[3.33] Move quarkus-cxf-test-util-internal in camel-quarkus-integration-test-cxf-soap-grouped back to use correct property#1981
Conversation
…-cxf-soap-grouped back to use correct property
|
/cc @aloubyansky (3.33), @gsmet (3.33), @jmartisk (3.33), @rsvoboda (3.33) |
|
cc @ppalaga |
|
This change won't have any effect, since the This is long standing issue. Basically, you should pretty much ignore which properties are used for versions in the generated project, unless the actual values appear to be wrong. Here is my guess why It looks like somehow the test artifact used for this test got unaligned with the rest of the CXF artifacts integrated in the platform, which is why you noticed a "hardcoded" 3.33.2 version first. To fix this, the integrated test artifact (that's used for this test module generation) needs to be aligned with the rest of the CXF artifacts. |
|
Note: to get this fixed in the upcoming 3.33.2 platform release, I'd need a fix by tomorrow, @ppalaga |
|
A quick workaround could be to enforce the proper version of this artifact in the platform itself, I.e. by adding it to the cxf bom. I am not sure if there is more appropriate quick fix. |
|
Given that
I would probably defer the fix to 3.33.3 |
|
I have looked at the reported issue as much as I could and here is my take: The fact that I do not see anything Camel Quarkus or Quarkus CXF could do better. The issue is that the BOM generator chooses I see two options to fix this: A. The BOM generator stops choosing the wrong properties for the artifacts. B. The QE team will regenerate the BOMs when building their custom platform - something like |
|
It's the second option. Platform BOMs are flattened and do not include properties. So not regenerating them will not be equivalent to actually updating the Quarkus version in the platform. |
|
@aloubyansky the issue we have is not with the generated BOMs, but with generated test modules. Those are generated by GeneratePlatformProjectMojo too. The flat version literal -> property backtracking in GeneratePlatformProjectMojo delivers wrong results. I think the option A is still valid. |
|
What would be the point of running tests with a Quarkus version that different from the generated BOMs though? Or am I misunderstanding the flow? |
It'd be great to improve on that in any case, of course. |
We (Quarkus QE) building custom platform for our testing. In last build we spotted failure when building platform for 3.33. In 9f12b59#diff-405d042550bf01fefb20a9c7759b93fa0c37ca2eddcc3fce410c6f2d410b30f7L33 it was changed to hard coded version (where I don't see that much problem) and in c17fb5c#diff-405d042550bf01fefb20a9c7759b93fa0c37ca2eddcc3fce410c6f2d410b30f7L33 it was changed again but to
quarkus.versionwhich should not be. When we building with custom Quarkus version it failing to build as it can't find thequarkus-cxf-test-util-internalwith our custom Quarkus version (3.33.999-SNAPSHOT).I don't know if if the change in from
quarkus-cxf.versionwas for some specific reason or not so atm I proposing go back toquarkus-cxf.versioncc @ppalaga @jmartisk as you are the one which updated this
Edit when I run the
./mvnw -DsyncI see that this was overwritten back toquarkus.version