HTTPChannel should ignore Content-Length when Transfer-Encoding=chunked

	enolte reported
Trac ID	trac#6148
Type	defect
Created	2012-10-18 21:39:54Z

RFC 2616 (http://www.w3.org/Protocols/rfc2616/rfc2616-sec4.html#sec4.4) says that the Content-Length header should be ignored when the Transfer-Encoding is chunked.

twisted.web.http.HTTPChannel.headerReceived() mistakenly sets _transferDecoder to _IdentityTransferDecoder and length to non-None whenever it receives a "Content-Length: ###" header after a "Transfer-Encoding: chunked" header.

This is causes the Request to receive the wrong body since it includes the chunk headers as well as the chunk data.

A possible solution is:

    def headedReceived(self, line):
        ...
        if header == 'content-length' and not isinstance(self._transferDecoder, _ChunkedTransferDecoder):
        ...

Another solution is to set self._transferDecoder in allHeadersReceived() where the method has access to all the headers.

trac-id__6148 6148
type__defect defect
reporter__enolte enolte
priority__normal normal
milestone__ 
branch__ 
branch_author__ 
status__new new
resolution__None None
component__web web
keywords__ 
time__1350596394000000 1350596394000000
changetime__1584359757768872 1584359757768872
version__None None
owner__ 

Start in twisted.web.http.HTTPChannel.headerReceived() and compare its handling of Content-Length after Transfer-Encoding: chunked with RFC 2616 section 4.4. Verify that a chunked request body is decoded without including chunk headers or using the Content-Length value; add or update the relevant HTTP tests if the surrounding test location is identified.