[DO NOT MERGE] test: Scalpel shadow — jackson2 2.22.0→2.21.2#24336
[DO NOT MERGE] test: Scalpel shadow — jackson2 2.22.0→2.21.2#24336gnodet wants to merge 6 commits into
Conversation
Add a shadow comparison section to CI PR comments showing what Maveniverse Scalpel's skip-tests mode would have tested — without affecting actual test execution. Changes: - incremental-build.sh: configure Scalpel with skipTestsForDownstreamModules and fetchBaseBranch=false, add writeScalpelComparison() for collapsible PR comment section with failure reporting - pr-build-main.yml / sonar-build.yml: add base branch fetch step for Scalpel's merge-base detection in shallow CI clones, restore checkout v7 - CI-ARCHITECTURE.md: document shadow comparison approach and configuration - Scalpel upgraded to 0.3.7: fixes inflated affectedModules count for parent POM property changes (scalpel#39) and skipTestsForDownstreamModules
When a matrix build has one JDK failing and another cancelled, the cancelled JDK's cleanup steps upload a ci-comment artifact without the comment file, overwriting the failed JDK's artifact that had it. Only upload the ci-comment artifact when incremental-test-comment.md actually exists, preventing a cancelled build from clobbering a completed build's comment. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Scalpel's auto-detection of GITHUB_BASE_REF via Maven system properties
(env.GITHUB_BASE_REF) is fragile — it can fail in CI rerun contexts or
with certain Maven wrapper configurations, causing the report to silently
not be generated.
Fix: always pass -Dscalpel.baseBranch=origin/${GITHUB_BASE_REF:-main}
explicitly. Also add a git merge-base pre-check and improved diagnostics
(tail of Scalpel log, broader grep) when the report is not found.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Instead of a fixed --depth=200 fetch that can miss the merge base for long-lived or stale branches, try 200 → 1000 → unshallow until git merge-base succeeds. Most PRs resolve at depth 200 (no extra cost); only old branches need the deeper fetches. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
🌟 Thank you for your contribution to the Apache Camel project! 🌟 🐫 Apache Camel Committers, please review the following items:
|
|
🧪 CI tested the following changed modules:
✅ POM dependency changes: targeted tests included Changed properties: jackson2-version Modules affected by dependency changes (6)
🔬 Scalpel shadow comparison — compile: +51, test: +22Maveniverse Scalpel detected 57 affected modules via effective POM comparison (vs 6 from grep-based detection). Changed properties: jackson2-version Changed managed dependencies: com.fasterxml.jackson:jackson-bom Skip-tests mode would test 28 modules (7 direct + 21 downstream), skip tests for 29 (generated code, meta-modules) Modules Scalpel would test (28)
Modules with tests skipped (29)
All tested modules (59 modules)
|
Move all Scalpel-related output below a separator line at the end of the CI comment, with a one-line diff summary showing what Scalpel would add/remove vs the current grep-based detection (e.g. "compile: +51, test: +22"). The top section now shows only what the existing grep mechanism found, making it easy to see what each approach contributes. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
502b203 to
4c29a89
Compare
Test PR for validating the Scalpel shadow comparison in #22524.
Bumps
jackson2-version2.22.0 → 2.21.2 to trigger Scalpel's managed dependency detection. Jackson is widely used via BOM import — Scalpel should find ~157 affected modules vs grep's ~6.Do not merge — close after CI validates the shadow comparison section.
Claude Code on behalf of @gnodet