Auto-close stale Known Build Error issues with zero monthly hits

[MichaelSimons]

Summary

Adds automation to the existing stale workflow that closes Known Build Error (KBE) issues that have had 0 hits for over a month.

How it works

1. Step 1 (github-script): Scans open KBE issues, parses the hit count table from each issue body. If 1-month count is 0 → adds stale label + comment. If hits have resumed on a stale issue → removes stale label.

2. Step 2 (actions/stale): Closes KBE issues that have been stale for 7+ days.

3. Existing stale job: Now exempts KBE issues via exempt-issue-labels to avoid conflicts.

Opting out with keep-open

The standard escape hatches don't work here:

• Removing stale — the automation re-applies it on the next run as long as the 1-month hit count is 0.
• Removing Known Build Error — changes the issue's meaning; the issue is still a known build error, you just don't want it auto-closed.

Instead, add the keep-open label to any KBE issue that should remain open despite zero hits. The script skips issues with this label entirely.

Changes

• .github/workflows/stale.yml: Added stale-known-build-errors job, added env.KBE_LABEL shared variable, added exempt-issue-labels to existing stale job.

Testing

Tested on fork (MichaelSimons/sdk) with workflow_dispatch and sample issues. Verified:

• Zero-hit issues get stale label and comment
• Active-hit issues are skipped
• Resumed-hit stale issues get stale label removed
• Stale issues are closed by actions/stale

Affected Issues

Once merged, the following issues will be marked as stale and closed:

• Package reference alias tests are disabled (Package reference alias tests are disabled #39172)
• .NET Framework build job test failure (.NET Framework build job test failure #40160)
• ToolInstallCommandTests.WhenRunWithRoot is flaky (ToolInstallCommandTests.WhenRunWithRoot is flaky #42346)
• Build failure in VMR Vertical Build Windows_x64: Downloading vswhere - Process cannot access sdk.txt because it is being used by another
  process (Build failure in VMR Vertical Build Windows_x64: Downloading vswhere - Process cannot access sdk.txt because it is being used by another process #45809)
• CanIgnoreTemplateGroupsWithConstraints flaky test failing in CI (CanIgnoreTemplateGroupsWithConstraints flaky test failing in CI #46212)
• dotnet watch - Failure in DeleteSubfolder test with polling (dotnet watch - Failure in DeleteSubfolder test with polling #46454)
• Containers integration tests are flaky when the network fails to load the registry (Containers integration tests are flaky when the network fails to load the registry #49502)
• GenerateBundle Task fails writing to singlefilehost.exe (GenerateBundle Task fails writing to singlefilehost.exe #50784)
• Tests Failing in Template Engine: Failed to update search cache. (The SSL connection could not be established, see inner
  exception.) (Tests Failing in Template Engine: Failed to update search cache. (The SSL connection could not be established, see inner exception.) #51147)
• Containers Test Issue: Unknown command docker buildx (Containers Test Issue: Unknown command docker buildx #51161)

[Copilot]

Pull request overview

Adds GitHub Actions automation to identify and close stale Known Build Error (KBE) issues based on a “0 hits in the last month” signal, while preventing the repo’s general stale workflow from acting on KBEs.

Changes:

• Added a shared KBE_LABEL env var and exempted KBE issues from the existing actions/stale job.
• Added a stale-known-build-errors job that:
  • uses actions/github-script to apply/remove the stale label based on parsed monthly hit count
  • uses actions/stale to close KBE issues that have been stale for 7+ days

[MichaelSimons]

Merging ahead of validation since it doesn't cover the stale action.