Mailing-List: contact dev-help@activemq.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@activemq.apache.org
Received-SPF: pass (athena.apache.org: local policy)
Subject: Re: ActiveMQ-CPP on AIX using IBM compilers dynamic_cast in
	TCPTransport.cpp always returning NULL
From: Timothy Bish <tim.bish@sensis.com>
To: dev@activemq.apache.org
In-Reply-To: <11830194.post@talk.nabble.com>
References: <11820667.post@talk.nabble.com> <46A92BC1.3030702@twcny.rr.com>
	 <11830194.post@talk.nabble.com>
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Date: Fri, 27 Jul 2007 10:19:31 -0400
Message-Id: <1185545971.4555.1.camel@tbish-laptop.ats.sensis.com>
Mime-Version: 1.0

The 2.1 Release candidate is available for download and supports
building as a shared lib, you could try that to see if it makes any
difference.

Get it here:
http://activemq.apache.org/cms/activemq-cpp-21-release.html

Regards
Tim.


On Fri, 2007-07-27 at 07:15 -0700, stevejsaunders wrote:
> 
> Tim Bish wrote:
> > 
> > stevejsaunders wrote:
> >> We have been testing some things with building ActiveMQ-CPP on Windows,
> >> Linux
> >> (RedHat Enterprise Linux v3.1), HP-UX (v11i (B.11.11)), Solaris 9 SPARC
> >> and
> >> AIX v5.2 and v5.3 all using the native commercial compilers (e.g. GNU for
> >> Linux, aCC for HP-UX, Studio 8 on Solaris 9, v6 and v8 Visual C/C++
> >> Enterprise on AIX.   
> >>
> >> With some minor coding changes to get it to build on Solaris and HP-UX
> >> related to STL and getting the build work with our other DSOs and EXEs we
> >> have been successfull so far.  
> >>
> >> One thing we can't get to work is on AIX.   On this OS we are using just
> >> the
> >> src/main/activemq/examples/example program and in
> >> src/main/activemq/transport/TcpTransport.cpp line 61 the dynamic_cast
> >> line
> >> (IOTransport* ioTransport = dynamic_cast<IOTransport*>( next );) always
> >> returns a NULL.   
> >> The next instance when dereferenced in the in the debugger shows a valid
> >> Transport class.   
> >>
> >> We are smart enough to know about having the RTTI turned on via the
> >> compiler
> >> and we have created several stand alone programs to verify that
> >> dynamic_cast
> >> works on these servers using several different compiler options.  We even
> >> created a static library that uses it and an EXE that loads it directly
> >> and
> >> indirectly through a DSO and it works fine.  
> >>
> >> I am at a lost on what to do next.   I have tried to compile it 50
> >> different
> >> ways and such and nothing is working.   
> >>
> >> Attached is the config.log from the configure run.  
> >> Here is what shows up on the command-line from running the example
> >> program:
> >> atl1ax03(rrc):/home/rrc/activemq-cpp-2.0.1-src/src/examples$ ./example
> >> =====================================================
> >> Starting the example:
> >> -----------------------------------------------------
> >> TcpTransport::TcpTransport - transport must be of type IOTransport
> >>         FILE: activemq/transport/TcpTransport.cpp, LINE: 64
> >>         FILE: activemq/transport/TcpTransport.cpp, LINE: 89
> >>         FILE: activemq/transport/TcpTransportFactory.cpp, LINE: 70
> >>         FILE: activemq/core/ActiveMQConnectionFactory.cpp, LINE: 190
> >> TcpTransport::TcpTransport - transport must be of type IOTransport
> >>         FILE: activemq/transport/TcpTransport.cpp, LINE: 64
> >>         FILE: activemq/transport/TcpTransport.cpp, LINE: 89
> >>         FILE: activemq/transport/TcpTransportFactory.cpp, LINE: 70
> >>         FILE: activemq/core/ActiveMQConnectionFactory.cpp, LINE: 190
> >> -----------------------------------------------------
> >> Finished with the example, ignore errors from this
> >> point on as the sockets breaks when we shutdown.
> >> =====================================================
> >>
> >>
> >>
> >>
> >> I even changed the dynamic_cast to a static_cast just to see what whould
> >> happen and here is the dbx output:
> >> atl1ax03(rrc):/home/rrc/activemq-cpp-2.0.1-src/src/examples$ dbx
> >> `pwd`/example core
> >> Type 'help' for help.
> >> [using memory image in core]
> >> reading symbolic information ...
> >>
> >> Illegal instruction (illegal opcode) in . at 0x0 ($t3)
> >> warning: Unable to access address 0x0 from core
> >> (dbx) where
> >> warning: Unable to access address 0xfffffffffffffffc from core
> >> warning: Unable to access address 0xfffffffffffffffc from core
> >> warning: Unable to access address 0xfffffffffffffffc from core
> >> warning: Unable to access address 0xfffffffffffffffc from core
> >> .() at 0x0
> >> unnamed block $b21117, line 120 in "WireFormatInfoMarshaller.cpp"
> >> unnamed block $b21116, line 120 in "WireFormatInfoMarshaller.cpp"
> >> activemq::connector::openwire::marshal::v2::WireFormatInfoMarshaller::looseUnmarshal(activemq::connector::openwire::OpenWireFormat*,activemq::connector::openwire::commands::DataStructure*,activemq::io::DataInputStream*)(this
> >> = 0x301899c8, wireFormat = 0x30189068, dataStructure = 0x301a855c, dataIn
> >> =
> >> 0x3018a008), line 120 in "WireFormatInfoMarshaller.cpp"
> >> unnamed block $b5610, line 246 in "OpenWireFormat.cpp"
> >> unnamed block $b5609, line 246 in "OpenWireFormat.cpp"
> >> unnamed block $b5608, line 246 in "OpenWireFormat.cpp"
> >> doUnmarshal(activemq::io::DataInputStream*)(this = 0x30189068, dis =
> >> 0x3018a008), line 246 in "OpenWireFormat.cpp"
> >> unnamed block $b5600, line 185 in "OpenWireFormat.cpp"
> >> unnamed block $b5599, line 185 in "OpenWireFormat.cpp"
> >> unmarshal(activemq::io::DataInputStream*)(this = 0x30189068, dis =
> >> 0x3018a008), line 185 in "OpenWireFormat.cpp"
> >> unnamed block $b5956, line 71 in "OpenWireCommandReader.cpp"
> >> unnamed block $b5955, line 71 in "OpenWireCommandReader.cpp"
> >> activemq::connector::openwire::OpenWireCommandReader::readCommand()(this
> >> =
> >> 0x30188b37), line 71 in "OpenWireCommandReader.cpp"
> >> unnamed block $b8254, line 166 in "IOTransport.cpp"
> >> activemq::transport::IOTransport::run()(this = 0x3012ee78), line 166 in
> >> "IOTransport.cpp"
> >> runCallback(void*)(param = 0x3018a118), line 152 in "Thread.cpp"
> >>
> >>
> >> Here is the activemq server startup end:
> >> INFO  TransportServerThreadSupport   - Listening for connections at:
> >> tcp://atl1ax03:61616
> >> INFO  TransportConnector             - Connector default Started
> >> INFO  TransportServerThreadSupport   - Listening for connections at:
> >> stomp://atl1ax03:61613
> >> INFO  TransportConnector             - Connector stomp Started
> >> INFO  BrokerService                  - ActiveMQ JMS Message Broker
> >> (localhost, ID:atl1ax03-61252-1185490832763-1:0) started
> >>
> >>
> >> http://www.nabble.com/file/p11820667/config.log config.log 
> >> http://www.nabble.com/file/p11820667/config.status config.status 
> >>
> >>
> >>
> >>   
> > We don't do any testing on AIX so its hard to say, it really sounds like 
> > rtti isn't enabled.  You will definitely need it as ActiveMQ-CPP relies 
> > quite heavily on it, and no static_cast won't work as it doesn't 
> > navigate the inheritance hierarchy, so you won't get a vtable pointer 
> > that works.
> > 
> > Its possible the configure script isn't getting CXXFLAGS setup correctly 
> > for AIX.  You can compare the output of the compile to your own sample 
> > and see what flags are missing.  You can have the script add more to 
> > CXXFLAGS by running ./configure CXXFLAGS=<yout flags here>
> > 
> > regards
> > Tim.
> > 
> > 
> > 
> > 
> 
> I kinda figured that the exposure to other Unix flavors other than Linux was
> limited based on the postings I have seen.   
> 
> Thanks for your suggestions but these two things have already been checked
> and double checked.  The RTTI flag is on, this is what we suspected from the
> beginning based on the behavior.   The testing program and this program have
> been compiled with the same exact arguments, the output from the builds were
> captured and the compile and archive and link of the ActiveMQ-CPP example
> were compared.  
> 
> I am starting to think because the very large size of the static library
> generated for ActiveMQ-CPP that we are having some other sort of problem
> with the linker.  
> 
> Thanks,
> Steve
> 
>