don't omit CRLF when those are the only bytes the in the chunk

cachecontrol/adapter.py

@@ -144,6 +144,10 @@ def _update_chunk_length(
                         super_update_chunk_length(self)
                         if self.chunk_left == 0:
                             self._fp._close()  # type: ignore[union-attr]
+                        elif self.chunk_left is not None:
+                            self._fp._set_chunk_bytes_remaining(  # type: ignore[union-attr]
+                                self.chunk_left
+                            )
 
                     response._update_chunk_length = functools.partial(  # type: ignore[method-assign]
                         _update_chunk_length, weakref.ref(response)

cachecontrol/filewrapper.py

@@ -37,6 +37,7 @@ def __init__(
         self.__buf = NamedTemporaryFile("rb+", delete=True)
         self.__fp = fp
         self.__callback = callback
+        self.__chunk_bytes_remaining = 0
 
     def __getattr__(self, name: str) -> Any:
         # The vagaries of garbage collection means that self.__fp is
@@ -107,14 +108,16 @@ def read(self, amt: int | None = None) -> bytes:
 
         return data
 
+    def _set_chunk_bytes_remaining(self, chunk_bytes_remaining: int) -> None:
+        self.__chunk_bytes_remaining = chunk_bytes_remaining
+
     def _safe_read(self, amt: int) -> bytes:
         data: bytes = self.__fp._safe_read(amt)  # type: ignore[attr-defined]
-        if amt == 2 and data == b"\r\n":
-            # urllib executes this read to toss the CRLF at the end
-            # of the chunk.
+        if self.__chunk_bytes_remaining == 0 and amt == 2:
             return data
 
         self.__buf.write(data)
+        self.__chunk_bytes_remaining -= len(data)
         if self.__is_fp_closed():
             self._close()
 

tests/conftest.py

@@ -8,6 +8,7 @@
 
 import cherrypy
 import pytest
+from cheroot.server import HTTPRequest
 
 
 class SimpleApp:
@@ -101,6 +102,14 @@ def stream(self, env, start_response):
         for i in range(10):
             yield pformat(i).encode("utf8")
 
+    def stream_with_crlf(self, env, start_response):
+        headers = [("Content-Type", "text/plain"), ("Cache-Control", "max-age=5000")]
+        start_response("200 OK", headers)
+
+        yield b"AA"
+        yield b"\r\n"
+        yield b"BB"
+
     def fixed_length(self, env, start_response):
         body = b"0123456789"
         headers = [
@@ -147,6 +156,21 @@ def url(server):
     return "http://%s:%s/" % server.bind_addr
 
 
+@pytest.fixture()
+def malformed_chunk_delimiters(monkeypatch):
+    """Make the test server send XX instead of each chunk's trailing CRLF."""
+    write = HTTPRequest.write
+
+    def write_malformed_chunk(request, chunk):
+        if request.chunked_write and chunk:
+            data = f"{len(chunk):x}\r\n".encode() + chunk + b"XX"
+            request.conn.wfile.write(data)
+        else:
+            write(request, chunk)
+
+    monkeypatch.setattr(HTTPRequest, "write", write_malformed_chunk)
+
+
 def get_free_port():
     s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
     s.bind(("", 0))

tests/test_chunked_response.py

@@ -43,14 +43,32 @@ def test_stream_is_cached(self, url, sess):
         content_1 = resp_1.content
 
         resp_2 = sess.get(url + "stream")
-        content_2 = resp_1.content
+        content_2 = resp_2.content
 
         assert not resp_1.from_cache
         assert resp_2.from_cache
         assert content_1 == content_2
 
+    def test_stream_with_crlf_chunk_is_cached_without_corruption(self, url, sess):
+        resp_1 = sess.get(url + "stream_with_crlf")
+        resp_2 = sess.get(url + "stream_with_crlf")
+
+        assert resp_1.content == b"AA\r\nBB"
+        assert resp_2.from_cache
+        assert resp_2.content == resp_1.content
+
     def test_stream_is_not_cached_when_content_is_not_read(self, url, sess):
         sess.get(url + "stream", stream=True)
         resp = sess.get(url + "stream", stream=True)
 
         assert not resp.from_cache
+
+    def test_stream_with_malformed_delimiters_is_cached_without_corruption(
+        self, url, sess, malformed_chunk_delimiters
+    ):
+        resp_1 = sess.get(url + "stream_with_crlf")
+        resp_2 = sess.get(url + "stream_with_crlf")
+
+        assert resp_1.content == b"AA\r\nBB"
+        assert resp_2.from_cache
+        assert resp_2.content == resp_1.content