Bug 326354

Comment 1 Claudiu Soroiu 2010-09-27 16:58:46 EDT

Created attachment 179689 [details]
test case to reproduce the problem

Comment 2 Claudiu Soroiu 2010-09-27 17:05:51 EDT

i was able to reproduce the bug on windows and linux.

Comment 3 Olivier Thomann 2010-09-27 18:48:58 EDT

I'll investigate.

Comment 4 Olivier Thomann 2010-09-29 16:06:59 EDT

The difference with javac is that the addFirst bridge method in FilterChainImpl is not calling the right method:
  // Method descriptor #96 (Ljava/lang/String;Lp/IMessageFilter;)V
  // Stack: 3, Locals: 3
  public bridge synthetic void addFirst(String arg0, IMessageFilter arg1);
    0  aload_0 [this]
    1  aload_1 [arg0]
    2  aload_2 [arg1]
    3  invokespecial AbstractFilterChain.addFirst(String, IMessageFilter) : void [111]
    6  return
      Line numbers:
        [pc: 0, line: 1]


where javac is generating:
  // Method descriptor #62 (Ljava/lang/String;Lp/IMessageFilter;)V
  // Stack: 3, Locals: 3
  public bridge synthetic void addFirst(String arg0, IMessageFilter arg1);
    0  aload_0 [this]
    1  aload_1 [arg0]
    2  aload_2 [arg1]
    3  invokespecial AbstractFilterChain.addFirst(String, Object) : void [25]
    6  return
      Line numbers:
        [pc: 0, line: 5]

AbstractFilterChain is only defining:
`  // Method descriptor #75 (Ljava/lang/String;Ljava/lang/Object;)V
  // Signature: (Ljava/lang/String;TT;)V
  // Stack: 4, Locals: 3
  public void addFirst(String arg0, Object arg1);
     0  aload_0 [this]
     1  aload_1 [arg0]
     2  invokespecial AbstractFilterChain.checkFilter(String) : void [26]
     5  aload_0 [this]
     6  aload_0 [this]
     7  getfield AbstractFilterChain.head : AbstractFilterChain.Node [5]
    10  aload_1 [arg0]
    11  aload_2 [arg1]
    12  invokespecial AbstractFilterChain.registerFilter(AbstractFilterChain$Node, String, Object) : void [28]
    15  return
      Line numbers:
        [pc: 0, line: 96]
        [pc: 5, line: 97]
        [pc: 15, line: 98]

so we must generate the resolve signature to point to addFirst(Ljava/lang/String;Ljava/lang/Object;)V.

Comment 5 Srikanth Sankaran 2010-10-21 04:05:24 EDT

Reduced single file test case:

public class X extends Y<I>  implements I {
    public static void main(String[] args) {
        ((I) new X()).foo(null);
    }
}

interface I {
    public void foo(I i);
}
 
abstract class Y<T> {
	   public void foo(T t) {}
}

-----------------

Under investigation.

Comment 6 Srikanth Sankaran 2010-10-21 04:23:30 EDT

Interestingly, both the original test case and
the abbreviated test case pass alright on 1.5
compliance and fail only with 1.6 mode.

Comment 7 Srikanth Sankaran 2010-10-21 05:44:48 EDT

Likely a side effect of bug 288658.
See that the problem goes away if
AbstractFilterChain is public instead
of being package private.

Comment 8 Srikanth Sankaran 2010-10-21 06:16:25 EDT

Created attachment 181378 [details]
Patch under test

This patch fixes the problem. Under test.

Basically the target of the bridge method
introduced by the fix for bug 288658 is
incorrect in case of the inherited method
using type parameters and such. We should be
bridging to the "original" method in this
case.

Comment 9 Srikanth Sankaran 2010-10-21 08:13:30 EDT

All tests pass. Olivier, please review. Do you want this for 3.6.2 ?

Comment 10 Srikanth Sankaran 2010-10-22 00:30:03 EDT

Released in HEAD for 3.7 M3

Will decide about 3.6.2 after code review.

Comment 11 Olivier Thomann 2010-10-22 07:45:40 EDT

(In reply to comment #9)
> All tests pass. Olivier, please review. Do you want this for 3.6.2 ?
Patch looks good and since there is no good workaround and the code doesn't run with it, yes I would be tempted to release into 3.6.2.
Daniel ?

Comment 12 Srikanth Sankaran 2010-10-25 01:22:52 EDT

Released in 3.6.x maintenance stream for 3.6.2

Comment 13 Jay Arthanareeswaran 2010-10-26 13:10:59 EDT

Verified for 3.7M3 using build I20101025-0901

Comment 14 Satyam Kandula 2011-01-20 03:51:01 EST

Verified for 3.6.2 RC2 using build M20110119-0834