Add Azure DevOps flaky-history annotations and quarantine awareness

[Evangelink]

Part 2 of the brainstorm in #5951 — adds opt-in flaky-test history annotations and quarantine awareness to Microsoft.Testing.Extensions.AzureDevOpsReport. Decorates AzDO log issues with historical flake context and lets known-noisy failures be downgraded so PR gates aren't blocked.

One of three PRs derived from the #5951 brainstorm. The others:

• Add live Azure DevOps test result publishing #8297 — Live test-result publishing
• Auto-upload test artifacts and build tags to Azure DevOps #8295 — Auto-upload test artifacts and build tags (pipeline-level)

See issue comment for the broader plan.

Why

Today every failure in an AzDO log looks equally bad. There's no way for the build to say "this test failed 4/20 times in the last 14 days — known noise" vs "this test had zero failures in 14 days and just broke — likely regression". Teams that have moved to MTP currently mark failures as warnings manually with a separate task or live with red PR checks for known-flaky tests.

What

Three new opt-in CLI options on Microsoft.Testing.Extensions.AzureDevOpsReport:

Option	Type	Purpose
--report-azdo-flaky-history <days>	int (1–90)	Query AzDO REST history for the last N days and annotate failures with [flaky: failed K/N in last Md] or [REGRESSION] (only when ≥5 prior samples).
--report-azdo-quarantine-file <path>	string	Path to a text file (one FQN/glob per line, # comments allowed) listing tests considered quarantined. Their failures are demoted to warning and tagged [quarantined]; emits ##vso[build.addbuildtag]has-quarantined-test-failure exactly once.
--report-azdo-demote-known-flaky	zero-arity	Together with --report-azdo-flaky-history, auto-demote failures whose flake-rate ≥25% in the window to warning. Default OFF (annotate-only). Requires --report-azdo-flaky-history.

All three are opt-in; missing AzDO env vars (SYSTEM_ACCESSTOKEN/SYSTEM_COLLECTIONURI/SYSTEM_TEAMPROJECT/BUILD_DEFINITIONID) → log warning and no-op.

How it works

• Auth — Authorization: Basic base64(":<SYSTEM_ACCESSTOKEN>") (no Microsoft.TeamFoundationServer.Client dependency; just HttpClient + a source-generated JsonSerializerContext so it's AOT-safe).
• History query — GET {project}/_apis/test/Runs?definitions={pipelineDefinitionId}&minLastUpdatedDate=…&maxLastUpdatedDate=…&automated=true&$top=200 paginated with $skip up to MaxRunsToInspect = 200. Per run, GET …/results?api-version=7.1&outcomes=Failed,Passed paged with continuation token. Aggregated into Dictionary<automatedTestName, FlakyStats>.
• Bounded session start — history load has a wall-clock budget (default 30 s). If exceeded, log info and degrade to empty stats; tests start immediately.
• Resilience — REST calls retry on transient errors (3 attempts, exponential backoff, 429 honors Retry-After). Response bodies truncated to 500 chars in error messages. All callbacks catch everything except OperationCanceledException; history/quarantine failures never fail the test run.
• Quarantine file — line-based (# comments), case-sensitive ordinal matching against FQN, glob patterns (*, ?) compiled to a single alternation regex. Capped at 10 000 patterns / 4 KB per pattern with a logged warning.
• Build tag — ##vso[build.addbuildtag]has-quarantined-test-failure guarded by Interlocked.Exchange so concurrent failure events emit it exactly once.

Highlights from the expert-reviewer round

Implementation went through one full round of expert-reviewer. Critical issues addressed:

• C1 (feature was broken): the runs query was using buildIds=<pipelineDefinitionId> (the wrong AzDO parameter — buildIds filters by individual build run id, not pipeline). Switched to definitions=<pipelineDefinitionId> so the query actually returns data. Unit test now asserts the URL contains definitions=.
• C2: history load was synchronous & serial — up to 25 000 sequential HTTP requests blocking session start. Bounded with a 30 s wall-clock budget + cancellation; degrades to empty stats on exceed.
• C3: $top=501 was silently capped server-side. Now properly pages runs via $skip until MaxRunsToInspect is reached.
• C4: [REGRESSION] previously fired on a single prior sample, generating false positives everywhere. Now requires TotalCount >= 5 (configurable via MinSamplesForRegressionAnnotation).

Major items also addressed: User-Agent/Accept headers added (prevents AzDO WAF-side 403s), source-generated JsonSerializerContext (AOT-safe), error-body truncation, inner-exception preservation on retry exhaustion, quarantine pattern caps, ordinal-case-sensitive matching, IAzureDevOpsHistoryService interface + proper DI registration, guard-clause CLI validation, acceptance tests for invalid-value error paths.

Tests

546 unit tests pass. New coverage:

• AzureDevOpsHistoryServiceTests.cs — history aggregation, paging, time budget, regression boundary (4 vs 5 samples), quarantine tag-emitted-once.
• AzureDevOpsHistoryClientTests.cs — URL composition (definitions= parameter), User-Agent/Accept headers, 429 retry honoring Retry-After.
• AzureDevOpsCommandLineProviderTests.cs — cross-product validation (e.g. --demote-known-flaky without --flaky-history, --quarantine-file without --report-azdo).
• AzureDevOpsCommandLineTests.cs — acceptance-style coverage.

HelpInfoAllExtensionsTests expectations updated for the new options (both --help and --info blocks, alphabetical order preserved).

Build status (local)

• .\.dotnet\dotnet.exe build src\Platform\Microsoft.Testing.Extensions.AzureDevOpsReport\Microsoft.Testing.Extensions.AzureDevOpsReport.csproj -c Debug — 0 warnings, 0 errors.
• .\.dotnet\dotnet.exe test test\UnitTests\Microsoft.Testing.Extensions.UnitTests\Microsoft.Testing.Extensions.UnitTests.csproj — 546/546 passed.
• .\build.cmd -pack — 0 warnings, 0 errors.

Out of scope (deliberate)

• Posting a summary list of quarantined-skipped tests to the terminal reporter (show summary of retried and flaky tests #5362 territory).
• Per-suite or per-area aggregation of flake stats.

Checklist

• Critical & Major review findings addressed
• Localized via resx + xlf (regenerated with /t:UpdateXlf, not hand-edited)
• Help/info acceptance test expectations updated
• No new public API (besides one internal IAzureDevOpsHistoryService)
• .\build.cmd green (0 warnings, 0 errors)
• PR feedback addressed

Refs #5951

[Copilot]

Pull request overview

Adds opt-in Azure DevOps flaky-history annotations and quarantine awareness to the Microsoft.Testing.Extensions.AzureDevOpsReport extension. Failures can now be annotated with historical flake context, demoted from error to warning when known-flaky, and demoted with a [quarantined] tag when listed in a quarantine file (which also emits a one-shot ##vso[build.addbuildtag]has-quarantined-test-failure).

Changes:

• New CLI options --report-azdo-flaky-history, --report-azdo-quarantine-file, --report-azdo-demote-known-flaky with cross-option validation in AzureDevOpsCommandLineProvider.
• New AzureDevOpsHistoryService + AzureDevOpsHistoryClient (AOT-safe JsonSerializerContext) that queries the AzDO REST Runs/Results APIs under a 30 s wall-clock budget, with retries, 429 Retry-After honoring, paging caps, and a regression-annotation min-sample threshold.
• AzureDevOpsReporter now annotates errors with [flaky: failed K/N in last Md] / [REGRESSION] / [quarantined], and demotes severity per the quarantine file and known-flaky rule.

Show a summary per file

File	Description
src/.../AzureDevOpsCommandLineOptions.cs	Adds 3 new option name constants.
src/.../AzureDevOpsCommandLineProvider.cs	Registers new options and adds cross-option validation.
src/.../AzureDevOpsExtensions.cs	Wires AzureDevOpsHistoryService as data consumer + session lifetime handler.
src/.../AzureDevOpsHistoryClient.cs	New REST client (auth, paging, retries, AOT JSON).
src/.../AzureDevOpsHistoryClientJsonContext.cs	Source-generated JSON context for DTOs.
src/.../AzureDevOpsHistoryService.cs	Loads/aggregates flaky stats with a bounded budget; exposes TryGetStats/IsLikelyFlaky.
src/.../AzureDevOpsReporter.cs	Adds annotation suffix building, severity demotion, one-shot quarantine build tag.
src/.../FlakyStats.cs	Struct holding pass/fail counts and failure rate.
src/.../IAzureDevOpsHistoryService.cs	Internal abstraction over the history service.
src/.../QuarantineFile.cs	Parses quarantine file (globs, # comments, caps) into regex matchers.
src/.../Microsoft.Testing.Extensions.AzureDevOpsReport.csproj	Adds System.Text.Json dependency and DynamicProxyGenAssembly2 IVT for Moq.
Directory.Packages.props	Pins System.Text.Json version.
src/.../Resources/AzureDevOpsResources.resx	New strings for options, warnings, and annotation templates; fixes prior Eanble/AzureDev Ops typos.
src/.../Resources/xlf/*.xlf (12 locales)	Regenerated XLFs for new strings; Description/OptionDescription flipped to needs-review-translation after the English typo fix.
test/.../AzureDevOpsHistoryClientTests.cs	Asserts URL composition (definitions=), headers, and run-paging behavior.
test/.../AzureDevOpsHistoryServiceTests.cs	Covers aggregation, paging, time-budget timeout, regression threshold, demote, quarantine tag-once.
test/.../AzureDevOpsCommandLineProviderTests.cs	Validates cross-option error messages.
test/.../AzureDevOpsCommandLineTests.cs	Acceptance-style test for invalid CLI argument errors.
test/.../HelpInfoAllExtensionsTests.cs	Updates --help / --info expectations for new options.

Copilot's findings

• Files reviewed: 31/31 changed files
• Comments generated: 0