Hi Andy,
Yes, sure, the test suite in question is
org.apache.cxf.systest.jaxrs.JAXRSSoapBookTest, in
particular the test cases which use
org.apache.cxf.systest.jaxrs.BookSubresource. In the
nutshell, in case of @FormParam("") the key is empty string ("") but it is not
stored
as empty string in params (rather expanded). In this case the decode check is
noop
essentially since there are no values stored under empty key. Please let me know
if you need simpler test case(s), I can add a few shortly.
Thanks a lot for looking.
Best Regards,
Andriy Redko
AM> Hi Andriy,
AM> Do you have a test case that shows the problem with the empty string
AM> FormParam? It does look like it might double-decode unless we also add a
AM> check later in the method to avoid that case.
AM> Thanks for pointing this out.
AM> Andy
AM> On Sun, Apr 18, 2021 at 9:19 PM Andriy Redko <[email protected]> wrote:
>> Hey guys,
>> I was lucky enough to have time during the weekend to fix a few TCK test
>> cases caused by
>> deployment issues. @David you may be interested in backporting [1] to your
>> repository. So
>> we are down to 61 failing test cases (+2 errors related to async
>> processing if I am not
>> mistaken). Regarding master builds, I am not sure we need yet another job,
>> we usually
>> very disciplined in backporting the fixes to 3.4.x / 3.3.x.
>> @Andy I have run into the edge case with @Encoded, when parameter name is
>> empty, fe @FormParam("").
>> I believe this is not standard but CXF supports that. I came up with this
>> quick fix [2] but I am
>> not sure it won't lead to multiple decoding attempts, still looking for
>> better solution.
>> Thanks!
>> [1]
>> https://github.com/apache/cxf/commit/7b820e3f193dc6fd0d73272c646d9b9aae2a1377
>> [2]
>> https://github.com/apache/cxf/commit/3b1ddd7c5e7a450d0c768e40850e5a899a8720b3
>> Best Regards,
>> Andriy Redko
>> > Do we want another job to build master? Only mention that as I see
>> Andy's PR was merged there.
>> I just backported it to 3.4.x-fixes so that it will show up in 3.4.4.
>> Thanks,
>> Andy
>> On Fri, Apr 16, 2021 at 2:04 PM David Blevins <[email protected]>
>> wrote:
>> > On Apr 16, 2021, at 4:46 AM, Andriy Redko <[email protected]> wrote:
>> > Good idea to create a subtask for each failure, personally haven't found
>> time to do that, my bad,
>> > thank you.
>> Happy to help!
>> > I will go over them and point to the relevant pull requests, some of
>> these issues
>> > have been worked on already (but not integrated yet).
>> Thank you so much.
>> > FYI: Just a hint (if you haven't noticed) each CXF TCK run has reports
>> and jtr files + binaries,
>> > as build artifacts, for example [1].
>> I totally did miss that :) Thank you for that! I downloaded the zip and
>> it has it all right there. Nice work.
>> Do we want another job to build master? Only mention that as I see Andy's
>> PR was merged there.
>> -David
>> > [1]
>> https://ci-builds.apache.org/job/CXF/job/CXF-JAXRS-TCK/lastSuccessfulBuild/artifact/
>> > Best Regards,
>> > Andriy Redko
>> > DB> Hi All,
>> > DB> Spinning a new thread as to not overload the original "@Encoded TCK
>> issue" too much.
>> > DB> Thanks, Andriy, for the Jenkins information and TCK setup scripts.
>> Thanks also, Andy, for the link to the proper JIRA issue.
>> > DB> Here's what I've done. I converted the Jenkinsfile into a script
>> and hammered on it so it is more developer friendly. Specifically, you can
>> run one test or a chunk of tests. As well it won't redo any setup steps
>> unless needed. It calculates a sha on your cxf-core jar and if that hasn't
>> changed, it just goes straight to running the requested test. This should
>> make it easier for us to debug and work locally.
>> > DB> - https://github.com/tckwork/cxf
>> > DB> If we want this, tell me where it should go and I'll submit a PR.
>> On the TomEE side, we keep a separate repo just for the TCK setup, which is
>> kind of handy as it tends to grow over time and one setup can work for all
>> our branches. But anything works really.
>> > DB> I then downloaded the consoleText from the latest CXF-JAXRS-TCK job,
>> parsed out each failing test, ran it locally, saved all the output and
>> wrote a script to create a subtask on CXF-7996 for each failing test:
>> > DB> - https://issues.apache.org/jira/browse/CXF-7996
>> > DB> The description of each subtask is the test name, a chunk of the
>> output, and instructions on how to run the test. Attached to each subtask
>> is the full TCK test output and jtr file so people can browse the failures
>> without having to run them directly. Also attached is a `test.txt` file
>> containing the test name so there is an easy way for future automated tools
>> to know what issue to update as status changes.
>> > DB> If I've possibly overstepped creating all those subtasks, let me
>> know and I can easily move them or delete them. Totally fine.
>> > DB> -David
