Remy Maucherat wrote:
Filip Hanik - Dev Lists wrote:
gents, I know, back again, and totally bastardizing the HTTP protocol,
I have a test request that looks like this:
----------------------------------
POST /comet HTTP/1.1CRLF
User-Agent: Jakarta Commons-HttpClient/3.0.1CRLF
Host: 127.0.0.1:8080CRLF
Transfer-Encoding: chunkedCRLF
CRLF
10CRLF
test data test 1CRLF
---------------------------------------
If I keep sending up
---------------------------------------
10CRLF
test data test 1CRLF
---------------------------------------
we are all happy, however, if I bastardize the next request and send
up the same request twice, so the 2nd request of the HTTP request
above becomes the body of the first chunked request, then suddenly
the chunked input filter will set its remaining bytes to over 50k
remaining bytes, and everything gets streamed into the servlet.
The ChunkedInputFilter is written to handle correct chunks very well,
but when incorrect chunked bodies come through, it passes all the
content through, instead of throwing an error.
If no one objects, I would like to add in some error checking, even
though it is the clients responsibility to provide correct data, the
web server should not let incorrect data through as that could be a
semi http buffer overflow :)
The only thing which determines if a chunk is correct is if the chunk
length is valid and readable. The parseChunkHeader method includes the
ability to return false and throw an IOE if an invalid header is
received.
yes, and there is where the bug lies, it doesn't throw an IOException
when it gets an invalid header, I'll fix it and submit the patch for
review so that we can get it taken care of.
Filip
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
