fix: use glob matching instead of fnmatch. #1407

I didn't understand that fnmatch considers the entire string to be a
filename, even if it has slashes in it. This led to incorrect matching.
Now we use our own implementation of glob matching to get the correct
behavior.

3 files changed, 48 insertions, 23 deletions

diff --git a/doc/cmd.rst b/doc/cmd.rst
index cb9a147e..f8de0cb3 100644
--- a/doc/cmd.rst
+++ b/doc/cmd.rst
@@ -342,7 +342,7 @@ single directory, and use the **combine** command to combine them into one
 
     $ coverage combine
 
-You can also name directories or files on the command line::
+You can also name directories or files to be combined on the command line::
 
     $ coverage combine data1.dat windows_data_files/
 
@@ -364,22 +364,6 @@ An existing combined data file is ignored and re-written. If you want to use
 runs, use the ``--append`` switch on the **combine** command.  This behavior
 was the default before version 4.2.
 
-To combine data for a source file, coverage has to find its data in each of the
-data files.  Different test runs may run the same source file from different
-locations. For example, different operating systems will use different paths
-for the same file, or perhaps each Python version is run from a different
-subdirectory.  Coverage needs to know that different file paths are actually
-the same source file for reporting purposes.
-
-You can tell coverage.py how different source locations relate with a
-``[paths]`` section in your configuration file (see :ref:`config_paths`).
-It might be more convenient to use the ``[run] relative_files``
-setting to store relative file paths (see :ref:`relative_files
-<config_run_relative_files>`).
-
-If data isn't combining properly, you can see details about the inner workings
-with ``--debug=pathmap``.
-
 If any of the data files can't be read, coverage.py will print a warning
 indicating the file and the problem.
 
@@ -414,6 +398,28 @@ want to keep those files, use the ``--keep`` command-line option.
 .. [[[end]]] (checksum: 0bdd83f647ee76363c955bedd9ddf749)
 
 
+.. _cmd_combine_remapping:
+
+Re-mapping paths
+................
+
+To combine data for a source file, coverage has to find its data in each of the
+data files.  Different test runs may run the same source file from different
+locations. For example, different operating systems will use different paths
+for the same file, or perhaps each Python version is run from a different
+subdirectory.  Coverage needs to know that different file paths are actually
+the same source file for reporting purposes.
+
+You can tell coverage.py how different source locations relate with a
+``[paths]`` section in your configuration file (see :ref:`config_paths`).
+It might be more convenient to use the ``[run] relative_files``
+setting to store relative file paths (see :ref:`relative_files
+<config_run_relative_files>`).
+
+If data isn't combining properly, you can see details about the inner workings
+with ``--debug=pathmap``.
+
+
 .. _cmd_erase:
 
 Erase data: ``coverage erase``
diff --git a/doc/config.rst b/doc/config.rst
index 6b753579..c6f6442a 100644
--- a/doc/config.rst
+++ b/doc/config.rst
@@ -357,7 +357,7 @@ The first list that has a match will be used.
 The ``--debug=pathmap`` option can be used to log details of the re-mapping of
 paths.  See :ref:`the --debug option <cmd_run_debug>`.
 
-See :ref:`cmd_combine` for more information.
+See :ref:`cmd_combine_remapping` and :ref:`source_glob` for more information.
 
 
 .. _config_report:
diff --git a/doc/source.rst b/doc/source.rst
index cfd0e6fc..64ebd132 100644
--- a/doc/source.rst
+++ b/doc/source.rst
@@ -59,10 +59,10 @@ removed from the set.
 
 .. highlight:: ini
 
-The ``include`` and ``omit`` file name patterns follow typical shell syntax:
-``*`` matches any number of characters and ``?`` matches a single character.
-Patterns that start with a wildcard character are used as-is, other patterns
-are interpreted relative to the current directory::
+The ``include`` and ``omit`` file name patterns follow common shell syntax,
+described below in :ref:`source_glob`.  Patterns that start with a wildcard
+character are used as-is, other patterns are interpreted relative to the
+current directory::
 
     [run]
     omit =
@@ -77,7 +77,7 @@ The ``source``, ``include``, and ``omit`` values all work together to determine
 the source that will be measured.
 
 If both ``source`` and ``include`` are set, the ``include`` value is ignored
-and a warning is printed on the standard output.
+and a warning is issued.
 
 
 .. _source_reporting:
@@ -103,3 +103,22 @@ reporting.
 
 Note that these are ways of specifying files to measure.  You can also exclude
 individual source lines.  See :ref:`excluding` for details.
+
+
+.. _source_glob:
+
+File patterns
+-------------
+
+File path patterns are used for include and omit, and for combining path
+remapping.  They follow common shell syntax:
+
+- ``*`` matches any number of file name characters, not including the directory
+  separator.
+
+- ``?`` matches a single file name character.
+
+- ``**`` matches any number of nested directory names, including none.
+
+- Both ``/`` and ``\`` will match either a slash or a backslash, to make
+  cross-platform matching easier.