I'm using BaseHTTPServer's send_response (from within a class that inherits BaseHTTPRequestHandler) with the following:

    self.send_response(response.code, response.headers)
    self.end_headers()
    self.wfile.write(content)
    self.wfile.flush()
When response is a httplib's HTTPResponse object, and its headers inherits from rfc822.Message.

What I see is that message is printed as is, including all the headers trailing "\r\n", after which the send_response method (BaseHTTPServer.py:381) adds another "\r\n".
Then send_response adds the "Server" and "Date" headers.
Since the headers before Server & Date include "\r\n\r\n", Date & server are considered by the browser as the content.

Am I misusing BaseHTTPServer? If not, this is a bug and "\r\n" should be removed from line 381, or added only after a check that shows they are not already there at the headers end, or in case there are no input headers.

It seems to me that you're indeed misusing it.

The correct way would be something like this (assuming response is a HTTPResponse object from httplib):

self.send_response(response.status)
for name, value in response.getheaders():
    self.send_header(name, value)
self.end_headers()

This is because send_response's second argument is the HTTP's "reason" field, i.e. invoking:

    self.send_response(123, 'FOOBAR')

results in 

    HTTP/1.1 123 FOOBAR\r\n

to be sent, followed by "Server" and "Date" headers. The second argument is not meant to be used for sending headers.

(When the second argument is omitted, a standard reason for the given status code is used.)

Thanks for correcting me. I guess I assumed that the "message" variable is
an HTTPMessage.
Is send_response documented somewhere? I failed to find a reference.

On Sun, Jul 3, 2011 at 9:45 PM, Petri Lehtinen <report@bugs.python.org>wrote:

>
> Petri Lehtinen <petri@digip.org> added the comment:
>
> It seems to me that you're indeed misusing it.
>
> The correct way would be something like this (assuming response is a
> HTTPResponse object from httplib):
>
> self.send_response(response.status)
> for name, value in response.getheaders():
>    self.send_header(name, value)
> self.end_headers()
>
> This is because send_response's second argument is the HTTP's "reason"
> field, i.e. invoking:
>
>    self.send_response(123, 'FOOBAR')
>
> results in
>
>    HTTP/1.1 123 FOOBAR\r\n
>
> to be sent, followed by "Server" and "Date" headers. The second argument is
> not meant to be used for sending headers.
>
> (When the second argument is omitted, a standard reason for the given
> status code is used.)
>
> ----------
> nosy: +petri.lehtinen
> resolution:  -> invalid
>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <http://bugs.python.org/issue12439>
> _______________________________________
>

See http://docs.python.org/dev/library/http.server#http.server.BaseHTTPRequestHandler.send_response