Two WSGI environ bug fixes.

[dowski]

The Content-Type header was not being correctly handled. If it wasn't sent exactly as "CONTENT-TYPE" it was getting recorded in the environ as "HTTP_CONTENT_TYPE" which is against PEP-333.

In the pure-Python parser, the SCRIPT_NAME field was being mishandled due to mutation of the original environ. The code was looking for HTTP_SCRIPT_NAME which was already popped from the environ dictionary if it existed.

[dowski]

Here's an interpreter session that shows the problem:

>>> req = """GET /index HTTP/1.1\r
... Host: foo\r
... Content-Type: bar\r
... \r
... """
>>> parser = HttpParser()
>>> parser.execute(req, len(req))
53
>>> parser.get_wsgi_environ()
{'RAW_URI': '/index', 'SCRIPT_NAME': '', 'REQUEST_METHOD': 'GET', 'HTTP_HOST': 'foo', 'PATH_INFO': '/index', 'SERVER_PROTOCOL': 'HTTP/1.1', 'QUERY_STRING': '', 'HTTP_CONTENT_TYPE': 'bar'}

Notice how you get HTTP_CONTENT_TYPE in the returned environ dictionary, not CONTENT_TYPE.

[dowski]

Any thoughts on this pull request? It seems like it addresses a real and fundamental bug to me, but maybe I'm misusing the API or something?

For reference, we're using http-parser in diesel (http://diesel.io/) and we'd like to be able to count on the official release version and not have to bundle a fixed version to work around this (apparent) issue.

[dowski]

Thank you!

[benoitc]

On Thu, Oct 4, 2012 at 3:27 PM, Christian Wyglendowski <
notifications@github.com> wrote:

Thank you!

Sorry for the delay. I'm just starting to recover but I generally release
faster :)

• benoit