On 6/12/11 3:51 PM, Jim Talbut wrote:
Willem,
Am I right in thinking that you are looking at it on the client side?
Both Sun and OpenJDK send "application/octet-stream" as the
content-type, so that functionality isn't working for either of them.
And even if it was the test would be broken because the declared
content-type is not used server side.
Because the multipart attachment is saved as a file, current we can't
get the content type from the request message.
If you mean that should be on the server side I don't think it's right
for the server to try to guess the mime-type of a file given to it by a
client when the client has explicitly stated what the mime type should
be (though it'd be great to have a backup for when the content-type
isn't sent).
The test is failed because the OpenJDK can't build up right content-type
from the file name. As the server save the multipart attachments into
files rightly, we don't need spend lots of time to find out what
content-type of the attachment is.
So I see the root cause as the fact that jetty doesn't make the declared
content type available for files accessed via getAttributes.
You are right.
I can't see anywhere that CamelFileDataSource is used.
The test client uses the commons http client, which just uses a standard
FileDataSource (one could easily provide a CamelFilePart to use
CamelFileDataSource).
We are tent to test if the server can build up a right data handler from
multipart attachments. We could comment out assert of conent-type of the
test case by now.
Jim
On 12/06/2011 01:43, Willem Jiang wrote:
It relates to how the FileTypeMap implements.
In CamelFileDataSource you can see the content-type is determined by
the below method.
public String getContentType() {
if (typeMap == null) {
return FileTypeMap.getDefaultFileTypeMap().getContentType(fileName);
} else {
return typeMap.getContentType(fileName);
}
}
As you know Sun and OpenJDK has different implementation, so we get
the test failure.
According to the comments of javax.activation.MimetypesFileTypeMap[1]
of the OpenJDK. You can see the content-type of NOTICE.txt should be
mapped to "text/plain". But it could be something wrong when it lookup
the mime.types file.
I think you can fix it by apply a right mime.types file in your
environment.
[1]http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/6-b14/javax/activation/MimetypesFileTypeMap.java#MimetypesFileTypeMap
On 6/11/11 3:14 AM, Jim Talbut wrote:
I think both the test in camel-jetty and the code in camel-http are
wrong.
In camel-http (DefaultHttpBinding) we have:
protected void populateAttachments(HttpServletRequest request,
HttpMessage message) {
// check if there is multipart files, if so will put it into DataHandler
Enumeration names = request.getAttributeNames();
while (names.hasMoreElements()) {
String name = (String) names.nextElement();
Object object = request.getAttribute(name);
LOG.trace("HTTP attachment {} = {}", name, object);
if (object instanceof File) {
String fileName = request.getParameter(name);
message.addAttachment(fileName, new DataHandler(new
CamelFileDataSource((File)object, fileName)));
}
}
}
That creates the DataHandler for the uploaded file without specifying
the content-type.
So different implementations of DataHandler may well have different
defaults (hence the difference between Sun and OpenJDK).
This is wrong, but I can't see how to get the correct content-type for
the uploaded file in order to add it to the constructor args.
The test is explicitly testing that the DataHandler has the same
defaults as the Sun JDK, which is not helpful.
It ought to test that the content-type is application/octet-stream, but
that would require changing camel-http.
There are two options:
1. Fix camel-http so it does set the content-type correctly (though I
couldn't work out where to get it from).
We'd then have to fix the test to check for the correct content-type.
2. Remove the assertion from the test.
Jim
On 10/06/2011 07:20, Jim Talbut wrote:
On 09/06/2011 19:54, Jim Talbut wrote:
On 09/06/2011 15:32, James Talbut wrote:
On Thu, Jun 09, 2011 at 10:13:00PM +0800, Willem Jiang wrote:
Maybe it relates to the openjdk.
Can you try the latest SUN/Oracle JDK 1.6 to run the test ?
That's possible, I should be able to try later today.
I've also tried on Windows 7 and that works (breaks in camel-ftp,
but that's another matter).
Jim
That was it, how interesting.
I downloaded the Sun JDK, set JAVA_HOME to point to it and ran maven
and it went correctly.
Should Camel support alternative JVMs?
Jim
Curiouser and curiouser...
I'm starting to think that it's actually the OpenJDK that's right and
the test is wrong.
The test has:
Part[] parts = {
new StringPart("comment", "A binary file of some kind"),
new FilePart(file.getName(), file)
};
There is nothing there to specify the content-type and there seems to
be an assumption that it's binary, not text (though it is actually
text).
The source for FilePart has:
public static final String DEFAULT_CONTENT_TYPE =
"application/octet-stream";
and that is used unless the content type is specified in the
constructor (which it isn't).
So the string part should have a content type of "text/plain" and the
file part should have a content type of "application/octet-stream".
The test has:
DataHandler data = in.getAttachment("NOTICE.txt");
assertNotNull("Should get the DataHandle NOTICE.txt", data);
assertEquals("Get a wrong content type", "text/plain",
data.getContentType());
Verifying that the file part is "text/plain".
The on-the-wire data with OpenJDK is:
--lDIZtpp0tBIGfhQWzHIZ5kV7QpREJSj-Cmp3Uf8C
Content-Disposition: form-data; name="comment"
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 8bit
A binary file of some kind
--lDIZtpp0tBIGfhQWzHIZ5kV7QpREJSj-Cmp3Uf8C
Content-Disposition: form-data; name="NOTICE.txt";
filename="NOTICE.txt"
Content-Type: application/octet-stream; charset=ISO-8859-1
Content-Transfer-Encoding: binary
=========================================================================
== NOTICE file corresponding to the section 4 d
of ==
== the Apache License, Version
2.0, ==
== in this case for the Apache Camel
distribution. ==
=========================================================================
This product includes software developed by
The Apache Software Foundation (http://www.apache.org/).
Please read the different LICENSE files present in the licenses
directory of
this distribution.
--lDIZtpp0tBIGfhQWzHIZ5kV7QpREJSj-Cmp3Uf8C--
The on-the-wire data with Sun JDK is:
--qhLzs_ECq12ft9Tp3B1ayD-jjjzshV
Content-Disposition: form-data; name="comment"
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 8bit
A binary file of some kind
--qhLzs_ECq12ft9Tp3B1ayD-jjjzshV
Content-Disposition: form-data; name="NOTICE.txt";
filename="NOTICE.txt"
Content-Type: application/octet-stream; charset=ISO-8859-1
Content-Transfer-Encoding: binary
=========================================================================
== NOTICE file corresponding to the section 4 d
of ==
== the Apache License, Version
2.0, ==
== in this case for the Apache Camel
distribution. ==
=========================================================================
This product includes software developed by
The Apache Software Foundation (http://www.apache.org/).
Please read the different LICENSE files present in the licenses
directory of
this distribution.
--qhLzs_ECq12ft9Tp3B1ayD-jjjzshV--
Note that both have the correct content-type
(application/octet-stream) for the file.
So I think there is a bug in camel-jetty when used with the Sun JDK.
Still looking, but please stop me if you think I've gone wrong so far!
Jim
--
Willem
----------------------------------
FuseSource
Web: http://www.fusesource.com
Blog: http://willemjiang.blogspot.com (English)
http://jnn.javaeye.com (Chinese)
Twitter: willemjiang
Weibo: willemjiang
