55384 - Memory leak, when try to invoke method with nsCIDRef parameter

Status: RESOLVED WONTFIX

(Core Graveyard :: Java to XPCOM Bridge, defect, P3)

Description

[Alexei V. Mokeev]

When try to call some method with argument of type "in nsCIDRef",
mozilla crashes.

How to reproduce:
Please apply attached diff to files in mozilla/java/xpcom/java/test directory.
(This diff add one parameter of type nsCIDRef to method test1 of bcIJavaSample 
interface)
then compile test,run mozilla  and note,
that mozilla crashed.


Reproduced with BC sources from 05/10/2000,
Mozilla PR3 build, Windows NT4.0

Comment 1

[Alexei V. Mokeev]

Attachment: please apply this patch to mozilla/java/xpcom/java/test to reproduce bug

Comment 2

[Igor Kushnirskiy]

Resigning to Denis Sharypov 

Comment 3

[Igor Kushnirskiy]

I think it is working. Reporter will you please verify the bug?

Comment 4

[Igor Kushnirskiy]

I am suggesting to close this bug as a dupe for 71951
Reporter what do you say?

Comment 5

[Alexei V. Mokeev]

Will verify this bug with fresh build.
This bug can't be dup of 71951, but 
of course, it may be already fixed by fix for some other bug

Comment 6

[Igor Kushnirskiy]

Reporter, did you verify the bug?

Comment 7

[Igor Kushnirskiy]

Reasingn to myself.
reporter will you please update the bug?

Comment 8

[Igor Kushnirskiy]

Reporter, will you please update the bug?
I am going to close the bug as invalid

Comment 9

[Alexei V. Mokeev]

verified under Linux with latest sources.
Crash isn't more reproduced.
Method with "in nsCIDRef aClass" parameter 
("const nsCID & aClass" in CPP and "CID aClass" in Java).
If this method is implemented in CPP component and invoked 
from Java component, then all works well.
When this method is implemeted in Java and invoked from CPP component,
then exception thrown in console(please see below) and nothing more.

--bcJavaComponentFactory::CreateInstance
endDEbug:avm:bcJavaSample.test3(Object,CID)
java.lang.IllegalArgumentException: argument type mismatch
        at java.lang.reflect.Method.invoke(Native Method)
        at org.mozilla.xpcom.Utilities.callMethodByIndex(Utilities.java:57)
        at org.mozilla.xpcom.Utilities.callMethodByIndex(Native Method)
        at org.mozilla.xpcom.Utilities.callMethod(Utilities.java:75)
        at org.mozilla.xpcom.ProxyHandler.invoke(ProxyHandler.java:53)
        at $Proxy2.test3(Unknown Source)
        at bcJavaSample.test2(bcJavaSample.java:54)
        at java.lang.reflect.Method.invoke(Native Method)
        at org.mozilla.xpcom.Utilities.callMethodByIndex(Utilities.java:57)

Think, that this may be *other* bug. 
Original crash (due to bad memory managment ?) isn't reproduced

Comment 10

[Alexei V. Mokeev]

Change summary. Mozilla isn't crashed on invokation of such method, 
but memory leak occurs.

Comment 11

[Igor Kushnirskiy]

I created bug 81122 for all memory problems
closing this on as dup


*** This bug has been marked as a duplicate of 81122 ***

Comment 12

[Alexei V. Mokeev]

Reopened and removed Geetha from CC

Comment 13

[Marcus Fellinger]

The current version of the xpidl compiler has the problem that it ignores the
[ref] modifier, treating parameters with [ref] as pass-by-value ones.
xpidl_header.c generates an indirect pointer (nsISupports **). We could emulate
this in java by using holder classes. So a parameter of type

[ref] (in|out|inout) nsISupports x

would be mapped to

nsISupportsHolder x

See the Java-IDL-Specification
http://java.sun.com/j2se/1.3/docs/guide/idl/mapping/jidlMapping.html for further
details.

Comment 14

[Worcester12345]

mlk keyword needed?

Comment 15

[jhp (no longer active)]

Blackconnect no longer builds and hasn't been worked on since 2001.  -> WONTFIX