You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@qpid.apache.org by "Andrew Stitcher (JIRA)" <ji...@apache.org> on 2011/04/11 23:00:06 UTC

[jira] [Updated] (QPID-3199) Severe but difficult to diagnose lock error in qpid::sys::StateMonitor

     [ https://issues.apache.org/jira/browse/QPID-3199?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Andrew Stitcher updated QPID-3199:
----------------------------------

    Description: 
The qpid::sys::StateMonitor class has 4 member functions liker this:
  void waitFor(Enum s) { ScopedWait(*this); while (s != state) wait(); }

However the ScopeWait(*this); is in error, because it creates a temporary ScopedWait and then immediately destroys it. Instead of locking this and then unlocking it at the end of the function execution.

What was meant was:
  void waitFor(Enum s) { ScopedWait w(*this); while (s != state) wait(); }

which creates a local variable w which is destroyed at the end of the function execution unlocking this.

It is possible that the compiler/options we use do not actually exhibit this bug. As the correct behaviour may not actually be implemented by all compilers.

If this bug exhibits it might show as StateMonitor::waitFor() never detecting the desired state for instance. It is remarkable that this code has been in the codebase for 3.5 years and we've not noticed any bug caused by it.

This bug was detected by the work on QPID-3004: The clang++ compiler detected that the value being returned by the constructor was not being used at all.

  was:
The qpid::sys::StateMonitor class has 4 member functions liker this:
  void waitFor(Enum s) { ScopedWait(*this); while (s != state) wait(); }

However the ScopeWait(*this); is in error, because it creates a temporary ScopedWait and then immediately destroys it. Instead of locking this and then unlocking it at the end of the function execution.

What was meant was:
  void waitFor(Enum s) { ScopedWait w(*this); while (s != state) wait(); }

which creates a local variable w which is destroyed at the end of the function execution unlocking this.

It is possible that the compiler/options we use do not actually exhibit this bug or at least only with some options as this correct behaviour, may not actually be implemented by all compilers.

If this bug exhibits it might show as StateMonitor::waitFor() never detecting the desired state for instance. It is remarkable that this code has been in the codebase for 3.5 years and we've not noticed any bug caused by it.

This bug was detected by the work on QPID-3004: The clang++ compiler detected that the value being returned by the constructor was not being used at all.


> Severe but difficult to diagnose lock error in qpid::sys::StateMonitor
> ----------------------------------------------------------------------
>
>                 Key: QPID-3199
>                 URL: https://issues.apache.org/jira/browse/QPID-3199
>             Project: Qpid
>          Issue Type: Bug
>          Components: C++ Broker, C++ Client
>    Affects Versions: 0.8, 0.9
>            Reporter: Andrew Stitcher
>            Assignee: Andrew Stitcher
>             Fix For: 0.10
>
>
> The qpid::sys::StateMonitor class has 4 member functions liker this:
>   void waitFor(Enum s) { ScopedWait(*this); while (s != state) wait(); }
> However the ScopeWait(*this); is in error, because it creates a temporary ScopedWait and then immediately destroys it. Instead of locking this and then unlocking it at the end of the function execution.
> What was meant was:
>   void waitFor(Enum s) { ScopedWait w(*this); while (s != state) wait(); }
> which creates a local variable w which is destroyed at the end of the function execution unlocking this.
> It is possible that the compiler/options we use do not actually exhibit this bug. As the correct behaviour may not actually be implemented by all compilers.
> If this bug exhibits it might show as StateMonitor::waitFor() never detecting the desired state for instance. It is remarkable that this code has been in the codebase for 3.5 years and we've not noticed any bug caused by it.
> This bug was detected by the work on QPID-3004: The clang++ compiler detected that the value being returned by the constructor was not being used at all.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org