Wagon perf improvements

[kenkania]

Add a PR-triggered “Perf Test” GitHub Actions workflow with concurrency control and JDK 21 that authenticates to Google Cloud and introduce a CLI run_test.sh to run Maven dependency resolution tests for performance validation

• Add a CI workflow at perf-test.yaml that runs on pull_request with concurrency cancellation, configures JDK 21 with Gradle cache, authenticates to Google Cloud via workload identity provider and service account, defines a matrix over three test inputs, and executes mvn --version.
• Add a CLI script at run_test.sh that prepares a local Maven repo, writes .mvn/extensions.xml and settings.xml for a Google Artifact Registry wagon when configured, and runs mvn dependency:go-offline locally or in Docker with logging.
• Add minimal test Maven projects at tests/maven-0-artifacts/pom.xml, tests/maven-1-artifact/pom.xml, and tests/maven-100-artifacts/pom.xml targeting Java 21.
• Add additional sample projects at tests/quarkus-starter/pom.xml and tests/payment-svc/pom.xml for broader test coverage.
• Update ignores in .gitignore to include bin/ and target/.

📍Where to Start

Start with the workflow definition in perf-test.yaml to understand CI behavior, then review the command flow in run_test.sh and the test project POMs under tests/.

---

Macroscope summarized 7ce6ad9. (Automatic summaries will resume when PR exits draft mode or review begins).

[macroscopeapp]

A failed mvn (or Docker) command can be reported as success because the pipeline’s exit status is that of tee, not the build command.

This happens because set -e is enabled but set -o pipefail isn’t, so failures in pipelines are ignored.

Consider enabling pipefail (e.g. add set -o pipefail after set -e) so that the script exits on any command failure within pipelines.

+set -o pipefail

🚀 Reply to ask Macroscope to explain or update this suggestion.

👍 Helpful? React to give us feedback.

[macroscopeapp]

The script can unintentionally copy the entire filesystem root into the test directory if TEST_DIR is empty or unset, which is dangerous and may partially succeed before failing.

This occurs because cp -rf $TEST_DIR/. "$RUN_DIR" expands to cp -rf /. "$RUN_DIR" when TEST_DIR is empty or unset.

Consider validating that TEST_DIR is non-empty and points to an existing directory before performing the copy, and exit with an error if the check fails.

+if [[ -z "$TEST_DIR" || ! -d "$TEST_DIR" ]]; then
+    echo "Error: TEST_DIR must be a non-empty directory" >&2
+    exit 1
+fi

🚀 Reply to ask Macroscope to explain or update this suggestion.

👍 Helpful? React to give us feedback.