ref(wsgi): Update _werkzeug vendor to newer version

Fixes GH-3516

Author: mgaligniana

sentry_sdk/_werkzeug.py

@@ -39,60 +39,112 @@
     from typing import Iterator
     from typing import Tuple
 
+text_type = str
+iteritems = lambda d, *args, **kwargs: iter(d.items(*args, **kwargs))
+
 
 #
-# `get_headers` comes from `werkzeug.datastructures.EnvironHeaders`
-# https://github.com/pallets/werkzeug/blob/0.14.1/werkzeug/datastructures.py#L1361
+# `get_headers` comes from `werkzeug.datastructures.EnvironHeaders.__iter__`
+# https://github.com/pallets/werkzeug/blob/1.0.1/src/werkzeug/datastructures.py#L1470
 #
 # We need this function because Django does not give us a "pure" http header
 # dict. So we might as well use it for all WSGI integrations.
 #
 def _get_headers(environ):
     # type: (Dict[str, str]) -> Iterator[Tuple[str, str]]
-    """
-    Returns only proper HTTP headers.
-    """
-    for key, value in environ.items():
-        key = str(key)
+    for key, value in iteritems(environ):
         if key.startswith("HTTP_") and key not in (
             "HTTP_CONTENT_TYPE",
             "HTTP_CONTENT_LENGTH",
         ):
-            yield key[5:].replace("_", "-").title(), value
-        elif key in ("CONTENT_TYPE", "CONTENT_LENGTH"):
-            yield key.replace("_", "-").title(), value
+            yield (
+                key[5:].replace("_", "-").title(),
+                _unicodify_header_value(value),
+            )
+        elif key in ("CONTENT_TYPE", "CONTENT_LENGTH") and value:
+            yield (key.replace("_", "-").title(), _unicodify_header_value(value))
+
+
+def _unicodify_header_value(value):
+    # type: (str | bytes) -> str
+    if isinstance(value, bytes):
+        value = value.decode("latin-1")
+    if not isinstance(value, text_type):
+        value = text_type(value)
+    return value
 
 
 #
 # `get_host` comes from `werkzeug.wsgi.get_host`
-# https://github.com/pallets/werkzeug/blob/1.0.1/src/werkzeug/wsgi.py#L145
+# https://github.com/pallets/werkzeug/blob/3.1.3/src/werkzeug/wsgi.py#L86
 #
 def get_host(environ, use_x_forwarded_for=False):
     # type: (Dict[str, str], bool) -> str
     """
     Return the host for the given WSGI environment.
     """
     if use_x_forwarded_for and "HTTP_X_FORWARDED_HOST" in environ:
-        rv = environ["HTTP_X_FORWARDED_HOST"]
-        if environ["wsgi.url_scheme"] == "http" and rv.endswith(":80"):
-            rv = rv[:-3]
-        elif environ["wsgi.url_scheme"] == "https" and rv.endswith(":443"):
-            rv = rv[:-4]
-    elif environ.get("HTTP_HOST"):
-        rv = environ["HTTP_HOST"]
-        if environ["wsgi.url_scheme"] == "http" and rv.endswith(":80"):
-            rv = rv[:-3]
-        elif environ["wsgi.url_scheme"] == "https" and rv.endswith(":443"):
-            rv = rv[:-4]
-    elif environ.get("SERVER_NAME"):
-        rv = environ["SERVER_NAME"]
-        if (environ["wsgi.url_scheme"], environ["SERVER_PORT"]) not in (
-            ("https", "443"),
-            ("http", "80"),
-        ):
-            rv += ":" + environ["SERVER_PORT"]
+        host = environ["HTTP_X_FORWARDED_HOST"]
+        if environ.get("wsgi.url_scheme") == "http" and host.endswith(":80"):
+            host = host[:-3]
+        elif environ.get("wsgi.url_scheme") == "https" and host.endswith(":443"):
+            host = host[:-4]
     else:
-        # In spite of the WSGI spec, SERVER_NAME might not be present.
-        rv = "unknown"
+        host = _get_host(
+            environ["wsgi.url_scheme"],
+            environ.get("HTTP_HOST"),
+            _get_server(environ),
+        )
+
+    return host
+
+
+# `_get_host` comes from `werkzeug.sansio.utils`
+# https://github.com/pallets/werkzeug/blob/3.1.3/src/werkzeug/sansio/utils.py#L49
+def _get_host(
+    scheme: str,
+    host_header: str | None,
+    server: tuple[str, int | None] | None = None,
+) -> str:
+    """
+    Return the host for the given parameters.
+    """
+
+    host = ""
+
+    if host_header is not None:
+        host = host_header
+    elif server is not None:
+        host = server[0]
+
+        # If SERVER_NAME is IPv6, wrap it in [] to match Host header.
+        # Check for : because domain or IPv4 can't have that.
+        if ":" in host and host[0] != "[":
+            host = f"[{host}]"
+
+        if server[1] is not None:
+            host = f"{host}:{server[1]}"  # noqa: E231
+
+    if scheme in {"http", "ws"} and host.endswith(":80"):
+        host = host[:-3]
+    elif scheme in {"https", "wss"} and host.endswith(":443"):
+        host = host[:-4]
+
+    return host
+
+
+def _get_server(
+    environ,  # type: Dict[str, str]
+) -> tuple[str, int | None] | None:
+    name = environ.get("SERVER_NAME")
+
+    if name is None:
+        return None
+
+    try:
+        port: int | None = int(environ.get("SERVER_PORT", None))
+    except (TypeError, ValueError):
+        # unix socket
+        port = None
 
-    return rv
+    return name, port

tests/integrations/wsgi/test_wsgi.py

@@ -61,6 +61,28 @@ def test_basic(sentry_init, crashing_app, capture_events):
     }
 
 
+def test_basic_django(sentry_init, crashing_app, capture_events):
+    sentry_init(send_default_pii=True)
+    app = SentryWsgiMiddleware(crashing_app, use_x_forwarded_for=True)
+    client = Client(app)
+    events = capture_events()
+
+    with pytest.raises(ZeroDivisionError):
+        client.get("/", environ_overrides={"HTTP_X_FORWARDED_HOST": "localhost:80"})
+
+    (event,) = events
+
+    assert event["transaction"] == "generic WSGI request"
+
+    assert event["request"] == {
+        "env": {"SERVER_NAME": "localhost", "SERVER_PORT": "80"},
+        "headers": {"Host": "localhost", "X-Forwarded-Host": "localhost:80"},
+        "method": "GET",
+        "query_string": "",
+        "url": "http://localhost/",
+    }
+
+
 @pytest.mark.parametrize("path_info", ("bark/", "/bark/"))
 @pytest.mark.parametrize("script_name", ("woof/woof", "woof/woof/"))
 def test_script_name_is_respected(