6.2.20. Other
6.2.20.1. [Critical] Possible hang in application if something failed while adding a new observable resource *
An application might have hanged if something went wrong while adding a
new observable resource (for example, you created a DataWriter).
Before the hang, you would have seen an exception error in the
RTI_MonitoringResourceRegistry_assertResource
function. However, not
all errors in this function led to the hang. This issue is now fixed.
[RTI Issue ID MONITOR-534]
6.2.20.2. [Critical] Application may have hung when event and event snapshot were published simultaneously for same observable resource *
When RTI Observability Collector Service discovers a Connext application, RTI Monitoring Library 2.0 (previously called Observability Library) automatically sends a special sample named “event snapshot”. This sample contains the current values of event metrics for each observable resource. If an event (for example, liveliness change) was triggered for an observable resource at the same time as an event snapshot was being published for the same resource, the application may have hung. The hang occurred because the thread that published the event and the thread that published the snapshot took the same pair of semaphores in inverted order.
This hang is fixed. Now, both threads take the semaphores in the same order.
[RTI Issue ID MONITOR-584]
6.2.20.3. [Critical] Unable to start Launcher, Admin Console, Code Generator, and Monitor in Windows when the RTI Workspace contained white spaces *
On Windows systems, Launcher, Admin Console, Code Generator, and Monitor failed to start when the RTI Workspace contained white spaces. This issue has been fixed.
[RTI Issue ID TELEMETRY-28]
6.2.20.4. [Critical] Deadlock issue resolved when disabling Monitoring Library 2.0 during command processing *
In the previous release, a deadlock could occur if RTI Monitoring Library 2.0 (previously known as RTI Observability Library) was disabled while a remote administration command was being processed. The hang was caused because the thread that processed the command and the thread that disabled the Library took the same pair of semaphores in inverted order.
This issue has been addressed in this release. Disabling the Library while a remote administration command is being processed is now thread safe.
[RTI Issue ID MONITOR-609]
6.2.20.5. [Major] Native Android applications were not shipped
Native Android applications are now included in Android target bundles, along with the APKs.
[RTI Issue ID INSTALL-789]
6.2.20.6. [Major] References to missing header file in Connext Professional source bundle
The Connext Professional source bundle included references to a header
file in the xmlutils.1.0
module that is not part of the source bundle.
As a result, if you were building Connext from source, you were unable
to complete the build due to the missing header file. RTI has now removed
this dependency from the xmlutils.1.0
module.
[RTI Issue ID CORE-12846]
6.2.20.7. [Major] Access to an internal field of observable resources was not thread safe *
In Monitoring Library 2.0 (previously known as Observability Library), if a remote administration command was issued for an observable resource (such as changing the Forwarding verbosity level of an application) at the same time that periodic metrics were collected for the same resource, an internal field of the resource was accessed by the two threads unsafely. The value of the internal field could remain in an inconsistent state, which, in the worst case, might have led to a deadlock when deleting the resource.
This issue is fixed. Accesses to the internal field are now thread safe.
[RTI Issue ID MONITOR-575]
6.2.20.8. [Minor] Running rtisetenv_<arch>.bat caused issues in PATH environment *
In release 7.1.0, running rtisetenv_<arch>.bat
may have caused
issues in the PATH environment on Windows. This problem has been fixed.
[RTI Issue ID INSTALL-880]
6.2.20.9. [Minor] Error creating a DataWriter using durable writer history if setting property dds.data_writer.history.odbc_plugin.builtin.sample_cache_max_size to -1
Creating a DataWriter using durable writer history and setting the
property
dds.data_writer.history.odbc_plugin.builtin.sample_cache_max_size
to -1
may have failed with the following error:
!allocate sample buffer pool
Even if the DataWriter creation did not fail, the value of
dds.data_writer.history.odbc_plugin.builtin.sample_cache_max_size
would
be incorrectly applied. The value was set to
dds.data_writer.history.odbc_plugin.builtin.instance_cache_max_size
for keyed topics and 1 for unkeyed topics.
This problem has been resolved.
[RTI Issue ID CORE-13732]
6.2.20.10. [Trivial] Connext did not print array dimensions for aliases that were arrays
When printing the type information for a type that is an alias of an array type, the array dimensions are now output for both the IDL and XML representations.
[RTI Issue ID CORE-13651]
* This bug does not affect you if you are upgrading from 6.1.x or earlier.