[stable-2.15] Implement checks, and backwards compat change, to move forward with UTF-8 only (#80370) (#80545)

(cherry picked from commit 0ee7cfb)

changelogs/fragments/80258-defensive-display-non-utf8.yml

@@ -0,0 +1,4 @@
+bugfixes:
+- Display - Defensively configure writing to stdout and stderr with a custom encoding error handler that will replace invalid characters
+  while providing a deprecation warning that non-utf8 text will result in an error in a future version.
+- module responses - Ensure that module responses are utf-8 adhereing to JSON RFC and expectations of the core code.

docs/docsite/rst/dev_guide/developing_modules_best_practices.rst

@@ -124,6 +124,7 @@ Creating correct and informative module output
 
 Modules must output valid JSON only. Follow these guidelines for creating correct, useful module output:
 
+* Module return data must be encoded as strict UTF-8. Modules that cannot return UTF-8 encoded data should return the data encoded by something such as base64. Optionally modules can make the determination if they can encode as UTF-8 and utilize ``errors='replace'`` to replace non UTF-8 characters making the return values lossy.
 * Make your top-level return type a hash (dictionary).
 * Nest complex return values within the top-level hash.
 * Incorporate any lists or simple scalar values within the top-level return hash.

lib/ansible/config/base.yml

@@ -1752,6 +1752,17 @@ MODULE_IGNORE_EXTS:
   ini:
   - {key: module_ignore_exts, section: defaults}
   type: list
+MODULE_STRICT_UTF8_RESPONSE:
+  name: Module strict UTF-8 response
+  description:
+    - Enables whether module responses are evaluated for containing non UTF-8 data
+    - Disabling this may result in unexpected behavior
+    - Only ansible-core should evaluate this configuration
+  env: [{name: ANSIBLE_MODULE_STRICT_UTF8_RESPONSE}]
+  ini:
+    - {key: module_strict_utf8_response, section: defaults}
+  type: bool
+  default: True
 OLD_PLUGIN_CACHE_CLEARING:
   description: Previously Ansible would only clear some of the plugin loading caches when loading new roles, this led to some behaviours in which a plugin loaded in previous plays would be unexpectedly 'sticky'. This setting allows to return to that behaviour.
   env: [{name: ANSIBLE_OLD_PLUGIN_CACHE_CLEAR}]

lib/ansible/module_utils/common/text/converters.py

@@ -268,18 +268,13 @@ def _json_encode_fallback(obj):
 
 
 def jsonify(data, **kwargs):
+    # After 2.18, we should remove this loop, and hardcode to utf-8 in alignment with requiring utf-8 module responses
     for encoding in ("utf-8", "latin-1"):
         try:
-            return json.dumps(data, encoding=encoding, default=_json_encode_fallback, **kwargs)
-        # Old systems using old simplejson module does not support encoding keyword.
-        except TypeError:
-            try:
-                new_data = container_to_text(data, encoding=encoding)
-            except UnicodeDecodeError:
-                continue
-            return json.dumps(new_data, default=_json_encode_fallback, **kwargs)
+            new_data = container_to_text(data, encoding=encoding)
         except UnicodeDecodeError:
             continue
+        return json.dumps(new_data, default=_json_encode_fallback, **kwargs)
     raise UnicodeError('Invalid unicode encoding encountered')
 
 

lib/ansible/modules/command.py

@@ -101,6 +101,8 @@ notes:
     -  The C(executable) parameter is removed since version 2.4. If you have a need for this parameter, use the M(ansible.builtin.shell) module instead.
     -  For Windows targets, use the M(ansible.windows.win_command) module instead.
     -  For rebooting systems, use the M(ansible.builtin.reboot) or M(ansible.windows.win_reboot) module.
+    -  If the command returns non UTF-8 data, it must be encoded to avoid issues. This may necessitate using M(ansible.builtin.shell) so the output
+       can be piped through C(base64).
 seealso:
 - module: ansible.builtin.raw
 - module: ansible.builtin.script

lib/ansible/modules/expect.py

@@ -81,6 +81,8 @@ notes:
   - The M(ansible.builtin.expect) module is designed for simple scenarios.
     For more complex needs, consider the use of expect code with the M(ansible.builtin.shell)
     or M(ansible.builtin.script) modules. (An example is part of the M(ansible.builtin.shell) module documentation).
+  - If the command returns non UTF-8 data, it must be encoded to avoid issues. One option is to pipe
+    the output through C(base64).
 seealso:
 - module: ansible.builtin.script
 - module: ansible.builtin.shell

lib/ansible/modules/raw.py

@@ -39,6 +39,8 @@ description:
      - This module does not require python on the remote system, much like
        the M(ansible.builtin.script) module.
      - This module is also supported for Windows targets.
+     - If the command returns non UTF-8 data, it must be encoded to avoid issues. One option is to pipe
+       the output through C(base64).
 extends_documentation_fragment:
     - action_common_attributes
     - action_common_attributes.raw

lib/ansible/modules/script.py

@@ -47,6 +47,8 @@ notes:
     stderr is sent to stdout. If you depend on separated stdout and stderr result keys, please switch to a copy+command set of tasks instead of using script.
   - If the path to the local script contains spaces, it needs to be quoted.
   - This module is also supported for Windows targets.
+  - If the script returns non UTF-8 data, it must be encoded to avoid issues. One option is to pipe
+    the output through C(base64).
 seealso:
   - module: ansible.builtin.shell
   - module: ansible.windows.win_shell

lib/ansible/modules/shell.py

@@ -90,6 +90,8 @@ notes:
   - An alternative to using inline shell scripts with this module is to use
     the M(ansible.builtin.script) module possibly together with the M(ansible.builtin.template) module.
   - For rebooting systems, use the M(ansible.builtin.reboot) or M(ansible.windows.win_reboot) module.
+  - If the command returns non UTF-8 data, it must be encoded to avoid issues. One option is to pipe
+    the output through C(base64).
 seealso:
 - module: ansible.builtin.command
 - module: ansible.builtin.raw

lib/ansible/plugins/action/__init__.py

@@ -39,6 +39,18 @@ from ansible.utils.plugin_docs import get_versioned_doclink
 display = Display()
 
 
+def _validate_utf8_json(d):
+    if isinstance(d, text_type):
+        # Purposefully not using to_bytes here for performance reasons
+        d.encode(encoding='utf-8', errors='strict')
+    elif isinstance(d, dict):
+        for o in d.items():
+            _validate_utf8_json(o)
+    elif isinstance(d, (list, tuple)):
+        for o in d:
+            _validate_utf8_json(o)
+
+
 class ActionBase(ABC):
 
     '''
@@ -1232,6 +1244,18 @@ class ActionBase(ABC):
                 display.warning(w)
 
             data = json.loads(filtered_output)
+
+            if C.MODULE_STRICT_UTF8_RESPONSE and not data.pop('_ansible_trusted_utf8', None):
+                try:
+                    _validate_utf8_json(data)
+                except UnicodeEncodeError:
+                    # When removing this, also remove the loop and latin-1 from ansible.module_utils.common.text.converters.jsonify
+                    display.deprecated(
+                        f'Module "{self._task.resolved_action or self._task.action}" returned non UTF-8 data in '
+                        'the JSON response. This will become an error in the future',
+                        version='2.18',
+                    )
+
             data['_ansible_parsed'] = True
         except ValueError:
             # not valid json, lets try to capture error

lib/ansible/utils/display.py

@@ -26,6 +26,7 @@ else:
     # this will be set to False if curses.setupterm() fails
     HAS_CURSES = True
 
+import codecs
 import ctypes.util
 import fcntl
 import getpass
@@ -297,8 +298,24 @@ class Display(metaclass=Singleton):
         except Exception as ex:
             self.warning(f"failed to patch stdout/stderr for fork-safety: {ex}")
 
+        codecs.register_error('_replacing_warning_handler', self._replacing_warning_handler)
+        try:
+            sys.stdout.reconfigure(errors='_replacing_warning_handler')
+            sys.stderr.reconfigure(errors='_replacing_warning_handler')
+        except Exception as ex:
+            self.warning(f"failed to reconfigure stdout/stderr with custom encoding error handler: {ex}")
+
         self.setup_curses = False
 
+    def _replacing_warning_handler(self, exception):
+        # TODO: This should probably be deferred until after the current display is completed
+        #       this will require some amount of new functionality
+        self.deprecated(
+            'Non UTF-8 encoded data replaced with "?" while displaying text to stdout/stderr, this is temporary and will become an error',
+            version='2.18',
+        )
+        return '?', exception.end
+
     def set_queue(self, queue):
         """Set the _final_q on Display, so that we know to proxy display over the queue
         instead of directly writing to stdout/stderr from forks