Port conditional test selection changes to main#15742
Draft
radical wants to merge 23 commits into
Draft
Conversation
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
radical
added
the
NO-MERGE
Contributor
|
🚀 Dogfood this PR with:
curl -fsSL https://raw.githubusercontent.com/microsoft/aspire/main/eng/scripts/get-aspire-cli-pr.sh | bash -s -- 15742Or
iex "& { $(irm https://raw.githubusercontent.com/microsoft/aspire/main/eng/scripts/get-aspire-cli-pr.ps1) } 15742" |
…-selection-to-main # Conflicts: # .github/workflows/tests.yml # tests/Infrastructure.Tests/Infrastructure.Tests.csproj
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Restore PR coverage for CLI, polyglot, and extension tests - Prevent TestSelector fixture projects from compiling during repo builds - Remove duplicate unused Aspire.TestSelector tool copy Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…-selection-to-main
- Remove unconditional PR execution from standalone conditional jobs - Add workflow regression coverage for category guards Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
adamint
reviewed
Jun 4, 2026
adamint
left a comment
adamint
left a comment
Member
There was a problem hiding this comment.
High-confidence findings from review:
-
DotNetAffectedRunner.RunProcessAsyncreads stdout to completion before it starts reading stderr. Ifdotnet affected --verbosefills stderr while stdout is quiet, the child blocks on the stderr pipe and the parent waits forever on stdout. Start bothReadToEndAsync()tasks before awaiting, like theGitHelpercode in this PR already does. -
This port downgrades existing
actions/upload-artifactcall sites from the v7.0.0 SHA onmainback to v4.6.1, leavingtests.ymlwith a v4/v7 mix. Please re-pin the touched upload steps to the v7.0.0 SHA used bymain. -
The
Show test selection resultsstep printsACTIVE_SELECTION_REASON, but that variable is not in the stepenv:block. The diagnostic group will always showselection_reason=unknown; addACTIVE_SELECTION_REASON: ${{ steps.detect_active_raw.outputs.selection_reason }}.
Resolve conflicts: - extension/Extension.proj: take origin/main version (restores the COREPACK_HOME property block; the prior HEAD had a placeholder '<x/>' from an earlier bad conflict resolution). - tests/Infrastructure.Tests/Infrastructure.Tests.csproj: keep both TestSelector and TypeScriptApiCompat ProjectReferences. - .github/workflows/tests.yml: keep the conditional-selection gating pattern (run_all / run_extension / run_polyglot, matrix-non-empty) and add the new jobs introduced on main to the 'Fail if any dependency failed' check: build_cli_archive_macos_x64, prepare_winget_installer_artifacts, prepare_homebrew_installer_artifacts, build_cli_e2e_image (gated by tests_matrix_requires_cli_archive), extension_e2e_tests (gated by extension_e2e_changes), and typescript_api_compat. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The previous merge of origin/main produced two 'if:' keys on the polyglot_validation job: one from main gating on event/ref, and one from this PR gating on the conditional-selection outputs. GitHub's workflow parser rejected the file entirely (run failed with no jobs created), and actionlint flagged 'key "if" is duplicated'. Combine both into a single guard: the job runs when (a) the trigger is a PR or a push to main AND (b) the conditional selector requested either run_all or run_polyglot. Non-PR events still fall through because setup_for_tests pre-sets every run_* output to 'true' for them. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The Stabilization Check job invokes Arcade's pack pipeline, which enumerates ProjectToBuild from `tests/**/*.csproj` (eng/Build.props line 69). That enumeration was picking up the static-analyzer fixture projects under tests/Infrastructure.Tests/TestSelector/TestFixtures/. Those .csproj files are deliberate test fixtures consumed as text by TestSelector's analyzer tests (IgnorePathFilterTests, NuGetDependentTestDetectorTests, etc.) — they're never meant to be built. Their adjacent Directory.Build.targets neuters Build/Rebuild/ Restore/Pack/Test as no-ops, which is enough for `build.sh -build` but not for `build.sh -pack`: the NuGet Pack target still attempts to read the assembly output (e.g. ConditionalPackableProject.dll) and fails with NU5026 because Build was a no-op. Filter the fixtures out of the test-project enumeration so Arcade never tries to build or pack them. The TestSelector tool's own testProjectPatterns config already excludes `tests/**/TestFixtures/**` from its discovery, so this brings the build infrastructure in line with the same convention. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Replayed the 161 PRs merged to main between 2026-05-23 and 2026-06-06
through the audit selector. v1 (the prior file) resolved 89/161 PRs to
'selective' and dropped 26/161 to 'fallback_unmatched' (any unmatched
file forces RunAll). After this change all 161 PRs classify cleanly:
18 skip, 32 trigger_all, 111 selective, 0 fallback.
Additions:
* ignorePaths — files that never need a test gate but were forcing
fallback today: .vscode/**, .mcp.json, localhive.{sh,ps1},
.github/{extensions,policies,aw}/**, .agents/** (CLI agent skill
content, consumed only by humans/agents), eng/scripts/debug-*.{sh,ps1}
(developer-loop helpers), eng/scripts/cli-starter-validation*.ps1
(consumed by the always-on cli_starter_validation_windows job, which
isn't conditional-selector-gated).
* sourceToTestMappings — scripts and infra files that have a specific
test home but weren't wired up: eng/Publishing.props,
eng/Signing.props, eng/scripts/{pack,stage,verify,validate}-cli-*
and tools/TypeScriptApiCompat/** all map to Infrastructure.Tests
(its Pipelines/ and PowerShellScripts/ folders test these scripts).
eng/scripts/get-aspire-cli{,-pr}.{sh,ps1}, eng/scripts/verify-cli-archive.ps1,
eng/homebrew/** and eng/winget/** map to Aspire.Acquisition.Tests
(the script-suite that exercises them). eng/scripts/update-aspire-
skills-bundle.ps1 + verify-aspire-skills-bundle.ps1 map to
Aspire.Cli.Tests (AspireSkillsBundleTests). Two self-mappings for
tests directories without a 'tests/Aspire.*.Tests/**' shape:
tests/Infrastructure.Tests/** and
tests/Aspire.Hosting.CodeGeneration.TypeScript.JsTests/**.
* polyglot.triggerPaths — expanded beyond the workflow file itself to
cover the polyglot fixture tree (tests/PolyglotAppHosts/**) and the
per-language Aspire.Hosting.{JavaScript,TypeScript,Python,Go,Java,Rust}
+ CodeGeneration* projects. Without these, polyglot-impacting PRs
(#17545, #17499, #17419, #17400, #17382) all fell into RunAll.
Local verification: 251 TestSelector/PowerShellScripts/ConditionalSelection
tests still pass. The CI-trigger-pattern coverage test (which loads this
file from disk) is unchanged because we only added to ignorePaths and
sourceToTestMappings — no existing entries were removed or narrowed.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Contributor
|
Re-running the failed jobs in the CI workflow for this pull request because 1 job was identified as retry-safe transient failures in the CI run attempt.
|
CI on 5d9adab failed in 'Fail if any dependency failed' because extension_e2e_tests showed up as 'skipped' while extension_e2e_changes had output run_extension_e2e=true. Root cause: extension_e2e_tests lists extension_tests_win in its needs, and the conditional selector correctly skipped extension_tests_win on this PR (run_extension/run_all both false). GitHub Actions then auto-skips any dependent whose 'if' doesn't include a status function — so extension_e2e_tests was forced to skip transitively, even though run_extension_e2e == 'true' and the rest of its needs were green. extension_e2e_changes has its own broader path-detection (extension/, src/Aspire.Cli/, src/Aspire.Hosting*/, src/Aspire.Dashboard*/, tests/Aspire.Cli*/, get-aspire-cli scripts, certain workflow files) so it can fire when 'extension' category is NOT triggered — e.g. a pure CLI or hosting change should run extension E2E even though no extension code changed. The fix is to let extension_e2e_tests run when its own path-detection says yes, regardless of extension_tests_win's result. Use `!cancelled() && !failure() && run_extension_e2e == 'true'`. The status-function prefix disables the implicit-skip behavior; the failure/cancel checks still propagate real failures from any needed job (build_packages, build_cli_archive_*, extension_bootstrap_linux, extension_tests_win) but treat 'skipped' as acceptable. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
CI on 187550c (the previous attempt to fix this) made extension_e2e_tests survive extension_tests_win being skipped, but the E2E job depends on the VSIX artifact produced inside extension_tests_win's Package VSIX step. With extension_tests_win skipped, every E2E shard failed: Unable to download artifact(s): Artifact not found for name: aspire-extension The two jobs cannot be decoupled — the unit-test job is also the only producer of the VSIX. Run extension_tests_win whenever EITHER the conditional selector triggers the extension category OR extension_e2e_changes (a broader path-detection covering CLI / hosting / dashboard) wants E2E. The 'Fail if any dependency failed' gate is updated to match: extension_tests_win is only required when one of those three signals is true. Revert extension_e2e_tests's if back to the simple form. Now that extension_tests_win is guaranteed to run whenever run_extension_e2e == 'true', the implicit GitHub-Actions auto-skip on needs failure/skip is the desired behavior again — there's no legitimate case where run_extension_e2e is true and extension_tests_win is skipped. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…-selection-to-main
…ules
When several source patterns all run the same test project, the rules
file used to need one mapping entry per source — a lot of repetition
and visual noise. Example before this change (excerpt of audit.json):
{ "source": "eng/Publishing.props", "test": "tests/Infrastructure.Tests/Infrastructure.Tests.csproj" },
{ "source": "eng/Signing.props", "test": "tests/Infrastructure.Tests/Infrastructure.Tests.csproj" },
{ "source": "eng/scripts/pack-cli-npm-package.ps1", "test": "tests/Infrastructure.Tests/Infrastructure.Tests.csproj" },
{ "source": "eng/scripts/verify-cli-npm-package.ps1","test": "tests/Infrastructure.Tests/Infrastructure.Tests.csproj" },
... 4 more rows ...
Make `source` accept either a string (existing form) or an array, and
collapse those rows into one. Both shapes resolve identically; the
single-string form is just sugar for a one-element array.
Implementation:
* `SourceToTestMapping.Source` is now `List<string>`, deserialized
through a `StringOrStringArrayConverter` that accepts both shapes.
* `ProjectMappingResolver.CompiledMapping` keeps one compiled regex
per source pattern; `TryMatch` / `TryMatchWithDetails` iterate them
and short-circuit on the first hit. `exclude` is applied uniformly
across every source pattern in the mapping (one matcher, checked
before the per-pattern regex loop). The detail-recording path
reports the *specific* source pattern that matched, not the whole
list, so `selector-active.json` mapping diagnostics stay precise.
* Schema (`test-selection-rules.schema.json`) updated to `oneOf`
string / non-empty array. Existing tooling that points at the
schema gets the new shape with no other changes.
* README adds a short example of the array form alongside the existing
`{name}` example.
Audit JSON consolidation: 29 mapping entries → 6 (each test project
now has a single entry). Replayed against 161 PRs merged to main
between 2026-05-23 and 2026-06-06: zero per-PR outcome drift versus
the pre-consolidation rules. The CI-trigger-pattern coverage test
(loads the real audit JSON from disk) still passes.
Tests added:
* `LoadFromJson_SourceAsArray_CollectsAllPatterns` — verifies the
new shape deserializes to the expected list, in order.
* `LoadFromJson_SourceAsString_NormalizesToSingleElementList` —
pins the back-compat shape.
* `ResolveTestProjects_ArraySource_AllPatternsMapToSameProject`
(theory, 4 inputs) — equivalence with the prior N-mapping form.
* `ResolveTestProjects_ArraySource_ExcludeAppliesToAllPatterns` —
exclude is shared across patterns within a mapping.
Existing tests that assigned `Source = "..."` now use the collection
initializer form `Source = ["..."]`; behavior unchanged.
Verification: 258 TestSelector / PowerShellScripts / ConditionalSelection
tests pass (up from 251 by the four new tests, factoring out theory
expansion).
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ilename globs match nested paths Three automated PRs in the last 14d fired more tests than they needed to: #17672, #17263 [Automated] Update ATS API Surface Area #17534 Move repository skills to .agents They each touched only api/*.txt or only README/skill markdown, but were firing trigger_all or selective:integrations — burning the full ~25min CI critical path. Audit-replay over 161 merged PRs found them, plus a latent C# glob bug where bare-filename patterns silently ignored nested matches. Root causes — two independent bugs: 1. The category-trigger rescue logic in RescueCategoryTriggerFiles built a synthetic union of all category triggerPaths without consulting per-category excludePaths. An ignored file that textually matched some category's glob but was excluded from that category got rescued back to active and then hit fallback_unmatched (worse than staying ignored). For src/Aspire.Hosting.Foundry/api/*.txt this meant ATS-only PRs fired integrations even after `**/api/*.txt` was added to ignorePaths. 2. The four glob analyzers (CriticalFileDetector, IgnorePathFilter, CategoryMapper.CompiledCategory, ProjectMappingResolver.CompiledMapping) handed user-facing patterns directly to FileSystemGlobbing.Matcher. The Matcher anchors bare-filename patterns at the repo root, so `Directory.Build.props` matched only the root file, not `src/Directory.Build.props` or `tests/Directory.Build.props`. The Python audit-replay evaluator (eval_rules.py) already documented and applied a "prepend **/ to bare-filename patterns" rule; the C# analyzers did not, so the two evaluators silently disagreed on ~5 patterns across ignorePaths, triggerAllPaths, and sourceToTestMappings. The fix: - Rescue now passes config.Categories directly to CategoryMapper so CompiledCategory.Matches honors per-category excludes. A file is rescued only when at least one category would actually fire on it. - New PatternNormalization.NormalizeGlob prepends `**/` to any pattern without a path separator. Every glob entry point applies it: the four analyzers above plus ProjectMappingResolver's regex compiler. - Rules: integrations.excludePaths gains `tests/Aspire.Acquisition.Tests/**`, `tests/Infrastructure.Tests/**`, `**/*.md`, `**/api/*.txt`. The same `**/api/*.txt` exclude is added to every category so an ignored ATS file can't be rescued back by any category. ignorePaths gains `**/api/*.txt`. Acquisition mapping's source list gains the missing self-mapping `tests/Aspire.Acquisition.Tests/**` (Templates and Infrastructure mappings already had this; Acquisition was an oversight exposed only after the new exclude was added). Verification: - Audit replay over 161 merged PRs: 4 outcomes change (#17263, #17534, #17549, #17672 all move to `skip`); zero regressions; zero fallback_unmatched. - New AuditFixtureTests xUnit [Theory] replays 28 hand-validated PRs against the live audit rules. Each row is a separate test, so any future rule edit that changes a row's outcome shows up as a visible CI failure. Coverage includes templates (#16447), CLI native build (#17567), extension multi-category (#17881/17698/17772), Hosting-core trigger_all (#17879), polyglot (#17948), and the regression canaries for previous fallback_unmatched cases. - Per-component regression tests pin both bugs: two new tests in EndToEndEvaluationTests for rescue+excludes; two more for bare-filename matching at nested paths. - Three pre-existing analyzer tests had asserted the buggy bare-filename behavior as expected (e.g. `*.md` not matching `docs/guide.md`). Updated with comments explaining the user-intent rule. - Full TestSelector namespace: 290 tests, all pass. No collateral damage on the wider Infrastructure.Tests suite (5 pre-existing baseline failures unchanged). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The previous commit added a `RestrictedProjectFilter.Apply` call in `TestEvaluator.FilterAndCombineTestProjects` but never `git add`ed the new `RestrictedProjectFilter.cs` or its test. A clean checkout of HEAD therefore fails to compile — the type is unresolved. The build only worked locally because the files existed untracked in the working tree. Add the two files so HEAD builds on its own and the restricted-test-project rule is covered by its unit tests. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
When the integrations category matched a changed file but `dotnet-affected` attributed the change to no project at all, the selector produced an empty affected-test-projects set and selected nothing. The common trigger is a changed file that is not an MSBuild input of any project (e.g. a file under a project with `EnableDefaultItems=false`, or outside the default item globs): `dotnet-affected`'s reverse-dependency walk returns nothing, so a known change to a covered source area runs zero integration tests. Add a conservative guard (`CheckMatchedButZeroProjects`, evaluator step 9b): if the `integrations` category fired but both the affected-project set and the resolved test-project set are empty, run all tests instead of selecting none. Scope is deliberate: - Gated on the `integrations` category — the boolean-driven categories (cli_e2e, extension, polyglot) gate their own jobs via `run_<category>` regardless of the affected matrix, so an empty matrix is expected for them and must not force a full run. - Requires zero affected projects — when `dotnet-affected` did see the change but it resolved only to a restricted opt-out project or a source project with no dependent test, that is a deliberate outcome, not a blind spot, so the guard stays out of the way. The guard is unreachable in the current active (templates-pilot) config: its `ignorePaths: ['**']` plus sourceToTestMappings covering all template trigger paths route every active file through the mappings-only shortcut before the guard runs. It applies to the audit (observational) selector and to any future promotion of the audit rules to the gating config. Also bump `dotnet-affected` 6.2.0-preview-1 -> 6.2.0 (now released). Tests: 4 cases in EndToEndEvaluationTests pinning the fire / no-fire conditions. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Changes under four shared test-helper / app-host fixture directories matched no audit rule, so the selector fell through to its conservative RunAll fallback (fallback_unmatched) — every change to a test utility re-ran the whole suite: - tests/Aspire.TestUtilities - tests/Aspire.Components.Common.TestUtilities - tests/testproject - tests/TestingAppHost1 Each is referenced by its dependents via ProjectReference (e.g. Aspire.Hosting.Tests, Aspire.Hosting.Testing.Tests, Aspire.Hosting.Docker.Tests), so dotnet-affected can fan a change out to exactly the affected test projects. Add the four directories to the integrations category triggerPaths so a change is "accounted for" and routed through dotnet-affected instead of forcing a full run. This is safe because every tracked file under these directories is an MSBuild input (default Compile/None/Content globs; none set EnableDefaultItems=false), so dotnet-affected sees the change. A future non-input file would resolve to zero projects and is caught by the matched-but-zero guard, which runs all tests. Tests: AuditRules_SharedTestHelperChange_RoutedToIntegrations asserts each directory now resolves to a selective integrations run rather than fallback_unmatched. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The active conditional test-selection config set `ignorePaths: ["**"]`,
which routed every non-template change into the "ignored" bucket. A PR
that touched a template file alongside any other source was narrowed to
Aspire.Templates.Tests only, and the rest of the suite was silently
skipped while CI reported green:
--changed-files "tests/Aspire.Templates.Tests/X.cs,src/Aspire.Hosting.Redis/RedisBuilder.cs"
=> runAllTests=false, affected=[Aspire.Templates.Tests], ignored=[RedisBuilder.cs]
The same catch-all ignore meant the selector never emitted `run_polyglot`,
so the `polyglot_validation` gate (`run_all || run_polyglot`) was false on
every PR and polyglot validation never ran.
Root cause: `ignorePaths` is a "never trigger a full run" signal, but the
`["**"]` entry was treated as "safe to skip." Ignored files bypass the
unmatched-files safety net that otherwise forces a full run.
Fix: drop the catch-all (`ignorePaths: []`). Non-template changes are now
unmatched and force a full run; pure-template PRs still narrow to
Aspire.Templates.Tests. Polyglot is covered for free because any
non-template change sets `run_all=true`, satisfying its gate.
Also in this change:
- ProjectMappingResolver: a normalized bare-filename pattern (`**/X`)
compiled to `^.*/X$`, which cannot match the file at the repo root.
Translate a leading `**/` to `(?:.*/)?` so root and nested both match.
- Audit comparison steps run pwsh with `$ErrorActionPreference='Stop'`
but had no guard; a malformed or locked TRX could fail an otherwise
green PR. Mark both observational steps `continue-on-error: true`.
- Restore four `upload-artifact` pins from v4.6.1 back to the repo
standard v7.0.0.
- Remove the superseded hybrid/comparison design artifacts; keep the
unified target-design docs that live specs still reference.
Tests: update two active-config tests that had asserted the unsafe
behavior, add a regression guard that fails if a mixed PR stops running
the full suite, and add a bare-filename root-match test.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Three review-driven fixes to the conditional test selector. ## Force runtime-coupled tests via category testProjects A category's tests only run under selective scope if dotnet-affected put them in `affected_test_projects`. `filter-test-matrix-by-scope.ps1` drops every matrix entry whose test project is not in that list, and the `tests_requires_cli_archive` job runs only when its filtered matrix is non-empty. `run_cli_e2e` gates test enumeration, not this filtering. Aspire.Cli.EndToEnd.Tests has no `ProjectReference` to src/Aspire.Cli or src/Aspire.Hosting; it consumes a built CLI archive at runtime. So dotnet-affected never selects it, and under any selective scope the CLI end-to-end job would be silently skipped even for direct CLI or Hosting changes. Add a `testProjects` list to a category. When the category is triggered, its projects are injected into the affected set additively — after the matched-but-zero guard (so that guard still sees a genuinely empty integrations resolution) and after restricted filtering (forced projects are explicit opt-ins). Wired in the audit ruleset so cli_e2e forces Aspire.Cli.EndToEnd.Tests for src/Aspire.Cli, src/Aspire.Hosting, and eng/clipack changes. Source-to-integration coverage already flows through dotnet-affected's project graph (a component change reaches its test via ProjectReference), so only the runtime-only couplings need explicit expression here. ## Read the real RequiresNugets property The NuGet-dependent test detector scanned csproj files for a `<RequiredNuGetsForTesting>` element, which no project defines. The actual repo property is `<RequiresNugets>` (eng/testing/CITestsProperties.props), set by the CLI, Templates, and EndToEnd test projects. The detector never matched anything. Renamed the scan and helper to `RequiresNugets`. ## Drop the unused unified rules config eng/scripts/test-selection-rules.unified.json and docs/ci/conditional-test-selection-unified-rules.md described a target design that nothing references. Removed both and the doc pointers to them. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ngine) The test-selection config has been accreting policy verbs into data: each new use case added a JSON field plus an interpreter branch (restrictedTestProjects = suppress, category testProjects = force, excludePaths = exclude). The cli_e2e gap that needed the "force" knob is a symptom: relationships and decisions are tangled, and precedence is implicit across two languages (C# selective + PowerShell RunAll). Add a forward-looking design spec proposing one line: data = relationships only (a typed-edge graph), code = decisions only (one fixed selection engine). Under that split the "force" and "restrict" knobs disappear — force becomes an ordinary runtime-typed edge, restrict becomes a per-node inferDeps fact (or edge negation). The design is validated against prior art (Nx project graph + affected, Pants dependency inference + explicit-edge augmentation, Bazel data edges + target-determinator, the dorny/paths-filter and Develocity/Azure-TIA extremes) and the mechanism-vs-policy principle. Two backbone claims (Bazel's overapproximation invariant, Pants' hermetic BUILD files) were verified against primary sources this session; the rest are cited for spot-checking during review. The spec is design-only and changes no live CI behavior. The override mechanisms it reshapes live only in the audit config today, dormant in production. It documents open decisions (negation granularity, edge direction, the C#/PowerShell dual-enforcement, implementation language) and a migration path that lands behind the audit config first. Add forward pointers from the original spec and the current-behavior doc. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Two companion docs to the test-selection redesign spec, for review and implementation in a follow-up session. - test-selection-redesign-eval.md: a review rubric (Design + Implementation criteria with PASS/FAIL and severity, a "cases the schema still can't express" section, and a blocker-only Go/No-Go gate). Grounded against the current tool code. - test-selection-redesign-plan.md: a phased implementation plan (Phases 0-6), resolved stances on the five open decisions, per-branch falsifiable tests, risk/rollback, and an audit->active promotion gate. The three docs are cross-linked; the design spec now points to both siblings. Open fork flagged at the top of both docs (rubric item D7): there is no first-class test->category label, so the design's claim that run_<category> booleans can never disagree with the matrix is not achievable as written -- deriving the boolean still re-matches test paths against category triggerPaths globs. The rubric recommends a category label on the edge target (relationship in data); the plan resolves it via a selected-set membership check in the engine (relationship in code). To be decided in review. Design-only: no live CI behavior changes. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Contributor
|
❓ CLI E2E Tests unknown — 113 passed, 0 failed, 2 unknown (commit View all recordings
📹 Recordings uploaded automatically from CI run #27181539476 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Ports the conditional test selection work from
conditional-test-selection-unified-rules-designontomainas a squashed change.This brings over the workflow updates, selector tooling, Infrastructure coverage, and supporting docs as a single forward port so the branch-only work can be reviewed against current
mainwithout replaying the original merge-heavy history.Validation performed:
./restore.sh./dotnet.sh build tools/TestSelector/TestSelector.csproj -c Release --nologo./dotnet.sh build tools/Aspire.TestSelector/Aspire.TestSelector.csproj -c Release --nologo./dotnet.sh build tools/GenerateTestSummary/GenerateTestSummary.csproj -c Release --nologo./dotnet.sh build tests/Aspire.Cli.EndToEnd.Tests/Aspire.Cli.EndToEnd.Tests.csproj -c Release --nologo./dotnet.sh build tests/Aspire.EndToEnd.Tests/Aspire.EndToEnd.Tests.csproj -c Release --nologo./dotnet.sh build tests/Aspire.Templates.Tests/Aspire.Templates.Tests.csproj -c Release --nologo./dotnet.sh test tests/Infrastructure.Tests/Infrastructure.Tests.csproj -c Release -- --filter-namespace "*TestSelector*" --filter-namespace "*.PowerShellScripts" --filter-namespace "*.ConditionalSelection" --filter-not-trait "quarantined=true" --filter-not-trait "outerloop=true"Fixes # (issue)
Checklist
<remarks />and<code />elements on your triple slash comments?aspire.devissue: