public class DoingImpl implements Doing {
	public <S extends Something, T extends S> T get(Class<S> clazz) {
		// TODO Auto-generated method stub
		return null;
	}
}
----end----
a 'Name clash' error is raised by eclipse, (despite this being eclipse generated code).  Eclipse 3.1 & javac 1.5.0_03 both compile this code successfully.  

Incidentally the problem appears to resolve inconsistently if the Doing and DoingImpl are written as inner classes of Something.
Comment 1 Olivier Thomann CLA 2006-11-01 09:46:42 EST 
Reproduced with 3.2 maintenance stream.
This seems to be a incremental build problem.
If a full build is done, no error is reported.
If I make a change in DoingImpl and recompile, then an error (name clash) is reported.
Comment 2 Olivier Thomann CLA 2006-11-01 10:14:50 EST 
Same bug in HEAD.
I'll prepare a regression test in GenericTypeTest.
Comment 3 Olivier Thomann CLA 2006-11-01 10:22:01 EST 
Added disabled regression tests in org.eclipse.jdt.core.tests.compiler.regression.GenericTypeTest. test1067/1068.
1067 checks the full build scenario and 1068 check the incremental build scenario.
Comment 4 Maxime Daniel CLA 2006-11-07 04:28:14 EST 
Enabled test1067, which passes.
test1068 shows that the problem comes from the binary type binding for S (type parameter declared as S extends Something) being marked with AccInterface - which the matching source type binding of test1067 is not. This happens in BinaryTypeBinding#initializeTypeVariable. This results into the method bindings being considered as incompatible since one of the type parameters is an interface (binary) while the other is a class (source).
Comment 5 Kent Johnson CLA 2006-12-11 16:21:10 EST 
*** Bug 167268 has been marked as a duplicate of this bug. ***
Comment 6 Philipe Mulet CLA 2007-01-08 09:31:52 EST 
Setting the AccInterface bit is indeed a mistake (leftover from some early days). 
Removed the offending bit setting, enabled test GenericTypeTest#test1068.
Comment 7 Philipe Mulet CLA 2007-01-08 09:32:19 EST 
Created attachment 56552 [details]
Proposed patch for 3.2.2
Comment 8 Philipe Mulet CLA 2007-01-08 09:33:37 EST 
Maxime - pls add a regression test for the dup bug 167268.
Comment 9 Philipe Mulet CLA 2007-01-08 09:52:34 EST 
Released for 3.2.2.
Comment 10 Philipe Mulet CLA 2007-01-08 09:56:06 EST 
Released for 3.3M5.
Fixed
Comment 11 Eric Jodet CLA 2007-01-15 08:44:42 EST 
verified for 3.2.2 using build M20070112-1200