421 lines
14 KiB
Python
421 lines
14 KiB
Python
"""
|
|
WSGI Protocol Linter
|
|
====================
|
|
|
|
This module provides a middleware that performs sanity checks on the
|
|
behavior of the WSGI server and application. It checks that the
|
|
:pep:`3333` WSGI spec is properly implemented. It also warns on some
|
|
common HTTP errors such as non-empty responses for 304 status codes.
|
|
|
|
.. autoclass:: LintMiddleware
|
|
|
|
:copyright: 2007 Pallets
|
|
:license: BSD-3-Clause
|
|
"""
|
|
import typing as t
|
|
from types import TracebackType
|
|
from urllib.parse import urlparse
|
|
from warnings import warn
|
|
|
|
from ..datastructures import Headers
|
|
from ..http import is_entity_header
|
|
from ..wsgi import FileWrapper
|
|
|
|
if t.TYPE_CHECKING:
|
|
from _typeshed.wsgi import StartResponse
|
|
from _typeshed.wsgi import WSGIApplication
|
|
from _typeshed.wsgi import WSGIEnvironment
|
|
|
|
|
|
class WSGIWarning(Warning):
|
|
"""Warning class for WSGI warnings."""
|
|
|
|
|
|
class HTTPWarning(Warning):
|
|
"""Warning class for HTTP warnings."""
|
|
|
|
|
|
def check_type(context: str, obj: object, need: t.Type = str) -> None:
|
|
if type(obj) is not need:
|
|
warn(
|
|
f"{context!r} requires {need.__name__!r}, got {type(obj).__name__!r}.",
|
|
WSGIWarning,
|
|
stacklevel=3,
|
|
)
|
|
|
|
|
|
class InputStream:
|
|
def __init__(self, stream: t.IO[bytes]) -> None:
|
|
self._stream = stream
|
|
|
|
def read(self, *args: t.Any) -> bytes:
|
|
if len(args) == 0:
|
|
warn(
|
|
"WSGI does not guarantee an EOF marker on the input stream, thus making"
|
|
" calls to 'wsgi.input.read()' unsafe. Conforming servers may never"
|
|
" return from this call.",
|
|
WSGIWarning,
|
|
stacklevel=2,
|
|
)
|
|
elif len(args) != 1:
|
|
warn(
|
|
"Too many parameters passed to 'wsgi.input.read()'.",
|
|
WSGIWarning,
|
|
stacklevel=2,
|
|
)
|
|
return self._stream.read(*args)
|
|
|
|
def readline(self, *args: t.Any) -> bytes:
|
|
if len(args) == 0:
|
|
warn(
|
|
"Calls to 'wsgi.input.readline()' without arguments are unsafe. Use"
|
|
" 'wsgi.input.read()' instead.",
|
|
WSGIWarning,
|
|
stacklevel=2,
|
|
)
|
|
elif len(args) == 1:
|
|
warn(
|
|
"'wsgi.input.readline()' was called with a size hint. WSGI does not"
|
|
" support this, although it's available on all major servers.",
|
|
WSGIWarning,
|
|
stacklevel=2,
|
|
)
|
|
else:
|
|
raise TypeError("Too many arguments passed to 'wsgi.input.readline()'.")
|
|
return self._stream.readline(*args)
|
|
|
|
def __iter__(self) -> t.Iterator[bytes]:
|
|
try:
|
|
return iter(self._stream)
|
|
except TypeError:
|
|
warn("'wsgi.input' is not iterable.", WSGIWarning, stacklevel=2)
|
|
return iter(())
|
|
|
|
def close(self) -> None:
|
|
warn("The application closed the input stream!", WSGIWarning, stacklevel=2)
|
|
self._stream.close()
|
|
|
|
|
|
class ErrorStream:
|
|
def __init__(self, stream: t.IO[str]) -> None:
|
|
self._stream = stream
|
|
|
|
def write(self, s: str) -> None:
|
|
check_type("wsgi.error.write()", s, str)
|
|
self._stream.write(s)
|
|
|
|
def flush(self) -> None:
|
|
self._stream.flush()
|
|
|
|
def writelines(self, seq: t.Iterable[str]) -> None:
|
|
for line in seq:
|
|
self.write(line)
|
|
|
|
def close(self) -> None:
|
|
warn("The application closed the error stream!", WSGIWarning, stacklevel=2)
|
|
self._stream.close()
|
|
|
|
|
|
class GuardedWrite:
|
|
def __init__(self, write: t.Callable[[bytes], object], chunks: t.List[int]) -> None:
|
|
self._write = write
|
|
self._chunks = chunks
|
|
|
|
def __call__(self, s: bytes) -> None:
|
|
check_type("write()", s, bytes)
|
|
self._write(s)
|
|
self._chunks.append(len(s))
|
|
|
|
|
|
class GuardedIterator:
|
|
def __init__(
|
|
self,
|
|
iterator: t.Iterable[bytes],
|
|
headers_set: t.Tuple[int, Headers],
|
|
chunks: t.List[int],
|
|
) -> None:
|
|
self._iterator = iterator
|
|
self._next = iter(iterator).__next__
|
|
self.closed = False
|
|
self.headers_set = headers_set
|
|
self.chunks = chunks
|
|
|
|
def __iter__(self) -> "GuardedIterator":
|
|
return self
|
|
|
|
def __next__(self) -> bytes:
|
|
if self.closed:
|
|
warn("Iterated over closed 'app_iter'.", WSGIWarning, stacklevel=2)
|
|
|
|
rv = self._next()
|
|
|
|
if not self.headers_set:
|
|
warn(
|
|
"The application returned before it started the response.",
|
|
WSGIWarning,
|
|
stacklevel=2,
|
|
)
|
|
|
|
check_type("application iterator items", rv, bytes)
|
|
self.chunks.append(len(rv))
|
|
return rv
|
|
|
|
def close(self) -> None:
|
|
self.closed = True
|
|
|
|
if hasattr(self._iterator, "close"):
|
|
self._iterator.close()
|
|
|
|
if self.headers_set:
|
|
status_code, headers = self.headers_set
|
|
bytes_sent = sum(self.chunks)
|
|
content_length = headers.get("content-length", type=int)
|
|
|
|
if status_code == 304:
|
|
for key, _value in headers:
|
|
key = key.lower()
|
|
if key not in ("expires", "content-location") and is_entity_header(
|
|
key
|
|
):
|
|
warn(
|
|
f"Entity header {key!r} found in 304 response.", HTTPWarning
|
|
)
|
|
if bytes_sent:
|
|
warn("304 responses must not have a body.", HTTPWarning)
|
|
elif 100 <= status_code < 200 or status_code == 204:
|
|
if content_length != 0:
|
|
warn(
|
|
f"{status_code} responses must have an empty content length.",
|
|
HTTPWarning,
|
|
)
|
|
if bytes_sent:
|
|
warn(f"{status_code} responses must not have a body.", HTTPWarning)
|
|
elif content_length is not None and content_length != bytes_sent:
|
|
warn(
|
|
"Content-Length and the number of bytes sent to the"
|
|
" client do not match.",
|
|
WSGIWarning,
|
|
)
|
|
|
|
def __del__(self) -> None:
|
|
if not self.closed:
|
|
try:
|
|
warn(
|
|
"Iterator was garbage collected before it was closed.", WSGIWarning
|
|
)
|
|
except Exception:
|
|
pass
|
|
|
|
|
|
class LintMiddleware:
|
|
"""Warns about common errors in the WSGI and HTTP behavior of the
|
|
server and wrapped application. Some of the issues it checks are:
|
|
|
|
- invalid status codes
|
|
- non-bytes sent to the WSGI server
|
|
- strings returned from the WSGI application
|
|
- non-empty conditional responses
|
|
- unquoted etags
|
|
- relative URLs in the Location header
|
|
- unsafe calls to wsgi.input
|
|
- unclosed iterators
|
|
|
|
Error information is emitted using the :mod:`warnings` module.
|
|
|
|
:param app: The WSGI application to wrap.
|
|
|
|
.. code-block:: python
|
|
|
|
from werkzeug.middleware.lint import LintMiddleware
|
|
app = LintMiddleware(app)
|
|
"""
|
|
|
|
def __init__(self, app: "WSGIApplication") -> None:
|
|
self.app = app
|
|
|
|
def check_environ(self, environ: "WSGIEnvironment") -> None:
|
|
if type(environ) is not dict:
|
|
warn(
|
|
"WSGI environment is not a standard Python dict.",
|
|
WSGIWarning,
|
|
stacklevel=4,
|
|
)
|
|
for key in (
|
|
"REQUEST_METHOD",
|
|
"SERVER_NAME",
|
|
"SERVER_PORT",
|
|
"wsgi.version",
|
|
"wsgi.input",
|
|
"wsgi.errors",
|
|
"wsgi.multithread",
|
|
"wsgi.multiprocess",
|
|
"wsgi.run_once",
|
|
):
|
|
if key not in environ:
|
|
warn(
|
|
f"Required environment key {key!r} not found",
|
|
WSGIWarning,
|
|
stacklevel=3,
|
|
)
|
|
if environ["wsgi.version"] != (1, 0):
|
|
warn("Environ is not a WSGI 1.0 environ.", WSGIWarning, stacklevel=3)
|
|
|
|
script_name = environ.get("SCRIPT_NAME", "")
|
|
path_info = environ.get("PATH_INFO", "")
|
|
|
|
if script_name and script_name[0] != "/":
|
|
warn(
|
|
f"'SCRIPT_NAME' does not start with a slash: {script_name!r}",
|
|
WSGIWarning,
|
|
stacklevel=3,
|
|
)
|
|
|
|
if path_info and path_info[0] != "/":
|
|
warn(
|
|
f"'PATH_INFO' does not start with a slash: {path_info!r}",
|
|
WSGIWarning,
|
|
stacklevel=3,
|
|
)
|
|
|
|
def check_start_response(
|
|
self,
|
|
status: str,
|
|
headers: t.List[t.Tuple[str, str]],
|
|
exc_info: t.Optional[
|
|
t.Tuple[t.Type[BaseException], BaseException, TracebackType]
|
|
],
|
|
) -> t.Tuple[int, Headers]:
|
|
check_type("status", status, str)
|
|
status_code_str = status.split(None, 1)[0]
|
|
|
|
if len(status_code_str) != 3 or not status_code_str.isdecimal():
|
|
warn("Status code must be three digits.", WSGIWarning, stacklevel=3)
|
|
|
|
if len(status) < 4 or status[3] != " ":
|
|
warn(
|
|
f"Invalid value for status {status!r}. Valid status strings are three"
|
|
" digits, a space and a status explanation.",
|
|
WSGIWarning,
|
|
stacklevel=3,
|
|
)
|
|
|
|
status_code = int(status_code_str)
|
|
|
|
if status_code < 100:
|
|
warn("Status code < 100 detected.", WSGIWarning, stacklevel=3)
|
|
|
|
if type(headers) is not list:
|
|
warn("Header list is not a list.", WSGIWarning, stacklevel=3)
|
|
|
|
for item in headers:
|
|
if type(item) is not tuple or len(item) != 2:
|
|
warn("Header items must be 2-item tuples.", WSGIWarning, stacklevel=3)
|
|
name, value = item
|
|
if type(name) is not str or type(value) is not str:
|
|
warn(
|
|
"Header keys and values must be strings.", WSGIWarning, stacklevel=3
|
|
)
|
|
if name.lower() == "status":
|
|
warn(
|
|
"The status header is not supported due to"
|
|
" conflicts with the CGI spec.",
|
|
WSGIWarning,
|
|
stacklevel=3,
|
|
)
|
|
|
|
if exc_info is not None and not isinstance(exc_info, tuple):
|
|
warn("Invalid value for exc_info.", WSGIWarning, stacklevel=3)
|
|
|
|
headers = Headers(headers)
|
|
self.check_headers(headers)
|
|
|
|
return status_code, headers
|
|
|
|
def check_headers(self, headers: Headers) -> None:
|
|
etag = headers.get("etag")
|
|
|
|
if etag is not None:
|
|
if etag.startswith(("W/", "w/")):
|
|
if etag.startswith("w/"):
|
|
warn(
|
|
"Weak etag indicator should be upper case.",
|
|
HTTPWarning,
|
|
stacklevel=4,
|
|
)
|
|
|
|
etag = etag[2:]
|
|
|
|
if not (etag[:1] == etag[-1:] == '"'):
|
|
warn("Unquoted etag emitted.", HTTPWarning, stacklevel=4)
|
|
|
|
location = headers.get("location")
|
|
|
|
if location is not None:
|
|
if not urlparse(location).netloc:
|
|
warn(
|
|
"Absolute URLs required for location header.",
|
|
HTTPWarning,
|
|
stacklevel=4,
|
|
)
|
|
|
|
def check_iterator(self, app_iter: t.Iterable[bytes]) -> None:
|
|
if isinstance(app_iter, bytes):
|
|
warn(
|
|
"The application returned a bytestring. The response will send one"
|
|
" character at a time to the client, which will kill performance."
|
|
" Return a list or iterable instead.",
|
|
WSGIWarning,
|
|
stacklevel=3,
|
|
)
|
|
|
|
def __call__(self, *args: t.Any, **kwargs: t.Any) -> t.Iterable[bytes]:
|
|
if len(args) != 2:
|
|
warn("A WSGI app takes two arguments.", WSGIWarning, stacklevel=2)
|
|
|
|
if kwargs:
|
|
warn(
|
|
"A WSGI app does not take keyword arguments.", WSGIWarning, stacklevel=2
|
|
)
|
|
|
|
environ: "WSGIEnvironment" = args[0]
|
|
start_response: "StartResponse" = args[1]
|
|
|
|
self.check_environ(environ)
|
|
environ["wsgi.input"] = InputStream(environ["wsgi.input"])
|
|
environ["wsgi.errors"] = ErrorStream(environ["wsgi.errors"])
|
|
|
|
# Hook our own file wrapper in so that applications will always
|
|
# iterate to the end and we can check the content length.
|
|
environ["wsgi.file_wrapper"] = FileWrapper
|
|
|
|
headers_set: t.List[t.Any] = []
|
|
chunks: t.List[int] = []
|
|
|
|
def checking_start_response(
|
|
*args: t.Any, **kwargs: t.Any
|
|
) -> t.Callable[[bytes], None]:
|
|
if len(args) not in {2, 3}:
|
|
warn(
|
|
f"Invalid number of arguments: {len(args)}, expected 2 or 3.",
|
|
WSGIWarning,
|
|
stacklevel=2,
|
|
)
|
|
|
|
if kwargs:
|
|
warn("'start_response' does not take keyword arguments.", WSGIWarning)
|
|
|
|
status: str = args[0]
|
|
headers: t.List[t.Tuple[str, str]] = args[1]
|
|
exc_info: t.Optional[
|
|
t.Tuple[t.Type[BaseException], BaseException, TracebackType]
|
|
] = (args[2] if len(args) == 3 else None)
|
|
|
|
headers_set[:] = self.check_start_response(status, headers, exc_info)
|
|
return GuardedWrite(start_response(status, headers, exc_info), chunks)
|
|
|
|
app_iter = self.app(environ, t.cast("StartResponse", checking_start_response))
|
|
self.check_iterator(app_iter)
|
|
return GuardedIterator(
|
|
app_iter, t.cast(t.Tuple[int, Headers], headers_set), chunks
|
|
)
|