Jonathan Fortier created AMQCPP-490:
---------------------------------------
Summary: Exception lifetime confusion can cause the application to
crash
Key: AMQCPP-490
URL: https://issues.apache.org/jira/browse/AMQCPP-490
Project: ActiveMQ C++ Client
Issue Type: Bug
Components: CMS Impl, Decaf
Affects Versions: 3.7.0, 3.6.0, 3.5.0
Reporter: Jonathan Fortier
Assignee: Timothy Bish
In changeset #139775, the "decaf::lang::Exception" class changed its ownership
policy for its std::exception cause. Before the changeset, Exception class
would clone the cause, and after the change, the Exception class took ownership
of the std::exception.
However, many code paths still assume that the Exception class will clone the
std::exception, which makes the application crash when a std::exception is
thrown in their protected region.
Here is a incomplete list of places in the code that makes this bad assumption
(line numbers from version 3.7.0):
-decaf\util\concurrent\FutureTask.h, line 272
-decaf\util\concurrent\FutureTask.h, line 301
-decaf\util\concurrent\ThreadPoolExecutor.cpp, line 742
-activemq\core\ActiveMQTransactionContext.cpp, line 646
-activemq\core\ActiveMQTransactionContext.cpp, line 757
-decaf\internal\net\tcp\TcpSocket.cpp, line 361
Note that subclasses of Exception may also be affected, since they normally
forward the cause to their parent.
The comments on many of those classes still wrongly indicates that the cause
will be cloned. For example, the comments for Exception::Exception(const
std::exception*) is inconsistent with the comments for Exception::initCause()
An easy way to reproduce the bug is to throw a std::exception from an
onException callback, which should make the application crash.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
