[Ctk-developers] Ctk-developers Digest, Vol 26, Issue 6

Sascha Zelzer s.zelzer at dkfz-heidelberg.de
Mon Sep 19 16:35:46 UTC 2011


I can see how it helps to initialize static data for a singleton 
consistently for multiple compilation units within one library.

However, I do not understand how the duplication of the "nifty_counter" 
symbol in Stream.cpp (if compiled into a static lib) could be prevented 
when it is linked from multiple dlls?

- Sascha

On 09/19/2011 06:25 PM, Jean-Christophe Fillion-Robin wrote:
> ctkSingleton or a similar approach relying on a nifty counter ensures 
> that only one instance of the singleton will be available within 
> different shared library. It also ensure that the memory is properly 
> cleaned when the last library that was referencing the singleton is 
> unloaded.
>
> See 
> https://secure.wikimedia.org/wikibooks/en/wiki/More_C%2B%2B_Idioms/Nifty_Counter
>
> Jc
>
> On Mon, Sep 19, 2011 at 12:19 PM, Sascha Zelzer 
> <s.zelzer at dkfz-heidelberg.de <mailto:s.zelzer at dkfz-heidelberg.de>> wrote:
>
>     Hi Jc,
>
>     thanks for the pointer, but I would rather not modify the original
>     QtSOAP code. Besides, ctkSingleton.h wouldn't change the fact that
>     the singleton would be duplicated if it is defined in a static lib
>     and linked from multiple shared libs.
>
>     Thanks,
>     Sascha
>
>
>     On 09/19/2011 05:19 PM, Jean-Christophe Fillion-Robin wrote:
>>     Good catch !
>>
>>     For proper implementation of Singleton, you could refer to
>>     ctkSingleton.h
>>
>>     See
>>     https://github.com/commontk/CTK/blob/master/Libs/Core/ctkSingleton.h
>>
>>     Thanks
>>     Jc
>>
>>     On Mon, Sep 19, 2011 at 10:31 AM, Sascha Zelzer
>>     <s.zelzer at dkfz-heidelberg.de
>>     <mailto:s.zelzer at dkfz-heidelberg.de>> wrote:
>>
>>         You're welcome! Thanks to you for your debugging efforts and
>>         detailed reports here.
>>
>>         -Sascha
>>
>>
>>         On 09/19/2011 04:18 PM, Anthony Dass wrote:
>>>         Hello Sascha,
>>>
>>>         Its really great to know about this fix!. Thank you so much
>>>         for fixing this so quickly. It would really help us, i'll
>>>         try with this version.
>>>
>>>         Once again thank you very much!
>>>
>>>         Thanks,
>>>         ~Anthony
>>>
>>>
>>>
>>>
>>>         On Mon, Sep 19, 2011 at 7:59 AM, Sascha Zelzer
>>>         <s.zelzer at dkfz-heidelberg.de
>>>         <mailto:s.zelzer at dkfz-heidelberg.de>> wrote:
>>>
>>>             Hi Anthony,
>>>
>>>             I finally investigated the issue under Windows (I could
>>>             not reproduce it on Linux) and just pushed a fix for it
>>>             to CTK. Please update your CTK sources and build CTK at
>>>             the superbuild level (this should update the QtSOAP
>>>             library) to fix the namespace issue.
>>>
>>>             Here is the technical stuff why it went wrong ;-)
>>>
>>>             QtSOAP was build as a static library but it used a
>>>             singleton called QtSoapNamespace in its code. So using
>>>             singletons from a static library in multiple DLLs is a
>>>             bad idea and this is why the namespaces where missing
>>>             for some QtSoapMessage instances (they where initialized
>>>             in one DLL, which filled this DLLs QtSOAPNamespace
>>>             singleton with namespace entries, but serialized to XML
>>>             in an other DLL, whose singleton did not know anything
>>>             about the namespaces).
>>>
>>>             Best,
>>>             Sascha
>>>
>>>
>>>
>>>
>>>             On 09/14/2011 03:24 AM, Anthony Dass wrote:
>>>>             Hi Sascha and Ben,
>>>>
>>>>             I noticed that only on submitting the SOAP request the
>>>>             namespaces is getting lost but for the same submission
>>>>             the response shows the valid message with the namespace
>>>>             included.
>>>>
>>>>             So, in the ctkSoapClient class, I just tried creating
>>>>             the QtSoapType object (as below) to set  as a method
>>>>             arguments and this works! - gets the required namespace
>>>>             as expected and the SOAP message is valid one. However
>>>>             if we do the same construction in  ctkDicomHostService
>>>>             and pass to the ctkSimpleSoapClient object via function
>>>>             call, it doesn't seem to work. do you have any clue?
>>>>
>>>>             //created in submitSoapRequest of the
>>>>             ctkSimpleSoapClient  class
>>>>             QtSoapType* soapType = new
>>>>             QtSoapSimpleType(QtSoapQName("state"),
>>>>             ctkDicomSoapState::toStringValue(ctkDicomAppHosting::IDLE));
>>>>             request.addMethodArgument(soapType);
>>>>
>>>>
>>>>             Below is the SOAP message of the request and response.
>>>>
>>>>             submit Request(request, d->Path); - namespace missing
>>>>             in the type attribute
>>>>             =========================
>>>>             <SOAP-ENV:Envelope
>>>>             xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/"
>>>>             SOAP-ENV:encodingStyle="http://schemas.xmlsoap.org/soap/encoding/"
>>>>             xmlns:xsd="htt
>>>>             p://www.w3.org/1999/XMLSchema
>>>>             <http://www.w3.org/1999/XMLSchema>">
>>>>             <SOAP-ENV:Body
>>>>             xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/">
>>>>             <setState xmlns="http://wg23.dicom.nema.org/">
>>>>             <state :type="xsd:string">CANCELED</state>
>>>>             </setState>
>>>>             </SOAP-ENV:Body>
>>>>             </SOAP-ENV:Envelope>
>>>>
>>>>             Got Response. Response:  - name space for type
>>>>             attribute is present.
>>>>             =========================
>>>>             "<SOAP-ENV:Envelope
>>>>             xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/"
>>>>             SOAP-ENV:encodingStyle="http://schemas.xmlsoap.org/soap/encoding/"
>>>>             xml
>>>>             ns:xsd="http://www.w3.org/1999/XMLSchema">
>>>>             <SOAP-ENV:Body
>>>>             xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/">
>>>>             <SOAP-ENV:setState
>>>>             xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/">
>>>>             <SOAP-ENV:setStateResponse
>>>>             xmlns:SOAP-ENV="http://schemas.xmlsoap.org/soap/envelope/"
>>>>             xsi:type="xsd:boolean"
>>>>             xmlns:xsi="http://www.w3.org/1999/XMLSchema-instance">false</SOAP-ENV:setStateResponse>
>>>>             </SOAP-ENV:setState>
>>>>             </SOAP-ENV:Body>
>>>>             </SOAP-ENV:Envelope>
>>>>
>>>>             Thank you!
>>>>             /Anthony
>>>>
>>>>             On Fri, Sep 9, 2011 at 4:31 AM, Benoit Bleuze
>>>>             <benoit.bleuze at inria.fr
>>>>             <mailto:benoit.bleuze at inria.fr>> wrote:
>>>>
>>>>                 We indeed completely depend on QtSoap for creating
>>>>                 the Soap messages. If the soap message looks like
>>>>                 this before any parsong, there is definitely
>>>>                 something wrong in QtSoap.
>>>>                 However it was tha case, I would understand that
>>>>                 our implementation works only on our host, but it
>>>>                 also works with XIP, which is not based on QtSoap.
>>>>                 So either this attribute is simply ignored by the
>>>>                 java implementation, or there is some preprocessing
>>>>                 that strips part of the message?
>>>>                 Definitely worth investigating...
>>>>
>>>>                 ------------------------------------------------------------------------
>>>>
>>>>                     Hello Sascha and Ben,
>>>>
>>>>                     Thank you very much for the quick and detailed
>>>>                     answers.
>>>>
>>>>                     In addition, when we tried to validate the
>>>>                     message using the online validations -
>>>>                     http://www.xmlvalidation.com we also see the
>>>>                     type attribute :type="xsd:string" is not a
>>>>                     valid XML attribute string and also it looks
>>>>                     that the namespace name is missing here. The
>>>>                     string should be either type="xsd:string" if no
>>>>                     namespace is used(without leading colon), or
>>>>                     this xmlns:type="xsd:string" if xmlns is a
>>>>                     namespace name.
>>>>
>>>>                     Since the method addMethodAttribute
>>>>                     of QtSoapmessage is setting this, is that we
>>>>                     need to change in the QtSoap module? or any
>>>>                     other options? please advice.
>>>>
>>>>                     thank you very much!
>>>>                     /Anthony
>>>>
>>>>
>>>>                     On Thu, Sep 8, 2011 at 6:47 AM,
>>>>                     <ctk-developers-request at commontk.org
>>>>                     <mailto:ctk-developers-request at commontk.org>>
>>>>                     wrote:
>>>>
>>>>                         Send Ctk-developers mailing list submissions to
>>>>                         ctk-developers at commontk.org
>>>>                         <mailto:ctk-developers at commontk.org>
>>>>
>>>>                         To subscribe or unsubscribe via the World
>>>>                         Wide Web, visit
>>>>                         http://public.kitware.com/cgi-bin/mailman/listinfo/ctk-developers
>>>>                         or, via email, send a message with subject
>>>>                         or body 'help' to
>>>>                         ctk-developers-request at commontk.org
>>>>                         <mailto:ctk-developers-request at commontk.org>
>>>>
>>>>                         You can reach the person managing the list at
>>>>                         ctk-developers-owner at commontk.org
>>>>                         <mailto:ctk-developers-owner at commontk.org>
>>>>
>>>>                         When replying, please edit your Subject
>>>>                         line so it is more specific
>>>>                         than "Re: Contents of Ctk-developers digest..."
>>>>
>>>>
>>>>                         Today's Topics:
>>>>
>>>>                           1. Re: CTK based Hosted App fails when
>>>>                         launched from Non-CTK
>>>>                              Host (Benoit Bleuze)
>>>>
>>>>
>>>>                         ----------------------------------------------------------------------
>>>>
>>>>                         Message: 1
>>>>                         Date: Thu, 8 Sep 2011 12:46:30 +0200 (CEST)
>>>>                         From: Benoit Bleuze <benoit.bleuze at inria.fr
>>>>                         <mailto:benoit.bleuze at inria.fr>>
>>>>                         Subject: Re: [Ctk-developers] CTK based
>>>>                         Hosted App fails when launched
>>>>                                from Non-CTK Host
>>>>                         To: Sascha Zelzer
>>>>                         <s.zelzer at dkfz-heidelberg.de
>>>>                         <mailto:s.zelzer at dkfz-heidelberg.de>>
>>>>                         Cc: Anthony Dass <antdass at gmail.com
>>>>                         <mailto:antdass at gmail.com>>,
>>>>                         ctk-developers at commontk.org
>>>>                         <mailto:ctk-developers at commontk.org>
>>>>                         Message-ID:
>>>>                         <610925954.61455.1315478790178.JavaMail.root at zmbs4.inria.fr
>>>>                         <mailto:610925954.61455.1315478790178.JavaMail.root at zmbs4.inria.fr>>
>>>>                         Content-Type: text/plain; charset="utf-8"
>>>>
>>>>                         Hi Anthony, Sascha gave you all the
>>>>                         technical answers we had in stock so far,
>>>>                         but I can add a bit of context here, it may
>>>>                         help. As Sascha said, the state/newState
>>>>                         switch is due to XIP not following the
>>>>                         specifications. And your change is valid.
>>>>                         We started with "state", but the only other
>>>>                         host we could use being XIP, we changed it.
>>>>                         You are welcome to revert it. I think we
>>>>                         notified Larry (Lawrence Tarbox) of that
>>>>                         problem during the last hackfest in Chapel
>>>>                         Hill. Eventually XIP will change its API
>>>>                         there (or so I hope). Replacing the WSDL
>>>>                         files doesn't change anything since we
>>>>                         don't use them to generate the interface,
>>>>                         all the messages are hard coded soap
>>>>                         messages constructed from C++. Now I know
>>>>                         nothing of the WCF unfortunately, so I
>>>>                         can't tell you whether the "type:" keyword
>>>>                         is misinterpreted or whatever. Sascha's
>>>>                         suggestion of changing the casing sounds
>>>>                         promising to me. I remember we used
>>>>                         capitals in some actions , following the
>>>>                         WSDL actually, and had to change it
>>>>                         afterwards for
>>>>                          specifications and XIP compatibility. I
>>>>                         remember Larry telling us the WSDL provided
>>>>                         in the specs was not up to date, and thus
>>>>                         we provide an updated (or at least it was
>>>>                         valid last february) set of WSDL (and
>>>>                         siblings) files there:
>>>>                         http://www.commontk.org/index.php/File:Wsdl_ft.zip
>>>>                         Note that they also have actions starting
>>>>                         with capitals. I have a faint memory of
>>>>                         Larry giving us an explanation as to why
>>>>                         they were upper cased, but I can't remember
>>>>                         the reason, my guess is that the .Net
>>>>                         implementation forced them to do so. And it
>>>>                         is very well possible that the XIP java
>>>>                         implementation accepts both casing and the
>>>>                         .NET or WCF are stricter, so I guess you'll
>>>>                         have to choose between the proposed WSDL
>>>>                         and the specifications... Keep us posted of
>>>>                         your progress. Ben ----- Original Message -----
>>>>                         > Hi Anthony,
>>>>                         > Welcome to the ctk-developers list.
>>>>                         > > We've tried adapting one of our DICOM
>>>>                         application into a CTK Based
>>>>                         > > Hosted App. It works well when tried
>>>>                         with the ctkhost - we could
>>>>                         > > launch the app, load the DICOM data and
>>>>                         get through different states
>>>>                         > > as well.
>>>>                         > So far, good news.
>>>>                         > > However when tried launching the same
>>>>                         app with different non-ctk
>>>>                         > > based
>>>>                         > > host, it launches our app but fails to
>>>>                         load any data. This non-ctk
>>>>                         > > based host seems to be written using
>>>>                         the WCF as per specification in
>>>>                         > > DICOM 118.
>>>>                         > Please note that we could only test our
>>>>                         implementation against the XIP
>>>>                         > host (written in Java) so far. So the
>>>>                         problems you are experiencing
>>>>                         > might very well relate to our implementation.
>>>>                         > > On investigating, we noticed that the
>>>>                         initial ?IDLE? message sent by
>>>>                         > > the App upon launch is not getting
>>>>                         processed by the Non-CTK based
>>>>                         > > host.
>>>>                         > > When traced the SOAP communication, we
>>>>                         found out that the messages
>>>>                         > > from the plug-in are getting to the
>>>>                         host, but then they are not
>>>>                         > > processed by WCF because of some WSDL
>>>>                         contacts mismatch. Thus the
>>>>                         > > host
>>>>                         > > does not receive any messages. Here?s
>>>>                         the log message:
>>>>                         > > < Message > The message with Action ''
>>>>                         cannot be processed at the
>>>>                         > > receiver, due to a ContractFilter
>>>>                         mismatch at the
>>>>                         > > EndpointDispatcher.
>>>>                         > > This may be because of either a
>>>>                         contract mismatch (mismatched
>>>>                         > > Actions
>>>>                         > > between sender and receiver) or a
>>>>                         binding/security mismatch between
>>>>                         > > the sender and the receiver. Check that
>>>>                         sender and receiver have the
>>>>                         > > same contract and the same binding
>>>>                         (including security requirements,
>>>>                         > > e.g. Message, Transport, None). </
>>>>                         Message >
>>>>                         > > When debugging the WCF processing of
>>>>                         the message we get the
>>>>                         > > deserialization error:
>>>>                         > > "The ':' character, hexadecimal value
>>>>                         0x3A, cannot be included in a
>>>>                         > > name. Line 4, position 11."
>>>>                         > > Here?s the SOAP message braced between
>>>>                         the CTK-plugin and Non-CTK
>>>>                         > > based Host:
>>>>                         > > < SOAP-ENV:Envelope xmlns:SOAP-ENV = "
>>>>                         > >
>>>>                         http://schemas.xmlsoap.org/soap/envelope/ "
>>>>                         SOAP-ENV:encodingStyle =
>>>>                         > > "
>>>>                         > >
>>>>                         http://schemas.xmlsoap.org/soap/encoding/ "
>>>>                         xmlns:xsd = "
>>>>                         > > http://www.w3.org/1999/XMLSchema " >
>>>>                         > > < SOAP-ENV:Body xmlns:SOAP-ENV = "
>>>>                         > >
>>>>                         http://schemas.xmlsoap.org/soap/envelope/ " >
>>>>                         > > < notifyStateChanged xmlns = "
>>>>                         http://wg23.dicom.nema.org/ " >
>>>>                         > > < newState :type = " xsd:string " >
>>>>                         IDLE </ newState >
>>>>                         > > </ notifyStateChanged >
>>>>                         > > </ SOAP-ENV:Body >
>>>>                         > > </ SOAP-ENV:Envelope >
>>>>                         > > We can see the ?:type? item in the
>>>>                         newState element that?s causing
>>>>                         > > the
>>>>                         > > error in WCF. Also, per DICOM
>>>>                         supplement 118 it should be called
>>>>                         > > ?state?, not ?newState?.
>>>>                         > Looking at the debug output from the CTK
>>>>                         hosted app, the xml namespace
>>>>                         > for the "type" attribute in the
>>>>                         "newState" tag is intact. So I am
>>>>                         > wondering it the output above has already
>>>>                         been modified somehow by the
>>>>                         > WCF host or if it is the original message
>>>>                         from the CTK hosted app.
>>>>                         > > Here?s the example of the SOAP message
>>>>                         generated by Non-CTK based
>>>>                         > > HostedApp: < s:Envelope xmlns:s = "
>>>>                         > >
>>>>                         http://schemas.xmlsoap.org/soap/envelope/ "
>>>>                         > < s:Body > <
>>>>                         > > NotifyStateChanged xmlns = "
>>>>                         > >
>>>>                         http://dicom.nema.org/PS3.19/HostService-20100825
>>>>                         " > < state > IDLE
>>>>                         > > </ state > </ NotifyStateChanged > </
>>>>                         s:Body > </ s:Envelope >
>>>>                         > > Next, we tried changing ?newState? to
>>>>                         ?state? but again it fails
>>>>                         > > when
>>>>                         > > tried with the non-ctk based host and
>>>>                         producing the same error.
>>>>                         > I think it is still either related to the
>>>>                         ":type" attribute or the
>>>>                         > error described below. Concerning
>>>>                         state/newState, please also see my
>>>>                         > comments below.
>>>>                         > > ...
>>>>                         > > 2011-09-06 17:17:59,609 <faultstring
>>>>                         xsi:type="xsd:string"
>>>>                         > > xmlns:xsi="
>>>>                         > >
>>>>                         http://www.w3.org/1999/XMLSchema-instance
>>>>                         ">The message with Action
>>>>                         > > ''
>>>>                         > > cannot be processed at the receiver,
>>>>                         due to a ContractFilter
>>>>                         > > mismatch
>>>>                         > > at the EndpointDispatcher. This may be
>>>>                         because of either a contract
>>>>                         > > mismatch (mismatched Actions between
>>>>                         sender and receiver) or a
>>>>                         > > binding/security mismatch between the
>>>>                         sender and the receiver. Check
>>>>                         > > that sender and receiver have the same
>>>>                         contract and the same binding
>>>>                         > > (including security requirements, e.g.
>>>>                         Message, Transport,
>>>>                         > > None).</faultstring> ...
>>>>                         > This actually hints at a problem
>>>>                         concerning the action names. Looking
>>>>                         > at the WSDL, the operation names start
>>>>                         with upper case latters, where
>>>>                         > we are using lower case (unfortunately,
>>>>                         the text in the DICOM Part 19
>>>>                         > specs refer to the method names with
>>>>                         starting lower case letters).
>>>>                         > > I even tried replacing the WSDL files
>>>>                         with the one provided by
>>>>                         > > Non-CTK
>>>>                         > > based host, but still it doesn?t seem
>>>>                         to work.
>>>>                         > Yes, that does not change the actual
>>>>                         behavior.
>>>>                         > > When looked at the code, I noticed a
>>>>                         comment in the CTK code (below
>>>>                         > > methods) ?spec would be "state", java
>>>>                         has "newState" FIX
>>>>                         > > JAVA/STANDARD?. Could this be something
>>>>                         we need to look for? If so
>>>>                         > > may
>>>>                         > > I know what would be the change in Java
>>>>                         side.
>>>>                         > Unfortunately, again, the text in the
>>>>                         specs refer to "newState" for
>>>>                         > the argument name and this is what the
>>>>                         XIP host is using. So we used
>>>>                         > it too for testing with XIP. But I agree
>>>>                         that the WSDL should be the
>>>>                         > definite guide and we should use "state"
>>>>                         when not testing with XIP.
>>>>                         > My advice would be to keep your
>>>>                         modification ("newState" -> "state")
>>>>                         > and also try to modify the action
>>>>                         (method) names in
>>>>                         > ctkDicomAppService.cpp (lines 40 - 60) to
>>>>                         use upper case for the first
>>>>                         > letter.
>>>>                         > I would be happy if you could keep us
>>>>                         posted about your progress.
>>>>                         > Thanks,
>>>>                         > Sascha
>>>>                         >
>>>>                         _______________________________________________
>>>>                         > Ctk-developers mailing list
>>>>                         > Ctk-developers at commontk.org
>>>>                         <mailto:Ctk-developers at commontk.org>
>>>>                         >
>>>>                         http://public.kitware.com/cgi-bin/mailman/listinfo/ctk-developers
>>>>                         -------------- next part --------------
>>>>                         An HTML attachment was scrubbed...
>>>>                         URL:
>>>>                         <http://public.kitware.com/pipermail/ctk-developers/attachments/20110908/e75bdbbd/attachment.htm>
>>>>
>>>>                         ------------------------------
>>>>
>>>>                         _______________________________________________
>>>>                         Ctk-developers mailing list
>>>>                         Ctk-developers at commontk.org
>>>>                         <mailto:Ctk-developers at commontk.org>
>>>>                         http://public.kitware.com/cgi-bin/mailman/listinfo/ctk-developers
>>>>
>>>>
>>>>                         End of Ctk-developers Digest, Vol 26, Issue 6
>>>>                         *********************************************
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>>         _______________________________________________
>>         Ctk-developers mailing list
>>         Ctk-developers at commontk.org <mailto:Ctk-developers at commontk.org>
>>         http://public.kitware.com/cgi-bin/mailman/listinfo/ctk-developers
>>
>>
>>
>>
>>     -- 
>>     +1 919 869 8849 <tel:%2B1%20919%20869%208849>
>>
>
>
>
>
> -- 
> +1 919 869 8849
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/ctk-developers/attachments/20110919/77e6766a/attachment.htm>


More information about the Ctk-developers mailing list