[CI] Build broken on main — 20260518.4

[joperezr]

🤖 This issue was filed automatically by a daily monitoring bot.

The most recent completed run of the internal microsoft-aspire pipeline on refs/heads/main failed.

• Build: 20260518.4 (id: 2977772)
• Result: Failed
• Started: 2026-05-18T12:09:59Z · Finished: 2026-05-18T13:22:14Z
• Commit: db90b8eb0094a5c01ecc59c23094a111dd40e5a3
• Internal link: https://dev.azure.com/dnceng/internal/_build/results?buildId=2977772

First-level analysis

Only the Publish Assets job failed; all build and packaging jobs (win_x64, linux_x64, osx_arm64, Windows, SourceIndexStage1, etc.) succeeded. The failure is in the Maestro/BAR promotion step — PromoteBuild returned a non-zero exit code while trying to publish the produced BAR build to its default channels, ending with PowerShell exited with code '1'. Likely category: publishing/infra (Maestro promote), not a product build or test regression.

---

cc Ankit Jain (@radical) Jose Perez Rodriguez (@joperezr)

[joperezr]

🤖 Aspire triage pass — 2026-05-18 16:03 UTC

Area: area-engineering-systems
Kind: bug (CI / publishing infra failure)
Repro: none — automated build report; details in issue body
Affected version(s): main branch, internal pipeline build 20260518.4
Severity: S2 — Publish Assets job fails (Maestro PromoteBuild exits with code 1); build/test stages succeeded so no product regression, but BAR promotion is blocked for this build.

Urgency: ⚪ backlog — no milestone
Reason: CI / publishing-infra failure, not a product regression in shipping bits, so it doesn't fit the servicing (13.3.x) or vNext (13.4) tracks. Typically handled promptly outside the milestone process.

Suggested owner(s): Ankit Jain (@radical) (score ~68), backup Mitch Denny (@mitchdenny) (score ~3)
Evidence:

• Ankit Jain (@radical) merged a large share of recent area-engineering-systems PRs in the last 90 days (e.g. Avoid empty test summary artifact URL #16478, Add LocalArchive + DotnetTool CLI install modes for E2E tests #16454, feat(ci): add test failure retry support to auto-rerun workflow #16446, Publish and sign CLI install scripts #15127, [CI] Fix randomly failing tests on azdo due to docker+testcontainers/ryuk issues #15111, Automatically rerun transient CI failures #15110, [CI] Update to newer linux image on internal pipeline #15055, Reduce the number of PR validation jobs #15031, Reduce PR validation time by decoupling CLI archive dependencies #15023, [release/13.2][CI] Use large runner for Aspire.Hosting.Tests #15016, Add 3rd-party signing entries for Sigstore, Tuf, and NSec.Cryptography #14915, [release/13.2][CI] Unified test splitting infrastructure for CI parallelization #14858, Improve fix-flaky-test skill and reproduce workflow #14710, [release/13.2] Add Homebrew and WinGet packaging for Aspire CLI #14690)
• Ankit Jain (@radical) was the resolver/assignee on many area-engineering-systems issues closed as completed in the last 180 days (e.g. Aspire.Cli nuget package isn't being published #16563, [CI] AzDO PR builds fail on WinGet/Homebrew installer URL validation #15818, [CI] release/13.2 internal builds are failing - need the build image updated #15655, [CI] Internal builds are failing to publish #15543, [CI] Disable EndToEnd tests on macOS for Outerloop #15500, [CI] Winget manifest building step failing on internal pipeline #15498, [CI] Fix EndToEnd tests running on Outerloop for main. #15325, [CI] .net 10 tests broken on helix #13297, Investigate test failures on the weekly azdo build for week of Jun 16 #9911, [CI] Quarantined and flaky tests tracker #9583, [tests] Playground: Ensure that these are running against the nugets #8793)
• Mitch Denny (@mitchdenny) merged Remove test-scenarios infrastructure #14639 in area-engineering-systems
• (Issue author Jose Perez Rodriguez (@joperezr) excluded from owner candidates per workflow rule)

Confidence: high (area + dominant owner)
Needs human?: yes — infra/publishing failures don't fit the servicing/vNext milestone tracks; routing decision left to a human.

Dry-run mode — no labels or milestone applied. Reply /triage apply to promote, or edit and reapply.

[joperezr]

🤖 Daily monitor: build still failing.

• Build: 20260522.1 (id: 2981613) on main
• Result: Failed
• Commit: a4c4d9e4f4500bbc881971223d69d1ad4c3dd099
• Internal link: https://dev.azure.com/dnceng/internal/_build/results?buildId=2981613

First-level analysis

Failing job: Windows. Root cause: yarn install exited with code 1 in extension/Extension.proj (line 39, MSB3073). Downstream symptoms in the same job — "Manifest generation failed" and the missing publish path artifacts/packages/Release/vscode — are caused by that upstream yarn failure. Likely category: VS Code extension build error (yarn dependency install).

[joperezr]

🤖 Daily monitor: build still failing.

• Build: 20260523.1 (id: 2982459) on main
• Result: Failed
• Commit: 96e8734309f8f325c1e33a7ec5a422723eb5cfff
• Internal link: https://dev.azure.com/dnceng/internal/_build/results?buildId=2982459

First-level analysis

The Windows job failed: yarn install exited with code 1 while building extension/Extension.proj (the VS Code extension). The subsequent Manifest generation failed and Not found PathtoPublish: …\artifacts\packages\Release\vscode errors are downstream symptoms of the extension build not producing its output. Likely category: build error in the extension's yarn install step.

[joperezr]

🤖 Daily monitor: build still failing.

• Build: 20260524.1 (id: 2983037) on main
• Result: Failed
• Commit: 91465876608dbba43dac5ef662e6e3f30192f8b1
• Internal link: https://dev.azure.com/dnceng/internal/_build/results?buildId=2983037

First-level analysis

The osx_arm64 job failed: a Bash step exited with code 1, which left the artifacts/packages/ output directory missing. That subsequently caused the manifest-generation and "Publish build artifacts" steps in the same job to fail because PathtoPublish was not found. Other jobs only show cosmetic warnings (Component Governance, Container Security, CodeQL LD_PRELOAD notices) and did not fail. Root cause is the failing build/script step on the osx_arm64 leg, not the post-build SDL/publish symptoms.

[joperezr]

🤖 Daily monitor: build still failing.

• Build: 20260525.5 (id: 2983765) on main
• Result: Failed
• Commit: e8d57670bf7abc64a7d085201fadbef7318408e3
• Internal link: https://dev.azure.com/dnceng/internal/_build/results?buildId=2983765

First-level analysis

The Windows job failed: yarn install in extension/Extension.proj exited with code 1 (MSB3073), which prevented the VS Code extension packages from being staged and then caused the subsequent artifact publish to fail (missing artifacts/packages/Release/vscode). The TSA errors on the Linux legs look like SDL-infra noise (TSA onboarding PreconditionFailed / timeout against the DevDiv TFS instance) and are likely secondary.

[joperezr]

🤖 Daily monitor: build still failing.

• Build: 20260527.6 (id: 2985368) on main
• Result: Failed
• Commit: e610574dd3b8442db0ee35a5a95e7abcd3888efc
• Internal link: https://dev.azure.com/dnceng/internal/_build/results?buildId=2985368

First-level analysis

Failing jobs: SourceIndexStage1 and Windows. Root cause is a NuGet restore failure (NU1101) for the generated AzureFunctionsWithDts.Functions WorkerExtensions.csproj — three packages cannot be found in the configured feeds: Microsoft.DurableTask.Extensions.AzureBlobPayloads, Microsoft.DurableTask.AzureManagedBackend.Protobuf, Microsoft.DurableTask.Extensions.AzureBlobPayloads.AzureManaged. PackageSourceMapping excludes dotnet-libraries, so these need to either be mirrored into an allowed feed or added to the source-mapping for an existing one. (The CodeQL LD_PRELOAD red lines on linux jobs are benign noise.)

[joperezr]

🤖 Daily monitor: build still failing.

• Build: 20260528.2 (id: 2986215) on main
• Result: Failed
• Commit: 7f69b671df4f70a2b0827fb4ec143a9254e9917c
• Internal link: https://dev.azure.com/dnceng/internal/_build/results?buildId=2986215

[radical]

Stale.