Move from requests to human_curl.

It turns out that _none_ of the Python stdlib or anything that relies on it
supports CONNECT through a proxy. Beggars belief, but there you go.

1 files changed, 1 insertions, 0 deletions

diff --git a/libmproxy/proxy.py b/libmproxy/proxy.py
index 9249de86..2d4ee988 100644
--- a/libmproxy/proxy.py
+++ b/libmproxy/proxy.py
@@ -123,6 +123,7 @@ def read_http_body(rfile, connection, headers, all, limit):
         content = rfile.read(limit if limit else None)
         connection.close = True
     else:
+        connection.close = True
         content = ""
     return content