[vbox-dev] deleteConfig in VirtualBox 4.3.x

[Klaus Espenlaub]

On 06.03.2014 14:55, john alexander sanabria ordonez wrote:
> well, what could I say... your answer do not solve the problem.
All I can say is that I've seen working Java code which deletes VMs, 
based on this sample. The language should play no role (unless there is 
a bug), it's all the same API behind it, whether it's used by C++ (e.g. 
VBoxManage) or anything else.

Let's approach it from a different angle - can you provide VBoxSVC.log 
(location is platform dependent, usually in the .VirtualBox subdirectory 
of the home directory) for a case where you just produced the hang? If 
you make sure that no VirtualBox related activities (no GUI, ...) are 
running before your test it'd avoid unrelated log messages.

Klaus
>
>
> On 6 March 2014 05:35, Klaus Espenlaub <klaus.espenlaub at oracle.com 
> <mailto:klaus.espenlaub at oracle.com>> wrote:
>
>     John,
>
>     please stay on the mailing list. This is a public service, if you
>     want personal support I have to ask for your Oracle customer
>     support identifier...
>
>
>     On 05.03.2014 20:16, john alexander sanabria ordonez wrote:
>>     Hi Klaus,
>>
>>     I read the documentation but (from my experience) it does not
>>     behave as I was expecting. For instance, I do understand that the
>>     deleteConfig method does not reach the 100 percent if there is an
>>     error but why the library does not throw an exception informing
>>     about it. In addition, why if the previous method (delete) works
>>     with no errors this time "deleteConfig" will generate any error?
>     The asynchronous part of the deletion might fail (and that's
>     exactly the same as in older versions, just the method name has
>     been changed), and your code will simply hang as it doesn't
>     properly check for completion. Failed async operations usually
>     will never reach 100% - why should they, they failed before.
>
>     These failures will NOT cause exceptions from the progress object
>     method calls, you have to check for them explicitly. It's been
>     always like that, and there's lots of sample code out there which
>     shows how to do this properly.
>
>>     About the wait, this is the actual code
>>
>>                             out << "To delete "
>>                             mediums.each {
>>                                     out << it.name <http://it.name> + " "
>>                             }
>>                             def iprogress = machine.delete(mediums)
>>                             out << "\nDeleting "
>>                             while (iprogress.percent != 100) {
>     Do not check for 100%, check if iprogress.getCompletion returns
>     non-zero (dunno if the true/false stuff makes it all the way to
>     Groovy).
>>                                     out << "."
>>                                     out.flush()
>>                                     Thread.sleep(500)
>     Again, you're ignoring that you should wait using the method from
>     the progress object, iprogress.waitForCompletion(500) as this
>     gives the API middleware the opportunity to run its event queues
>     and so on.
>
>     PLEASE have a look at the Java sample (TestVBox.java), it contains
>     many useful code sequences which are verified to be sensible (not
>     that all of them go into extreme error checking, but at least do
>     the bare minimum necessary to avoid hangs).
>
>     Klaus
>
>>                             }
>>     John,
>>
>>
>>
>>     On 4 March 2014 13:55, Klaus Espenlaub
>>     <klaus.espenlaub at oracle.com <mailto:klaus.espenlaub at oracle.com>>
>>     wrote:
>>
>>         Hi,
>>
>>         On 04.03.2014 14:44, john alexander sanabria ordonez wrote:
>>         > Hi,
>>         >
>>         > I wrote a Groovy script to use the Java binding to manage
>>         virtual
>>         > machines through the VirtualBox web service interface. My
>>         script worked
>>         > well in VirtualBox 4.2.x but when I updated to version 4.3
>>         I noted that
>>         > my unregistervm procedure does not work. First, the
>>         IMachine.delete
>>         > method was not available and it was changed by
>>         IMachine.deleteConfig
>>         > which basically works similarly to the delete method. I
>>         made the
>>         > corresponding modifications however it does not delete the
>>         mediums
>>         > returned by the IMachine.unregister method.
>>
>>         This is documented in the SDK reference, for the API changes
>>         in 4.3.
>>         >
>>         > My unregistervm methods looks similar to this
>>         >
>>         > def mediums =
>>         machine.unregister(CleanupMode.DetachAllReturnHardDisksOnly)
>>         > def iprogress = machine.deleteConfig(mediums)
>>         > while (iprogress.percent < 100) {
>>         >    out << "."
>>         > }
>>         >
>>         > and it never ends because the iprogress.percent never
>>         reaches 100.
>>         >
>>         > What I am doing wrong? Thanks for your help.
>>
>>         The percentage will never reach 100 if there is any error.
>>         You should be
>>         adding more flexible progress checking (there should be
>>         enough working
>>         samples out there, including the Java sample code). What you
>>         have is
>>         asking for hangs. Also, you should at least have some wait for
>>         completion (if you prefer with very low timeout), as
>>         otherwise you're
>>         simply burning CPU cycles.
>>
>>         Klaus
>>
>>         >
>>         > John,
>>         > 
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.virtualbox.org/pipermail/vbox-dev/attachments/20140306/3c453512/attachment.html>