[SYNAPSE-662] ClientWorker overriding character encoding

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.2
• Fix Version/s: 2.0
• Assignee: Hiranya Kasub Jayathilaka
• Reporter: Myles Bunbury
• Created: 09/Jun/10 18:59
• Updated: 20/Sep/10 05:34
• Resolved: 20/Sep/10 05:34

Description

The following code exists in the org.apache.synapse.transport.nhttp.ClientWorker class' run() method:

responseMsgCtx.setProperty(
Constants.Configuration.CHARACTER_SET_ENCODING,
contentType.indexOf(HTTP.CHARSET_PARAM) > 0 ?
charSetEnc : MessageContext.DEFAULT_CHAR_SET_ENCODING);

This fails however for the following Content-Type HTTP header:
application/soap+xml; action="urn:echoResponse";charset=UTF-16

BuilderUtil.getCharSetEncoding(contentType) is called a few lines up and correctly extracts the UTF-16 character set value, but then overrides this and sets it to UTF-8. It does this because the value of org.apache.http.protocol.HTTP.CHARSET_PARAM is "; charset=". That is, it's failing because the response does not have a space between the charset parameter and the previous parameter.

Reading through the HTTP specs, I haven't come across anything that says either that whitespace is permissible or not permissible here. In my view the code should therfore be flexible enough to handle either case.

I would suggest simpifying the code to:

responseMsgCtx.setProperty(Constants.Configuration.CHARACTER_SET_ENCODING, charSetEnc);

as the BuilderUtil.getCharSetEncoding() code above the offending code seems to be sufficient at first glance.