bpo-33409: Clarify PEP 538/540 relationship (GH-7534)

While locale coercion and UTF-8 mode turned out to
be complementary ideas rather than competing ones,
it isn't immediately obvious why it's useful to
have both, or how they interact at runtime.

This updates both the Python 3.7 What's New doc
and the PYTHONCOERCECLOCALE and PYTHONUTF8
documentation in an attempt to clarify that
relationship:

- in the respective What's New sections, add a closing paragraph
  explaining which problem each one solves, and pointing to the
  other PEP's section for the specific aspects it relies on the other
  PEP to solve
- use "locale-aware mode" as a more descriptive term for the
  default non-UTF-8 mode
- improve wording conistenccy between the PYTHONCOERCECLOCALE
  and PYTHONUTF8 docs when they cover the same thing (mostly
  related to legacy locale detection and setting the standard
  stream error handler)
- improve the description of the locale coercion trigger conditions
  (including pointing out that setting LC_ALL turns off locale coercion)
- port the full description of the UTF-8 mode behaviour changes
  from PEP 540 into the PYTHONUTF8 documentation
- be explicit that PYTHONIOENCODING still overrides the settings
  for the standard streams
- mention concrete examples of things that do and don't get their
  text encoding assumptions adjusted by the two text encoding
  assumption override techniques
(cherry picked from commit 1bcb8a6)

Co-authored-by: Nick Coghlan <[email protected]>

Author: miss-islington

Doc/using/cmdline.rst

@@ -438,8 +438,10 @@ Miscellaneous options
      * Set the :attr:`~sys.flags.dev_mode` attribute of :attr:`sys.flags` to
        ``True``
 
-   * ``-X utf8`` enables the UTF-8 mode, whereas ``-X utf8=0`` disables the
-     UTF-8 mode.
+   * ``-X utf8`` enables UTF-8 mode for operating system interfaces, overriding
+     the default locale-aware mode. ``-X utf8=0`` explicitly disables UTF-8
+     mode (even when it would otherwise activate automatically).
+     See :envvar:`PYTHONUTF8` for more details.
 
    It also allows passing arbitrary values and retrieving them through the
    :data:`sys._xoptions` dictionary.
@@ -789,36 +791,49 @@ conflict.
 .. envvar:: PYTHONCOERCECLOCALE
 
    If set to the value ``0``, causes the main Python command line application
-   to skip coercing the legacy ASCII-based C locale to a more capable UTF-8
-   based alternative.
+   to skip coercing the legacy ASCII-based C and POSIX locales to a more
+   capable UTF-8 based alternative.
 
-   If this variable is *not* set, or is set to a value other than ``0``, and
-   the current locale reported for the ``LC_CTYPE`` category is the default
-   ``C`` locale, then the Python CLI will attempt to configure the following
-   locales for the ``LC_CTYPE`` category in the order listed before loading the
-   interpreter runtime:
+   If this variable is *not* set (or is set to a value other than ``0``), the
+   ``LC_ALL`` locale override environment variable is also not set, and the
+   current locale reported for the ``LC_CTYPE`` category is either the default
+   ``C`` locale, or else the explicitly ASCII-based ``POSIX`` locale, then the
+   Python CLI will attempt to configure the following locales for the
+   ``LC_CTYPE`` category in the order listed before loading the interpreter
+   runtime:
 
    * ``C.UTF-8``
    * ``C.utf8``
    * ``UTF-8``
 
    If setting one of these locale categories succeeds, then the ``LC_CTYPE``
    environment variable will also be set accordingly in the current process
-   environment before the Python runtime is initialized. This ensures the
-   updated setting is seen in subprocesses, as well as in operations that
-   query the environment rather than the current C locale (such as Python's
-   own :func:`locale.getdefaultlocale`).
+   environment before the Python runtime is initialized. This ensures that in
+   addition to being seen by both the interpreter itself and other locale-aware
+   components running in the same process (such as the GNU ``readline``
+   library), the updated setting is also seen in subprocesses (regardless of
+   whether or not those processes are running a Python interpreter), as well as
+   in operations that query the environment rather than the current C locale
+   (such as Python's own :func:`locale.getdefaultlocale`).
 
    Configuring one of these locales (either explicitly or via the above
-   implicit locale coercion) will automatically set the error handler for
-   :data:`sys.stdin` and :data:`sys.stdout` to ``surrogateescape``. This
-   behavior can be overridden using :envvar:`PYTHONIOENCODING` as usual.
+   implicit locale coercion) automatically enables the ``surrogateescape``
+   :ref:`error handler <error-handlers>` for :data:`sys.stdin` and
+   :data:`sys.stdout` (:data:`sys.stderr` continues to use ``backslashreplace``
+   as it does in any other locale). This stream handling behavior can be
+   overridden using :envvar:`PYTHONIOENCODING` as usual.
 
    For debugging purposes, setting ``PYTHONCOERCECLOCALE=warn`` will cause
    Python to emit warning messages on ``stderr`` if either the locale coercion
    activates, or else if a locale that *would* have triggered coercion is
    still active when the Python runtime is initialized.
 
+   Also note that even when locale coercion is disabled, or when it fails to
+   find a suitable target locale, :envvar:`PYTHONUTF8` will still activate by
+   default in legacy ASCII-based locales. Both features must be disabled in
+   order to force the interpreter to use ``ASCII`` instead of ``UTF-8`` for
+   system interfaces.
+
    Availability: \*nix
 
    .. versionadded:: 3.7
@@ -834,10 +849,56 @@ conflict.
 
 .. envvar:: PYTHONUTF8
 
-   If set to ``1``, enable the UTF-8 mode. If set to ``0``, disable the UTF-8
-   mode. Any other non-empty string cause an error.
+   If set to ``1``, enables the interpreter's UTF-8 mode, where ``UTF-8`` is
+   used as the text encoding for system interfaces, regardless of the
+   current locale setting.
+
+   This means that:
+
+    * :func:`sys.getfilesystemencoding()` returns ``'UTF-8'`` (the locale
+      encoding is ignored).
+    * :func:`locale.getpreferredencoding()` returns ``'UTF-8'`` (the locale
+      encoding is ignored, and the function's ``do_setlocale`` parameter has no
+      effect).
+    * :data:`sys.stdin`, :data:`sys.stdout`, and :data:`sys.stderr` all use
+      UTF-8 as their text encoding, with the ``surrogateescape``
+      :ref:`error handler <error-handlers>` being enabled for :data:`sys.stdin`
+      and :data:`sys.stdout` (:data:`sys.stderr` continues to use
+      ``backslashreplace`` as it does in the default locale-aware mode)
+
+   As a consequence of the changes in those lower level APIs, other higher
+   level APIs also exhibit different default behaviours:
+
+    * Command line arguments, environment variables and filenames are decoded
+      to text using the UTF-8 encoding.
+    * :func:`os.fsdecode()` and :func:`os.fsencode()` use the UTF-8 encoding.
+    * :func:`open()`, :func:`io.open()`, and :func:`codecs.open()` use the UTF-8
+      encoding by default. However, they still use the strict error handler by
+      default so that attempting to open a binary file in text mode is likely
+      to raise an exception rather than producing nonsense data.
+
+   Note that the standard stream settings in UTF-8 mode can be overridden by
+   :envvar:`PYTHONIOENCODING` (just as they can be in the default locale-aware
+   mode).
+
+   If set to ``0``, the interpreter runs in its default locale-aware mode.
+
+   Setting any other non-empty string causes an error during interpreter
+   initialisation.
+
+   If this environment variable is not set at all, then the interpreter defaults
+   to using the current locale settings, *unless* the current locale is
+   identified as a legacy ASCII-based locale
+   (as descibed for :envvar:`PYTHONCOERCECLOCALE`), and locale coercion is
+   either disabled or fails. In such legacy locales, the interpreter will
+   default to enabling UTF-8 mode unless explicitly instructed not to do so.
+
+   Also available as the :option:`-X` ``utf8`` option.
+
+   Availability: \*nix
 
    .. versionadded:: 3.7
+      See :pep:`540` for more details.
 
 
 Debug-mode variables

Doc/whatsnew/3.7.rst

@@ -97,9 +97,10 @@ Significant improvements in the standard library:
 
 CPython implementation improvements:
 
+* Avoiding the use of ASCII as a default text encoding:
+  * :ref:`PEP 538 <whatsnew37-pep538>`, legacy C locale coercion
+  * :ref:`PEP 540 <whatsnew37-pep540>`, forced UTF-8 runtime mode
 * :ref:`PEP 552 <whatsnew37-pep552>`, deterministic .pycs
-* :ref:`PEP 538 <whatsnew37-pep538>`, legacy C locale coercion
-* :ref:`PEP 540 <whatsnew37-pep540>`, forced UTF-8 runtime mode
 * :ref:`the new development runtime mode <whatsnew37-devmode>`
 * :ref:`PEP 565 <whatsnew37-pep565>`, improved :exc:`DeprecationWarning`
   handling
@@ -184,7 +185,8 @@ PEP 538: Legacy C Locale Coercion
 
 An ongoing challenge within the Python 3 series has been determining a sensible
 default strategy for handling the "7-bit ASCII" text encoding assumption
-currently implied by the use of the default C locale on non-Windows platforms.
+currently implied by the use of the default C or POSIX locale on non-Windows
+platforms.
 
 :pep:`538` updates the default interpreter command line interface to
 automatically coerce that locale to an available UTF-8 based locale as
@@ -205,10 +207,18 @@ continues to be ``backslashreplace``, regardless of locale.
 
 Locale coercion is silent by default, but to assist in debugging potentially
 locale related integration problems, explicit warnings (emitted directly on
-:data:`~sys.stderr` can be requested by setting ``PYTHONCOERCECLOCALE=warn``.
+:data:`~sys.stderr`) can be requested by setting ``PYTHONCOERCECLOCALE=warn``.
 This setting will also cause the Python runtime to emit a warning if the
 legacy C locale remains active when the core interpreter is initialized.
 
+While :pep:`538`'s locale coercion has the benefit of also affecting extension
+modules (such as GNU ``readline``), as well as child processes (including those
+running non-Python applications and older versions of Python), it has the
+downside of requiring that a suitable target locale be present on the running
+system. To better handle the case where no suitable target locale is available
+(as occurs on RHEL/CentOS 7, for example), Python 3.7 also implements
+:ref:`whatsnew37-pep540`.
+
 .. seealso::
 
     :pep:`538` -- Coercing the legacy C locale to a UTF-8 based locale
@@ -231,8 +241,17 @@ The forced UTF-8 mode can be used to change the text handling behavior in
 an embedded Python interpreter without changing the locale settings of
 an embedding application.
 
-The UTF-8 mode is enabled by default when the locale is "C".  See
-:ref:`whatsnew37-pep538` for details.
+While :pep:`540`'s UTF-8 mode has the benefit of working regardless of which
+locales are available on the running system, it has the downside of having no
+effect on extension modules (such as GNU ``readline``), child processes running
+non-Python applications, and child processes running older versions of Python.
+To reduce the risk of corrupting text data when communicating with such
+components, Python 3.7 also implements :ref:`whatsnew37-pep540`).
+
+The UTF-8 mode is enabled by default when the locale is ``C`` or ``POSIX``, and
+the :pep:`538` locale coercion feature fails to change it to a UTF-8 based
+alternative (whether that failure is due to ``PYTHONCOERCECLOCALE=0`` being set,
+``LC_ALL`` being set, or the lack of a suitable target locale).
 
 .. seealso::
 

Misc/NEWS.d/next/Documentation/2018-06-08-23-46-01.bpo-33409.r4z9MM.rst

@@ -0,0 +1,2 @@
+Clarified the relationship between PEP 538's PYTHONCOERCECLOCALE and PEP
+540's PYTHONUTF8 mode.