diff --git a/doc/cmd.rst b/doc/cmd.rst
index 1a54043d1..2ce7cc13d 100644
--- a/doc/cmd.rst
+++ b/doc/cmd.rst
@@ -103,6 +103,13 @@ but before the program invocation::
     $ coverage run --source=dir1,dir2 my_program.py arg1 arg2
     $ coverage run --source=dir1,dir2 -m packagename.modulename arg1 arg2
 
+.. note::
+   Specifying ``--source`` on the ``coverage run`` command line won't affect
+   subsequent reporting commands like ``coverage xml``.  Use the
+   :ref:`source <config_run_source>` setting in the configuration file to apply
+   the setting uniformly to all commands.
+
+
 Coverage.py can measure multi-threaded programs by default. If you are using
 more exotic concurrency, with the `multiprocessing`_, `greenlet`_, `eventlet`_,
 or `gevent`_ libraries, then coverage.py will get very confused.  Use the
diff --git a/doc/config.rst b/doc/config.rst
index 8d0fc7e2a..e44390e58 100644
--- a/doc/config.rst
+++ b/doc/config.rst
@@ -203,6 +203,10 @@ file paths in the data file.  This makes it easier to measure code in one (or
 multiple) environments, and then report in another. See :ref:`cmd_combine`
 for details.
 
+Note that setting ``source`` has to be done in the configuration file rather
+than the command line for this option to work, since the reporting commands
+need to know the source origin.
+
 .. versionadded:: 5.0
 
 .. _config_run_source: