The configure script could enable C++11 for gcc 4.7 or newer as part of the generated platform_macros.GNU file. This enables the user to use C++11 features

@jwillemsen: There's an error on build dds_doc_master_mosquito_linux_gcc_d0o1i0c1 which disables built-in topics at compile time. The PR added #include "dds/DdsDcpsCoreTypeSupportC.h" in tests/DCPS/StaticDiscovery/DataReaderListenerImpl.cpp. That include can be guarded by an ifdef for DDS_HAS_MINIMUM_BIT.

The OpenDDS configure script invokes MPC but doesn't make use of the parallel MPC workers, could save time with each configure

The configure script lacks support for Visual Studio 2017, this has compiler version 19.10.25017

The QoS XML library uses dds4ccm_opendds as part of its MPC files. This looks an old dependency and should be removed to my idea. At the moment OpenDDS is on a system where there is xerces and ace_xml_utils this library should be compiled and available, independent of dds4ccm_opendds

Are you planning any support fo Matlab/Simulink in future releases?

Maybe a simulink blockset for reading and writing data between models?

It would be an interesting enhancement when OpenDDS would be able to support in addition to IDL to C++ defined types to also support IDL to C++11 defined types. This would at least need some extensions in the opendds idl compiler to support a different generated type system, maybe some more rework in the core.

On some systems I get the message below, it would help new users when this message would point to for example a wiki page describing the background of the warning and how to fix it on various platforms.

(10062|139710143137856) WARNING: Could not find FQDN. Using "127.0.0.1" as fully qualified hostname, please correct system configuration.

There is no unit test that does test read_next_instance_w_condition

The OpenDDS configure script has no option for enabling xerces which controls the compilation of the QoS XML Handler library

Hey guys,

it seems, that your webpage is not online anymore. Is that correct?

Kind Regards,
Max

Hi ,

Because I don't understand how to use your /performance-tests about throughput and latency ,
I have done a program , modified from sample code of other vendor , to test throughput .

But I saw the result was not well in rtps discovery mode .

When I use rtps discovery mode to publish sequence in 8192 payload size , I get 26 Mbps.
When I use DCPSInfoRepo discovery mode to publish sequence in 8192 payload size , I get 500 Mbps .

Would you help me fix rtps.ini to improve throughput ?
I think the code is fine .
my github repo

OS: Ubuntu 16.04 x86_64 VirtualBox

Thanks

After initial installation, configure and setting enviroment variables by following OpenDDS steps, there isn't a DCPSInfoRepo.exe file in bin folder. Why?

OS Windows 10, VS 2015.

With versioned namespaces enabled, the following optional features fail to build:

• java
• wireshark
• qt4

For reference, on Ubuntu 16.04 I configured OpenDDS with these additional options:

• Enable JNI: --java=/usr/lib/jvm/default-java
• Enable wireshark plugin: --wireshark=/my/path/to/wireshark-1.12.13 --glib --rapidjson
• Enable monitor GUI and ishapes: --qt --boost

Pull request #422 is now reverted from master. After the OpenDDS 3.10 release, we can drop support for TAO 2.0a which has no XML_Utils. This issue will be our reminder to effectively re-merge #422 after the release.

Running configure with no arguments warns that $opts{'safety-profile'} is not defined on line 466.

There is no MPC base project that only provides the C++ compiler flags and libraries without pulling in the automatic IDL_Files rules. Inheriting from dcps gives also the IDL compiler flags which requires the user to specify empty IDL_Files sections in his MPC file at the moment he has a project that only compiles/links with OpenDDS C++ code and there are other IDL files in the directory that are not for OpenDDS

As listed on the openDDS website, http://opendds.org/documents/architecture.html, currently openDDS does not fully implement the whole specification.

dds_qos.cpp:5702

::std::basic_string< ACE_TCHAR > n (::XSCRT::XML::uq_name (e.name ()));
if (n == "sec")

@jwillemsen should we edit dds_qos.cpp directly or fix XSC first?

When working on the versioned namespace support (issue #316) I noticed that when using OpenDDS a lot of headers are pulled into user code. A lot of coming through the type specific data reader, data writer, and topic. I think it would help our users when we can reduce the amount of headers they pull in when using OpenDDS. I would expect they only need the IDL generated types and stubs and not a lot of the internal headers.

Currently ZeroCopyDataSeq::~ZeroCopyDataSeq pulls in the full DataReaderImpl.h just to call auto_return_loan, maybe move this to a seperate class so that the zero sequence implementation doesn't pull in the full datareaderimpl with everything it uses

hello,i am newer into dds.
i want to test pub-sub on two indepent windows with rtps.ini like below:

[common]
DCPSGlobalTransportConfig=$file
DCPSDefaultDiscovery=DEFAULT_RTPS

[transport/the_rtps_transport]
transport_type=rtps_udp

but the pub-sub can't find each other.of course,it works(find each other)intra-host,but i want to test inter-host,how can i set the rtps.ini?

thank you very much!

FACE/TS_common.hpp defines NOSIGNAL in enum SEND_FLAG_TYPE, but NOSIGNAL is an already existing define on Solaris 11.3. The enum field should probably be

• either renamed, or
• an ifdef/undef NOSIGNAL should be added to the top of FACE/TS_common.hpp

The spec requires a Time-Based Filter reader to eventually see the latest sample per instance. In cases where that latest sample would normally be suppressed by the TBFilter, the current implementation does not eventually deliver that later sample. This isn't an issue for periodic writers.

In terms of implementation, consider reusing much of the code for reader-side Deadline (timer support, etc.) to make the most recent sample available if the writer stops updating the instance.

This issue was originally on the SourceForge tracker, which we are no longer using.

operator<<(Serializer& serializer, CoherentChangeControl& value) only checks the serializer.good_bit() at the end, wouldn't it be much better to check each insertion independently.

operator>>(Serializer& serializer, CoherentChangeControl& value) could be optimized by just checking the result of each member >>

hey guys,

am i doing something wrong or does the OpenDDS Modeling SDK CodeGen function not generate a .mwc file which is needed for building with mwc?

Thanks.

2015-04-10 (DDS 1.4) 2.2.2.3.6.2

This operation returns the default name for the data-type represented by the TypeSupport.

When working on new windows builds I had a build run where nothing was compiled but several tests seem to run 30 seconds because they don't check if the spawn did success.

tests/DCPS/SharedTransport/run_test.pl udp


X:\BUILD\DOC_ROOT\DDS\bin\.\\DCPSInfoRepo.EXE -o repo.ior
ERROR: Cannot Spawn: <X:\BUILD\DOC_ROOT\DDS\bin\.\\DCPSInfoRepo.EXE> not found
ERROR: waiting for Info Repo IOR file

auto_run_tests_finished: tests/DCPS/SharedTransport/run_test.pl udp Time:31s Result:0

A quick parse gives the following tests that need some work

• tests/DCPS/Prst_delayed_subscriber/run_test.pl
• tests/DCPS/DCPSInfoRepo/run_test.pl
• tests/DCPS/FooTest3_0/run_test.pl
• tests/DCPS/DcpsIntegration/run_test-topic.pl
• tests/DCPS/FooTest4/run_test.pl
• tests/DCPS/MultiRepoTest/run_test.pl
• tests/DCPS/Presentation/run_test.pl
• tests/DCPS/Reconnect/run_test.pl
• tests/DCPS/BuiltInTopicTest/prst_repo_run_test.pl
• tests/transport/simple/run_test.pl bp
• performance-tests/DCPS/InfoRepo_population/run_test.pl
• DevGuideExamples/DCPS/Messenger/run_test.pl
• DevGuideExamples/DCPS/Messenger_ZeroCopy/run_test.pl
• tests/DCPS/Messenger/run_corbaloc_test.pl
• tests/DCPS/Messenger/run_ns_test.pl
• examples/DCPS/Messenger_Imr/run_test.pl
• tests/DCPS/TransientDurability/run_test.pl
• tests/DCPS/PersistentDurability/run_test.pl
• tests/DCPS/StatusCondition/run_test.pl
• tests/DCPS/QueryCondition/run_test.pl
• tests/DCPS/ContentFilteredTopic/run_test.pl
• tests/DCPS/ContentFilteredTopic/run_test.pl nopub
• tests/DCPS/MultiTopic/run_test.pl
• tests/DCPS/Federation/run_test.pl
• tests/DCPS/Priority/run_test.pl
• tests/DCPS/SharedTransport/run_test.pl
• tests/DCPS/Ownership/run_test.pl
• tests/DCPS/GroupPresentation/run_test.pl
• tests/DCPS/ConfigFile/run_test.pl
• tests/DCPS/SubscriberCycle/run_test.pl
• tests/DCPS/PersistentInfoRepo/run_test.pl

As far our automated testing shows OpenDDS tests fail when using MinGW-64. Created this issue to make people aware of this and work on analyzing this further and provide fixes.

Request:
With Presentation QoS access_scope set to group and ordered_access true the DDS middleware should guarantee that samples from multiple writers that are associated with the same publisher are delivered in the same order as written. This is not implemented for RTPS. Part of the problem is that although this feature is described in the DDS specification the current RTPS specification does not include any standard mechanism to implement this on the wire (similar situation as for coherent_access). Apparently, some commercial vendors have implemented this with an RTPS extension, probably by including a publisher-scoped packet count.

Response:
I'm not sure the publisher-scoped counter is enough. Consider an example where data readers A and B (on different topics) are matched with corresponding writers belonging to the same publisher, using the QoS you have listed. The publishing side writes two samples and the overall order is A1, B1. Sample A1 was lost on the network so reader B sees B1 first. B1's "publisher-scoped" counter is 2. Now the subscribing side needs determine whether or not to deliver B1 to the application. It does not have a way to know if sample 1 was lost or if it came from data writer C, for which the subscriber has no corresponding reader (incompatible QoS or a separate Topic).

One solution to this problem would be to include in the headers of each sample a list of (writer GUID, sequence number) pairs indicating the state of each other data writer belonging to that publisher at the time the sample was written.

This issue was originally on the SourceForge tracker, which we are no longer using.

ACE and TAO have support to put everything into a versioned namespace. Would be useful for IDL2C++11 when at least the IDL generated code can support versioned namespaces.

Now that QOS_XML_Handler is enabled, the amount of extra warnings generated from its use of auto_ptr causes Travis CI builds to exceed the available log output space and fail.

@jwillemsen is there anything simple we can do to fix this? Looks like it is XSC-generated code? In the short term we may need to turn off xerces on Travis CI.

Lot of code still uses throw (CORBA::SystemException) in the method declaration as exception specification. This is not part of the IDL to C++ language mapping and doesn't add value, should be removed globally through all code

We normally put all generated files into a subdirectory to keep our user code directory clean. As a test I changed the ReadCondition test to have:

project: dcpsexe, dcps_transports_for_test {
  exename = ReadConditionTest
  idlflags += -SS
  TypeSupport_Files {
    gendir = GeneratedCode
    Messenger.idl
  }
}

This gives the command below, see that -o isn't passed to opendds_idl to put all generated files into a subdirectory

mkdir -p GeneratedCode
OpenDDS/bin/opendds_idl -Sa -St Messenger.idl

I was looking at the generated code and noticed that for example 'BooleanSeqTypeSupportImpl.cpp' the function 'void gen_skip_over(Serializer& ser, CORBA::BooleanSeq*)' is generated as below. The demarshaling of length is not checked, this is done in the 'bool operator>>(Serializer& strm, CORBA::BooleanSeq& seq)' function which is just above. At the moment we get incorrect data just skipping length independent of demarshaling succeeded or not looks a risk

void gen_skip_over(Serializer& ser, CORBA::BooleanSeq*)
{
  ACE_UNUSED_ARG(ser);
  ACE_CDR::ULong length;
  ser >> length;
  ser.skip(length, 1);
}

Looks there is no build using Coverity Scan (see https://scan.coverity.com/) yet for OpenDDS

Building OpenDDS 3.8 (or git HEAD) fails on Solaris 11.3 directly after invoking gmake. Error below.

Configure options:
./configure --java=/usr/jdk/instances/jdk1.8.0/

Error:

user@solaris:/opt/OpenDDS-3.8$ gmake
gmake[1]: Entering directory `/opt/OpenDDS-3.8/ACE_wrappers/ace'

GNUmakefile: /opt/OpenDDS-3.8/ACE_wrappers/ace/GNUmakefile.ACE MAKEFLAGS=w

CC -mt -g -library=Cstd  -DACE_HAS_KSTAT -DACE_HAS_CUSTOM_EXPORT_MACROS=0 -D_POSIX_PTHREAD_SEMANTICS   -I/opt/OpenDDS-3.8/ACE_wrappers -DACE_HAS_SCTP -DACE_HAS_LKSCTP -D__ACE_INLINE__ -I.. -DACE_BUILD_DLL  -c -KPIC -o .shobj/Local_Name_Space.o Local_Name_Space.cpp
"/opt/OpenDDS-3.8/ACE_wrappers/ace/config.h", line 1: Error: Could not open include file "ace/config-solaris.h".
"/opt/OpenDDS-3.8/ACE_wrappers/ace/Assert.h", line 23: Error: ACE_Proper_Export_Flag is not defined.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/Assert.h", line 23: Error: A declaration does not specify a tag or an identifier.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/Assert.h", line 23: Error: Use ";" to terminate declarations.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 48: Error: Expected an expression.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 55: Error: Multiple declaration for ACE_Errno_Guard.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 55: Warning: Implicit int is not supported in C++.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 58: Error: The destructor name must match the class name.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 58: Error: Illegal number of arguments for ~file level().
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 66: Error: operator=(int) must be a member function.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 69: Error: Illegal number of arguments for operator==(int).
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 72: Error: Illegal number of arguments for operator!=(int).
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 74: Error: A declaration was expected instead of "private".
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 76: Error: Type name expected instead of "ACE_Errno_Guard".
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 76: Error: Multiple declaration for ACE_Errno_Guard.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 76: Warning: Implicit int is not supported in C++.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 77: Error: Type name expected instead of "ACE_Errno_Guard".
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 77: Error: Type name expected instead of "ACE_Errno_Guard".
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 77: Error: operator=(?&) must be a member function.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.h", line 83: Error: A declaration was expected instead of "}".
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 8: Error: ACE_Errno_Guard is not defined.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 8: Error: Multiple declaration for ACE_Errno_Guard.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 8: Warning: Implicit int is not supported in C++.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 10: Error: Only a constructor can have constructor initializers.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 19: Warning: "ACE_Errno_Guard(int&, int)" is expected to return a value.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 22: Error: ACE_Errno_Guard is not defined.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 22: Error: Multiple declaration for ACE_Errno_Guard.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 22: Warning: Implicit int is not supported in C++.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 23: Error: Only a constructor can have constructor initializers.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 29: Warning: "ACE_Errno_Guard(int&)" is expected to return a value.
"/opt/OpenDDS-3.8/ACE_wrappers/ace/OS_Errno.inl", line 32: Error: ACE_Errno_Guard is not defined.
Compilation aborted, too many Error messages.
gmake[1]: *** [.shobj/Local_Name_Space.o] Fehler 2
gmake[1]: Leaving directory `/opt/OpenDDS-3.8/ACE_wrappers/ace'
gmake: *** [ACE] Fehler 2

The current iShapes example requires boost to compile and use. Instead of depending on boost it would to my idea be easier to depend on C++11 and remove the boost dependency completely.

We can assist the user by putting documentation about the core DDS API inside the core DDS IDL and also to the generated TypeSupport.idl, that way the user can more easily access that documentation.

Currently OpenDDS doesn't provide the functionality to unregister a type from its type support. Other DDS vendors provide this and as far as I remember there should be an issue on the DDS specification to add this.

It is a documented OpenDDS limitation that RTPS transport instances can not be shared by different Domain Participants. This has a non-obvious consequence for applications that communicate with multiple domains. Such applications necessarily use multiple Domain Participants, and if they rely on a default transport configuration in the OpenDDS configuration file (no domain-specific sections) the application will fail in very obscure, difficult to diagnose ways because it violates this limitation. The limitation should ideally be removed, or if not, then at least detected and an explicit error generated.

This issue was originally on the SourceForge tracker, which we are no longer using.

When using multi dimensional arrays we get non compiling generated code, see PR #459 for a test extension that triggers this issue.

We are using RTPS by default for discovery and we did this through the ace/config.h file as below. We now notice that the OpenDDS unit tests don't run anymore, for example the ReadCondition test fails as below. Would be much easier if a user can set RTPS as default discovery but also that the unit tests work at that moment.

#define DDS_DEFAULT_DISCOVERY_METHOD Discovery::DEFAULT_RTPS

./run_test.pl 
DCPSInfoRepo  -o repo.ior -NOBITS -DCPSPendingTimeout 3 
InfoRepo PID: 6363
(6363|139668423190336) ERROR: DCPSInfoRepo::init: Unable to set the ORB in InfoRepoDiscovery.
Unexpected initialization Error: Unable to set the ORB in InfoRepoDiscovery.                                                                                                                  

Testing OpenDDS 3.8, when trying the Java Messenger I get the attached error. Windows and Java version is in the log.

hs_err_pid8096.txt

Hi

My system is Ubuntu 14.04 x86_64 in a vagrant in Windows7.

I am trying to install OpenDDS 3.9.

And I followed the steps "./configure and make" .

Therefore , I got a lot of error messages like title .
Please check the link of screenshot for detail, http://imgur.com/a/JfKXi

Now I make failed in /DevGuideExamples and /examples .
Always the same messages .

I have tried to ignore them in /OpenDDS-3.9/GUNmakefile.dist , but there were too many things I should ignore .

I also used no_hidden_visibility = 1 in $(ACE_ROOT)/include/makeinclude/platform_macros.GNU.

I think it should be a system problem or environment setting problem ?

And what are the messages meaning ?

Thanks .

Kev

Currently on travis-ci only the default configuration is tested. Using the env matrix support more configurations could be tested which reduces the likelihood of problems on the scoreboard. We do this already for the ACE/TAO travis configuration

OpenDDS doesn't provide (read|take)_instance_w_condition. These are not yet part of the DDS specification but that seems to be an error, see http://issues.omg.org/browse/DDS15-26.

It seems they where intended to be added but never ended up in the specification, not sure why. Other vendors do provide these operations, the lack in OpenDDS would trigger rework in user code when they want to switch.

DDS-XTypes/IDL4 now specify @key for annotating the key fields of an IDL type. Seems OpenDDS doesn't support this yet.

See http://buildlogs.remedy.nl/win_msvc14_opendds_debug/