Conversation
Merging the goal into an execution the project already declares hands it that execution's phase and configuration. Bound to a phase that runs after the tests, the agent path is unresolved when Surefire forks.
The additions are independent of the surefire tag edit, so they can be chained onto the document directly. Ordering is preserved: the dependency plugin is added before surefire, as the scheduled queue did.
MBoegers
marked this pull request as ready for review
August 7, 2026 13:51
timtebeek
approved these changes
Aug 7, 2026
mergify Bot
added a commit
to robfrank/linklift
that referenced
this pull request
Aug 20, 2026
…41.0 to 3.42.0 [skip ci] Bumps [org.openrewrite.recipe:rewrite-migrate-java](https://github.com/openrewrite/rewrite-migrate-java) from 3.41.0 to 3.42.0. Release notes *Sourced from [org.openrewrite.recipe:rewrite-migrate-java's releases](https://github.com/openrewrite/rewrite-migrate-java/releases).* > 3.42.0 > ------ > > What's Changed > -------------- > > * Adopt upstream XML trailing-comment formatting by [`@timtebeek`](https://github.com/timtebeek) in [openrewrite/rewrite-migrate-java#1180](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1180) > * Add mapping for CheckForNull to JSpecify annotation by [`@zbynek`](https://github.com/zbynek) in [openrewrite/rewrite-migrate-java#1179](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1179) > * Skip `UseSetOf`/`UseListOf` for `HashSet`/`ArrayList` subclasses by [`@timtebeek`](https://github.com/timtebeek) in [openrewrite/rewrite-migrate-java#1182](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1182) > * Avoid hardcoded patch version in UpdateSdkManTest by [`@timtebeek`](https://github.com/timtebeek) in [openrewrite/rewrite-migrate-java#1183](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1183) > * Only add the Mockito surefire agent configuration when asked, and keep it minimal by [`@timtebeek`](https://github.com/timtebeek) in [openrewrite/rewrite-migrate-java#1184](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1184) > * Make subpackage recursion explicit in the Jackson JAX-RS JSON rename by [`@timtebeek`](https://github.com/timtebeek) in [openrewrite/rewrite-migrate-java#1185](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1185) > * Add the Mockito agent properties goal in its own execution by [`@MBoegers`](https://github.com/MBoegers) in [openrewrite/rewrite-migrate-java#1186](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1186) > * Derive SDKMAN test versions from the candidate list by [`@timtebeek`](https://github.com/timtebeek) in [openrewrite/rewrite-migrate-java#1188](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1188) > * Do not apply `var` when the generic return type needs the declared type by [`@jevanlingen`](https://github.com/jevanlingen) in [openrewrite/rewrite-migrate-java#1187](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1187) > * Cover legacy Bouncy Castle artifacts and their API changes by [`@timtebeek`](https://github.com/timtebeek) in [openrewrite/rewrite-migrate-java#1189](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1189) > * Add jakarta.validation-api dependency when migrating com.sun.istack.NotNull by [`@steve-aom-elliott`](https://github.com/steve-aom-elliott) in [openrewrite/rewrite-migrate-java#1190](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1190) > > New Contributors > ---------------- > > * [`@zbynek`](https://github.com/zbynek) made their first contribution in [openrewrite/rewrite-migrate-java#1179](https://redirect.github.com/openrewrite/rewrite-migrate-java/pull/1179) > > **Full Changelog**: <openrewrite/rewrite-migrate-java@v3.41.0...v3.42.0> Commits * [`1238ceb`](openrewrite/rewrite-migrate-java@1238ceb) OpenRewrite recipe best practices * [`01d0fe8`](openrewrite/rewrite-migrate-java@01d0fe8) Add `jakarta.validation-api` dependency when migrating `com.sun.istack.NotNul... * [`33354b5`](openrewrite/rewrite-migrate-java@33354b5) Cover legacy Bouncy Castle artifacts and their API changes ([#1189](https://redirect.github.com/openrewrite/rewrite-migrate-java/issues/1189)) * [`6c648a5`](openrewrite/rewrite-migrate-java@6c648a5) Update Gradle wrapper to 9.7.0 * [`7099ad3`](openrewrite/rewrite-migrate-java@7099ad3) Do not apply `var` when the generic return type needs the declared type ([#1187](https://redirect.github.com/openrewrite/rewrite-migrate-java/issues/1187)) * [`ccfa7b0`](openrewrite/rewrite-migrate-java@ccfa7b0) Derive SDKMAN test versions from the candidate list ([#1188](https://redirect.github.com/openrewrite/rewrite-migrate-java/issues/1188)) * [`96e8647`](openrewrite/rewrite-migrate-java@96e8647) [Auto] SDKMAN! Java candidates as of 2026-08-10T1102 * [`26f898e`](openrewrite/rewrite-migrate-java@26f898e) Add the Mockito agent properties goal in its own execution ([#1186](https://redirect.github.com/openrewrite/rewrite-migrate-java/issues/1186)) * [`d82dd47`](openrewrite/rewrite-migrate-java@d82dd47) OpenRewrite recipe best practices * [`55f2e93`](openrewrite/rewrite-migrate-java@55f2e93) Make subpackage recursion explicit in the Jackson JAX-RS JSON rename ([#1185](https://redirect.github.com/openrewrite/rewrite-migrate-java/issues/1185)) * Additional commits viewable in [compare view](openrewrite/rewrite-migrate-java@v3.41.0...v3.42.0)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The problem
The recipe adds
-javaagent:${org.mockito:mockito-core:jar}to Surefire'sargLine, and needs themaven-dependency-pluginpropertiesgoal to define that property. When a project already declared that plugin, the recipe merged<goal>properties</goal>into an execution the project owns:An execution owns the phase and the configuration of every goal it lists, so the merged goal inherited both. Bound to
package,propertiesruns after Surefire forks, the placeholder is passed through verbatim, and the build fails with:Two further problems came from the same merge:
.../executions/execution/goalshad no positional predicate and the visitor no one-shot guard, so a plugin with N executions got<goal>properties</goal>appended to all N goal lists.PropertiesMojoandAbstractDependencyMojo(the base ofcopy-dependencies,unpack,analyze, …) both declare askipparameter. Maven's configuration finalizer matches by parameter name, so an execution carrying<skip>silently skipped the merged goal — the same broken build, with a phase that looks perfectly fine.The fix
The
propertiesgoal now always lands in an execution of its own, never merged into one the project declares:No
<phase>, so Maven binds each goal of the execution to that goal's own default —initializefordependency:properties, verified against the mojo descriptor in every published plugin version from 2.4 to 3.11.0.The explicit
<id>is load-bearing rather than cosmetic. Maven merges executions by id — for parent inheritance and forpluginManagementinjection alike — and an omitted id isdefault. Without an id this execution would be captured by any id-less execution declared upstream and handed its<phase>, re-entering the same bug from another direction; and appending an id-less execution next to a project's own id-less one is a duplicate-id model validation error.The guard is
runsPropertiesGoalAtDefaultPhase: we rely on an existingpropertiesgoal only when its execution declares no<phase>, because that is the one case where its timing is self-evident. Where a project binds the goal to a phase of its own, deciding whether that lands before the tests would mean reasoning about lifecycle ordering, inheritance and property interpolation, so the goal gets its own execution instead. Worst case that is a second run of a goal which only sets properties; it is never a broken build. The check is mirrored against the document because the resolved model is not refreshed between cycles.This collapses the four-branch cascade to a single path and retires
ChangePluginExecutionsandAddOrUpdateChildTagfrom this recipe. Neither can express the operation:ChangePluginExecutionsreplaces the whole<executions>element, andAddOrUpdateChildTagis name-keyed and so cannot append a sibling<execution>. Dropping the former also closes a second route to the reported error — it matches groupId with.orElse(null)whereAddPluginVisitordefaults the implied Apache group, so a<plugin>declared without a<groupId>used to no-op silently while theargLinewas still written.Tests
Four scenarios added:
addsSeparateExecutionWhenExistingExecutionRunsAfterTests— the reported reproductionaddsSeparateExecutionWhenPropertiesGoalIsBoundToAPhase— we do not rely on a phase-boundpropertiesgoaladdsSingleExecutionWhenMavenDependencyPluginHasSeveralExecutions— one execution added regardless of how many exist, with an implied groupId<id>line, three moved the goal out of the project's execution into its ownNotes for review
The three restructured fixtures did not encode broken behaviour. Their executions are phase-less, and Maven binds each goal of a phase-less execution to that goal's own default, so the previous merge was functionally correct there. This change is a correctness fix for the explicit-late-phase case, plus hardening against configuration inheritance and the N-execution amplification.
One pre-existing gap is untouched and worth a separate look: branch selection reads the resolved model, which includes parent-inherited and active-profile plugins, while the edit is anchored at
/project/build/pluginsof the current document. A plugin declared only in a parent or only in a profile therefore gets theargLinewithout thepropertiesgoal. It is not the failure reported here, and no test covers it today.