Additions and updates

The TotalView 2026.4 release adds the following new product features and updates:

UI Launch for Single Python Debugging Sessions

The ability to launch a single Python debugging session from the UI is now available by selecting Debug a Python Program from the Start Page or File menu. Through the displayed session dialog choose a Python interpreter, select a script, and enter script arguments. You can now save and reopen single-process native Python sessions alongside other sessions. For more information, see Debug a Python program.

The ability to launch parallel Python debugging sessions from the UI will be available in the next TotalView release. To launch parallel native Python programs, use an MPI launcher from the command line. For more information, see Launch parallel Python programs.

Python exception handling

TotalView 2026.4 now supports Python exception handling. When you debug Python, you can configure TotalView to stop:

  • Whenever an exception is raised.

  • Only when no handler catches an exception.

  • When an exception from your code propagates into library or third-party code.

These stop conditions can be configured independently, so you can choose when to stop on an exception without setting breakpoints by hand. By default, TotalView stops only for uncaught exceptions. For more information, see Configure Python exceptions.

Python multiprocessing package

TotalView 2026.4 now debugs worker processes started by Python's multiprocessing.Process, multiprocessing.Pool, and concurrent.futures.ProcessPoolExecutor on Linux and macOS. Worker debugging is enabled by default. To learn more, see TV::python_multiprocessing.

To see worker processes and their threads grouped by script, process, thread name, and function in the Processes & Threads view, see Group Python processes and threads.

Improved support for Python processes blocked in a library call

TotalView 2026.4 adds improved support for Python processes blocked in a library call, such as time.sleep() or mpi4py.COMM_WORLD.Barrier().

To learn more, see Processes blocked in system calls in Native Python debugging limitations.

Python processes identified by script name in share-group labels

In the Processes & Threads view and the output of dstatus -g share_group, Python processes in a native Python debugging sessions now appear under the name of the script they run instead of the interpreter name. When a share group contains processes running different scripts, each script appears separately, with directory names added when needed to distinguish scripts with the same name. See dstatus.

Call stack filtering for Python debugging

Call stack filtering now works in native Python debugging sessions, where function() and image() rules can hide frames. If filtering hides the frame where execution stopped, the Source view and Local Variables view show the first visible frame. To learn more about stack filtering, see Enabling Stack Filtering.

Bug fixes, security updates, and performance improvements

Numerous bug fixes, security updates, and performance improvements have been addressed, including updates to open-source third-party software packages used by TotalView.