Skip to content

[skills] Improve CI flake triage and targeted retries - #12926

Merged
simonrozsival merged 4 commits into
mainfrom
simonrozsival-ci-flake-retries
Oct 6, 2026
Merged

simonrozsival merged 4 commits into
mainfrom
simonrozsival-ci-flake-retries

Conversation

@simonrozsival

@simonrozsival simonrozsival commented Sep 25, 2026 •

Copy link
Copy Markdown
Member

Summary

  • classify red dotnet-android builds as likely PR regressions, known flaky tests, transient infrastructure failures, timeout/crashes, or unknowns with evidence and confidence
  • search for exact flaky-CI trackers and recommend structured issue updates or new issues
  • prefer confirmation-gated Azure stage retries with forceRetryAllJobs: false over full /azp run retries
  • emit versioned machine-readable analysis from ci_failures.cs, including previous attempts, paginated test results, cross-run Auto-Retry correlation, stage-safe retry plans, and fail-closed error handling

Validation

  • read-only analysis of Azure build 1611818 recovered its three original infrastructure failures and recognized the successful targeted retry attempts

simonrozsival and others added 2 commits September 25, 2026 22:38
Add evidence-based failure classification, flaky issue matching guidance, targeted Azure stage retry recommendations, machine-readable analysis, and offline fixtures.

Co-authored-by: Copilot App <[email protected]>
Remove the fixture suite and fixture-only analyzer input paths to keep the skill focused on live CI analysis.

Co-authored-by: Copilot App <[email protected]>
@simonrozsival
simonrozsival marked this pull request as ready for review October 5, 2026 15:31
Copilot AI balanced review requested due to automatic review settings October 5, 2026 15:31

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot review overview

🟡 Changes recommended

The analyzer can conflate or misclassify test failures and inadvertently exclude valid targeted retries.

Review effort: Balanced
Findings: 3 Medium severity

Open (3)
What changed in this PR

Improves CI failure classification, flaky-issue tracking, and confirmation-gated Azure stage retries.

Changes:

  • Adds structured failure analysis and retry plans.
  • Documents classification and issue-matching rules.
  • Adds targeted stage-retry guidance.
File Description
SKILL.md Defines the updated CI triage workflow.
scripts/​ci_failures.cs Implements classification and retry planning.
references/​flaky-issues.md Adds flaky-issue matching guidance.
references/​error-patterns.md Defines failure categories and confidence rules.
references/​azdo-queries.md Documents targeted Azure retries.

Comment thread .github/skills/ci-status/scripts/ci_failures.cs
Comment thread .github/skills/ci-status/scripts/ci_failures.cs Outdated
Comment thread .github/skills/ci-status/scripts/ci_failures.cs Outdated
…idence

Suppress synthetic test gates only with same-stage/attempt test evidence, retain assembly and normalized root error when collecting outcomes and fingerprints, and keep sibling-only passes out of retry recommendations. Preserve each failing stage and hydrate missing summary errors only from an exact result ID.

Co-authored-by: Copilot App <[email protected]>
@simonrozsival

Copy link
Copy Markdown
Member Author

@dalexsoto stamp

@dalexsoto dalexsoto left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Stamping as requested by @simonrozsival; no code review performed.

@simonrozsival
simonrozsival merged commit 2a20d99 into main Oct 6, 2026
44 checks passed
@simonrozsival
simonrozsival deleted the simonrozsival-ci-flake-retries branch October 6, 2026 06:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants