Croway opened a new pull request, #1959:
URL: https://github.com/apache/camel-spring-boot/pull/1959

   Cherry-pick of #1932 onto `camel-spring-boot-4.22.x`.
   
   **Original PR:** #1932 — platform-http path variables follow the path Spring 
matched
   **JIRA:** [CAMEL-24577](https://issues.apache.org/jira/browse/CAMEL-24577)
   
   ### What it fixes
   
   `SpringBootPlatformHttpBinding` resolved the `{placeholder}` values of a 
`platform-http` consumer's
   path against `getRawPath(request)` — the undecoded request URI with the 
context path stripped off.
   Spring, however, dispatches the request against its own parsed 
`RequestPath`, whose segments are
   percent-decoded and have any matrix parameters removed. When the two 
disagreed, the header exposed
   to the route did not match what Spring actually matched: a request to 
`/greeting/%61dmin` was routed
   to the `admin` value but the header carried the raw `%61dmin`, and 
`/greeting/name;v=1` carried
   `name;v=1` instead of `name`.
   
   The fix evaluates the placeholders against the segments Spring matched, 
obtained via
   `ServletRequestPathUtils.parse(request)`, instead of the raw request path. 
This also brings the
   Spring MVC engine's behavior in line with the vertx platform-http engine, 
which already decodes path
   parameters. The path is parsed fresh rather than read from the `RequestPath` 
attribute Spring caches
   on the request, because the platform-http consumer services the request on 
its own executor and by
   the time it runs, Spring's dispatch may have already cleared that attribute 
— reading it back would
   make correctness depend on timing.
   
   `getRawPath()` itself is untouched, so `Exchange.HTTP_PATH` still reports 
the raw path, preserving the
   context-path handling from CAMEL-22116 and CAMEL-23191.
   
   ### Verification on this branch
   
   - The cherry-pick (`git cherry-pick -x`) applied with a clean auto-merge (no 
conflict markers) on top
     of this branch's current tip — the task brief anticipated a possible 
conflict with the CAMEL-24496
     backport on the same file, but that backport (#1953) is still an open, 
unmerged PR at the time of
     this backport, so there was nothing in `origin`'s actual history to 
conflict with.
   - Both cherry-picked test files compiled and ran as-is: this branch already 
matches `main`'s test
     conventions here 
(`org.apache.camel.test.spring.junit6.CamelSpringBootTest`, and
     `SpringBootPlatformHttpConsumer.service()` returning 
`CompletableFuture<Void>`), so no adaptation
     was needed (unlike the 4.18.x backport, which required fixing both of 
those for that branch's older
     conventions).
   - Ran `../../mvnw verify` in 
`components-starter/camel-platform-http-starter` (against the module
     built with its parent/`camel-version` temporarily repointed to the 
released 4.22.0, since the
     branch's 4.22.1-SNAPSHOT parent isn't published, plus the same repoint for 
the literal
     `org.apache.camel` core SNAPSHOT versions in 
`tooling/camel-spring-boot-dependencies/pom.xml`): a
     first run showed `SpringBootPlatformHttpRequestTimeoutTest.testGetAsync` 
failing all 3 surefire
     attempts with `IllegalStateException: The response object has been 
recycled...`. I reproduced the
     identical failure on this branch's tip *before* the cherry-pick as well 
when running the full
     suite, then re-ran the full suite on both the pre-cherry-pick tip and the 
cherry-picked branch again
     and got a clean pass both times (123 unit tests + 11 integration tests, 0 
failures), and the test
     also passes reliably in isolation on both. This points to a 
load/timing-sensitive flake in that
     timeout test unrelated to this change — the change here doesn't touch 
response timeout/writing
     logic at all — rather than a regression caused by this cherry-pick.
   - `mvn install -pl components-starter/camel-platform-http-starter -am 
-DskipTests -Dfastinstall` built
     cleanly. The temporary parent/`camel-version` and dependencies-pom edits, 
plus any catalog
     regeneration artifacts, were reverted before pushing; the pushed commit 
touches only the same files
     as the original PR.
   
   _Claude Code on behalf of Federico Mariani_
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to