Spring-hateoas: Links are relative to request context

[cheinema]

Since 0.22.0.RELEASE, links are built relative to the request context. The following test case in ControllerLinkBuildUnitTest runs with success against the previous release, but fails with the current.

@Test
public void createsLinkRelativeToContextRoot() {

    request.setContextPath("/ctx");
    request.setServletPath("/foo");
    request.setRequestURI("/ctx/foo");

    assertThat(linkTo(PersonControllerImpl.class).withSelfRel().getHref(), endsWith("/ctx/people"));
}

java.lang.AssertionError: 
Expected: a string ending with "/ctx/people"
     but: was "http://localhost/ctx/foo/people"
Expected :a string ending with "/ctx/people"

This would be a regression to our application due to the invalid links.

[cheinema]

I suspect that this behavior was introduced with #519.

[odrotbohm]

Thanks for giving this a spin so quickly. Indeed our migration to the Spring APIs introduced that glitch. I have a fix ready to be polished here. We should be able to craft a release quickly.

[odrotbohm]

0.23.0.BUILD-SNAPSHOT binaries available. Care to give them a spin before I start the machinery?

[feffef]

I am happy to try that out, we also just noticed a similar regression after upgrading to 0.22.0 (as we've been eagerly awaiting the fix for #525 / #511)

[odrotbohm]

Yeah, please go ahead. Sorry for the trouble. I'm pretty puzzled to see that going out unnoticed :(.

The more important parts of #525 / #511 were fixed in Spring Framework 4.3.5. So you should see very decent improvements by just moving to the latter. 4.3.5 has not really been a required release API wise, I just wanted to wait for it so that we can ship a dependency to it to make sure people get the changes that actually fix the reported issues against Spring HATEOAS.

[feffef]

I can confirm that all links in our project that are generated with ControllerLinkBuilder are construced correctly again (i.e. not relative to the path of the incoming request) when using 0.23.0.BUILD-SNAPSHOT.

[cheinema]

fb0ba25 LGTM

[otrosien]

grml - mea culpa (47cefe6869c9f41fef17d0187377a7cd379a488e)..

[odrotbohm]

Alright. Looks like the issue hasn't appeared downstream as none of the projects (Spring Data) actually use the ControllerLinkBuilder.

@otrosien — Nevermind. We should've had tests for that in the first place.

About to push the buttons…

[odrotbohm]

0.23 is live at the Spring repositories. Maven Central sync seems to have general hickups currently.

[feffef]

Thanks a lot for fixing this so quickly! Everything is working fine with 0.23 - I only had to add the https://repo.spring.io/libs-release repository, as the release indeed doesn't yet seem to be available on maven central.

[otrosien]

@olivergierke - also the git tag has not been pushed back to github yet.

[odrotbohm]

The artifacts are now available via Maven Central. However, the search index doesn't seem to list them yet.

[odrotbohm]

Should be up now.