Re: [Quickfix-developers] PATCH - Java crashes solved

[<OM...@th...>]

Well, I would be careful before making the assumption that the creation of
C++ threads is the cause of the problem.  This is a perfectly valid thing
to do if you make a call to AttachCurrentThread (which is done).  This
article discusses this:
http://www.complex-ite.net/tutorials/JDCBook/jniref.html#thrd

JNI will often return a null pointer on one call when a previous call
caused a serious problem in the JVM.  Could be a synchronization problem (
either in QuickFIX or your callback code ).  But I believe the core of the
problem should be solvable without changing the semantics of the library.
It is also essential for something like the ThreadedSocketAcceptor to exist
at all!

--oren



|---------+----------------------------------------------->
|         |           Alex Hornby <al...@an...>        |
|         |           Sent by:                            |
|         |           qui...@li...ur|
|         |           ceforge.net                         |
|         |                                               |
|         |                                               |
|         |           05/01/2003 10:39 AM                 |
|         |                                               |
|---------+----------------------------------------------->
  >----------------------------------------------------------------------------------------------|
  |                                                                                              |
  |       To:       Oren Miller <OM...@th...>                                       |
  |       cc:       quickfix-developers <qui...@li...>              |
  |       Subject:  [Quickfix-developers] PATCH - Java crashes solved                            |
  >----------------------------------------------------------------------------------------------|




Hi Oren,

Heres a patch that adds a blockingStart() method to the Acceptor class
and makes the java SocketAcceptor wrapper use it.

This is useful because it means there is no long a requirement for C++
thread creation in the call stack from Java.

This is vital for the JVM, as if a thread is created in C++ via
pthread_create() the JVM won't return an environment for it from
ENV::get(). This was the cause of a lot out our java crashes, as when
the environment is unknown a null pointer is returned (and then used).

ChangeLog entry below.

2003-05-01 Alex Hornby <al...@an...>

             * src/java/org_quickfix_SocketAcceptor.cpp: use blockingStart
to
             start acceptor. This is important as it means that the thread
             can be
created in java rather than in C++, which means that the JVM
             knows about it and can give sensible results in ENV::get().

             * src/C++/Application.h: add blockingStart method

             * src/C++/Application.cpp: add blockingStart method

Cheers,
Alex.

Index: src/C++/Acceptor.h
===================================================================
RCS file: /cvsroot/quickfix/quickfix/src/C++/Acceptor.h,v
retrieving revision 1.7
diff -u -r1.7 Acceptor.h
--- src/C++/Acceptor.h         10 Apr 2003 05:13:31 -0000          1.7
+++ src/C++/Acceptor.h         1 May 2003 15:28:29 -0000
@@ -89,6 +89,9 @@

   /// Start acceptor.
   void start() throw ( ConfigError&, RuntimeError& );
+  void blockingStart() throw ( ConfigError&, RuntimeError& );
+
+
   /// Stop acceptor.
   void stop();

Index: src/C++/Acceptor.cpp
===================================================================
RCS file: /cvsroot/quickfix/quickfix/src/C++/Acceptor.cpp,v
retrieving revision 1.9
diff -u -r1.9 Acceptor.cpp
--- src/C++/Acceptor.cpp             10 Apr 2003 05:13:31 -0000
1.9
+++ src/C++/Acceptor.cpp             1 May 2003 15:28:30 -0000
@@ -168,6 +168,17 @@
   QF_STACK_POP
 }

+void Acceptor::blockingStart() throw ( ConfigError&, RuntimeError& )
+{ QF_STACK_PUSH( Acceptor::start )
+
+  onConfigure( m_settings );
+  onInitialize( m_settings );
+
+  startThread(this);
+
+  QF_STACK_POP
+}
+
 void Acceptor::stop()
 { QF_STACK_PUSH( Acceptor::stop )

Index: src/java/org_quickfix_SocketAcceptor.cpp
===================================================================
RCS file:
/cvsroot/quickfix/quickfix/src/java/org_quickfix_SocketAcceptor.cpp,v
retrieving revision 1.7
diff -u -r1.7 org_quickfix_SocketAcceptor.cpp
--- src/java/org_quickfix_SocketAcceptor.cpp           29 Mar 2003 08:26:17
-0000        1.7
+++ src/java/org_quickfix_SocketAcceptor.cpp           1 May 2003 15:28:30
-0000
@@ -139,7 +139,7 @@
   JVM::set( pEnv );
   try
   {
-    getCPPSocketAcceptor( obj ) ->start();
+    getCPPSocketAcceptor( obj ) ->blockingStart();
   }
   catch( FIX::ConfigError &e )
   {



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Quickfix-developers mailing list
Qui...@li...
https://lists.sourceforge.net/lists/listinfo/quickfix-developers