XMLHttpRequest: PUT, DELETE, HEAD and all other methods but POST actually do a GET

• Component: DOM
• Status: RESOLVED FIXED

[Henri Sivonen]

Steps to reproduce:
1) Try to do a PUT or a DELETE using XMLHttpRequest.

Actual results:
A GET request is issued.

Expected results:
Expected the full suite of HTTP verbs to be supported or at least the API to
fail detectably when a verb is not supported.

Additional information:
In order to implement a client for a clean RESTful Web service supporting full
CRUD capabilities, PUT and DELETE are needed, too.

[Adam Trachtenberg]

This also applies to other verbs not explicitly defined in the HTTP spec, such
as SEARCH used in HTTP-DAV.

In my case, this prevents me from writing a client to access Microsoft Exchange
servers.

[Stuart Morgan]

Comfirmed in ToT by source inspection; anything unknown is treated as a GET.

[Eric Seidel (no email)]

Would be nice to have a test case.  Although w/o server side support, that might be hard.

[Alexey Proskuryakov]

Writing a synthetic test shouldn't be too hard (one could take one from bug 3810 as a model), but it would 
be very helpful to have an example of a "real life" application that uses this.

[Alexey Proskuryakov]

*** Bug 7390 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

Created attachment 6776 [details]
proposed patch

There are some differences between this implementaion and Firefox:
1) For null data, "Content-Length: 0" header is sent (isn't sent in Firefox). Looks like better behavior to me; besides, I don't think we have a choice, because this header is added by NSURLConnection.
2) HEAD is not implemented - there seems to be a crashing bug in NSURLConnection, <rdar://4460899>.
3) After #2 is fixed, we won't be sending a body in HEAD requests, just like we don't send it in GET ones. Firefox ignores the body in GET requests, but sends it in HEAD ones, which looks like a bug to me.

Also fixed a bug: send(null) was actually sending characters "null" instead of an empty body.

[Darin Adler]

Comment on attachment 6776 [details]
proposed patch

Should use String instead of QString in new code, like the new parameter to TransferJob. String converts itself into NSString without even calling getNSString().

[Alexey Proskuryakov]

Created attachment 6786 [details]
proposed patch

Addressed the comments.

[Darin Adler]

Comment on attachment 6786 [details]
proposed patch

Change looks fine as is, but I have three comments:

    1) Is there a test for the change in JSXMLHttpRequest.cpp? That change looks fine, but I'd like to see a test that was broken before and fixed now.

    2) Since method is now a String you can use equalIgnoringCase rather than calling lower when comparing the method with head.

    3) I would like the client pointer to be the first parameter to TransferJob and the method the second parameter, because the client pointer sort of a "back pointer to the claler" while the rest of the parameters are the specification of what the job is to do.

The most important of these three is (1), but I think none are a big enough deal to prevent landing the patch. I'll say r=me.

[Alexey Proskuryakov]

(In reply to comment #9)
1) Yes, the tests call send(null) for all methods, and there were a lot of failures before I fixed this.
2) Actually, method in XMLHTTPRequest class is still a QString, I haven't changed it.
3) Done.

[Sergio Villar Senin]

Committed r78043: <http://trac.webkit.org/changeset/78043>

[Sergio Villar Senin]

(In reply to comment #11)
> Committed r78043: <http://trac.webkit.org/changeset/78043>

Sorry for the noise. I just mentioned this bug in a ChangeLog, it seems that webkit-patch was too smart this time.

[Alexey Proskuryakov]

This still remains fixed :)

[Lucas Forschler]

Mass moving XML DOM bugs to the "DOM" Component.