Hi,
Sorry for the double mail.
I forgot to add that, while I know the .jar versions are not the latest,
I checked the code in Fisheye and from what I understood the issue
corresponds with the latest code. The problems seems to arise from the
code in the line 627
<https://fisheye.apache.org/browse/cxf/rt/frontend/jaxrs/src/main/java/org/apache/cxf/jaxrs/impl/UriBuilderImpl.java?hb=true#to627>
[2]: if the path does not start with a "/", the varirable
schemeSpecificPart is set to null, which in turn prevents the code block
in the line 161
<https://fisheye.apache.org/browse/cxf/rt/frontend/jaxrs/src/main/java/org/apache/cxf/jaxrs/impl/UriBuilderImpl.java?hb=true#to161>
[3] to run.
Regards
On 16/06/17 23:52, Rubén Pérez wrote:
Hi,
This is my first email to the list, so I hope I will not miss
something or do something wrong. I'm writing here instead of reporting
a bug directly, mainly because I'm not sure if I can even register in
the bug tracker, so apologies if that was the expected way to go.
Cutting to the chase, I believe there is a bug in the UriBuilderImpl.
Specifically, when an instance of the class is created through the
'.fromURI' method, and the provided URI does not start with a "/"
(i.e. it has no schema and authority and its path is not absolute),
UriBuilderImpl assumes incorrectly that the URI is non-hierarchical
and therefore the '.build()' method will always return the provided
URI without modification, regardless of which edit methods (e.g.
"queryParam", "path", etc.) are called on the instance.
In order to illustrate what I mean, I created a very simple program
(attached). I compiled it with:
javac -cp '.:<path_to_the_jar>/javax.ws.rs-api-2.0.1.jar' UriBuilderBug.java
and runned it with:
java -cp
'.:/<path_to_the_jar>/javax.ws.rs-api-2.0.1.jar:<path_to_the_jar2>/cxf-rt-frontend-jaxrs-3.1.11.jar:<path_to_the_jar3>/cxf-core-3.1.11.jar'
UriBuilderBug
The RFC-3986, in its section 4.2
<https://tools.ietf.org/html/rfc3986#section-4.2> [1] defines the
syntax for valid relative URI references and explicitly allows
references without schema or authority that do *not* start with a
forward slash. Perhaps part of the confusion comes from the fact that
the ABNF defining the relative references is as follows:
relative-part = "//" authority path-abempty
/ path-absolute
/ path-noscheme
/ path-empty
, where the "/" symbol representing the different alternatives can be
easily mistaken for a literal "/" in the URI itself.
So that's it. Thanks for reading till here. Please let know your
opinions about this and, if I'm right, it would be great to get it
fixed soon. If something is not clear or you have any questions, I'm
willing to answer them. If there's a reason for the current behaviour,
or if I am incorrectly interpreting the standard, I'd also be very
glad to know.
Best regards
--
Rubén Pérez Vázquez
*Universität zu Köln*
/Regionales Rechenzentrum (RRZK)/
Weyertal 121, Raum 4.05
D-50931 Köln
✆: +49-221-470-89603
[1] https://tools.ietf.org/html/rfc3986#section-4.2
--
Rubén Pérez Vázquez
*Universität zu Köln*
/Regionales Rechenzentrum (RRZK)/
Weyertal 121, Raum 4.05
D-50931 Köln
✆: +49-221-470-89603
[2]
https://fisheye.apache.org/browse/cxf/rt/frontend/jaxrs/src/main/java/org/apache/cxf/jaxrs/impl/UriBuilderImpl.java?hb=true#to627
[3]
https://fisheye.apache.org/browse/cxf/rt/frontend/jaxrs/src/main/java/org/apache/cxf/jaxrs/impl/UriBuilderImpl.java?hb=true#to161
