feat(ci): preliminary AI analysis of testrun failures

[mkoura]

Add a regression-workflow step that, on testrun failure, invokes anthropics/claude-code-base-action against the still-uncompressed allure-results/, latest_artifacts/ symlink, errors_all.log and friends, producing failure_analysis.md. The analysis is uploaded with testrun files and inlined into the failure mail body.

To keep originals available without untarring, save_artifacts.sh now creates a latest_artifacts symlink to the renamed artifacts_/ dir, and create_results.sh no longer wipes allure-results/ after archiving.

[Copilot]

Pull request overview

Adds CI support for preliminary AI-assisted analysis of failed regression test runs, preserving uncompressed failure artifacts so they can be inspected before upload and mail notification.

Changes:

• Adds a Claude analysis step to the reusable regression workflow and includes its output in artifacts/mail.
• Keeps allure-results/ after archiving and creates a latest_artifacts symlink.
• Passes the Claude OAuth token through regression and nightly workflow callers.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
runner/save_artifacts.sh	Creates latest_artifacts symlink to the timestamped artifacts directory.
runner/create_results.sh	Preserves allure-results/ after creating the archive.
.github/workflows/regression.yaml	Passes Claude token to reusable regression workflow.
.github/workflows/regression-dbsync.yaml	Passes Claude token to db-sync regression workflow.
.github/workflows/regression_reusable.yaml	Adds Claude failure analysis, uploads output, and includes it in failure emails.
.github/workflows/nightly.yaml	Passes Claude token to nightly regression workflow.
.github/workflows/nightly_pv11.yaml	Passes Claude token to PV11 nightly workflow.
.github/workflows/nightly_dbsync.yaml	Passes Claude token to db-sync nightly workflow.
.github/workflows/nightly_cli.yaml	Passes Claude token to CLI nightly workflow.

Comments suppressed due to low confidence (2)

.github/workflows/regression_reusable.yaml:161

• This step has the same implicit success() guard problem: once the test step has failed, the read step is skipped, so FAILURE_ANALYSIS is never populated for the mail body even if the analysis file exists. Include an explicit failure-tolerant status check here, matching the later upload/mail steps.

        if: steps.testing-step.outcome == 'failure'

.github/workflows/regression_reusable.yaml:168

• This writes AI-generated Markdown to $GITHUB_ENV with a fixed delimiter. If the analysis (or quoted test output) contains __EOF_ANALYSIS__ on its own line, the env command is terminated early and the remaining lines can be parsed as separate environment assignments or make the step fail. Use a delimiter that cannot appear in the content, or pass the file content to the mail step without re-encoding it through $GITHUB_ENV.

              echo 'FAILURE_ANALYSIS<<__EOF_ANALYSIS__'
              cat failure_analysis.md
              echo '__EOF_ANALYSIS__'
            } >> "$GITHUB_ENV"

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.