test(phase 2): add critical-path test coverage across Domain / Application / Infrastructure layers#342
Open
Vasar007 wants to merge 125 commits into
Open
test(phase 2): add critical-path test coverage across Domain / Application / Infrastructure layers#342Vasar007 wants to merge 125 commits into
Vasar007 wants to merge 125 commits into
Conversation
…d test-stack injection - Add Sources/Tests/Directory.Build.props that imports the parent Sources/Directory.Build.props (so child projects keep AppPlatforms / TestTargetFrameworks / CSharpLangVersion / ManagePackageVersionsCentrally) and injects the shared test stack as PackageReference items into every C# test csproj — coverlet.collector + Microsoft.NET.Test.Sdk + xunit + xunit.runner.visualstudio + AwesomeAssertions + NSubstitute (core), plus the opt-in stack Microsoft.AspNetCore.Mvc.Testing + Microsoft.Extensions.TimeProvider.Testing + Testcontainers.PostgreSql + WireMock.Net. - ItemGroup PackageReference items are conditioned on '$(MSBuildProjectExtension)' == '.csproj' so the F# ContentDirectories.Tests project keeps its own Unquote-based stack untouched (Decision D-04 + Specifics §4). - Add four new PackageVersion entries to Sources/Directory.Packages.props: Microsoft.AspNetCore.Mvc.Testing 10.0.8, Microsoft.Extensions.TimeProvider.Testing 10.0.0 (10.0.8 was not published; 10.0.0 is the closest matching version on nuget.org — minor deviation from PLAN), Testcontainers.PostgreSql 4.11.0, WireMock.Net 2.6.0. No Version= attributes on any PackageReference (CPM). - Drop the now-duplicate per-csproj PackageReference entries (coverlet, test sdk, xunit, xunit runner) from ProjectV.Appraisers.Tests and ProjectV.Common.Tests — the shared injection makes them duplicates that would error under TreatWarningsAsErrors=true (NU1504). The retrofit of test bodies + adding the Tests.Shared ProjectReference stays in Task 3 as planned. Decision references: D-02, D-03, D-04 (F# untouched).
…pers
Establishes the shared test-infrastructure library under
Sources/Tests/ProjectV.Tests.Shared/ that every downstream Phase 2 test
plan will reference (Decisions D-02, D-31..D-35).
- ProjectV.Tests.Shared.csproj — RootNamespace=ProjectV.Tests.Shared,
Library, IsPublishable/IsPackable=false. No PackageReference for the
core test stack (Directory.Build.props supplies it). ProjectReferences:
ProjectV.Appraisers, ProjectV.CommonCSharp, ProjectV.Models plus
Acolyte.NET + Microsoft.Extensions.{DependencyInjection,Hosting} via CPM.
- Usings/SharedUsings.cs — global usings exposed to every consumer:
System, Collections.Generic, Linq, Threading.Tasks, AwesomeAssertions,
NSubstitute (+ ExceptionExtensions / ReceivedExtensions), Xunit,
ProjectV.Tests.Shared.ForTests. Scoped #pragma warning disable IDE0005
on the file (intentional unused-in-this-assembly usings — downstream
test projects rely on them).
- ForTests/ base-class hierarchy (D-32):
- BaseTest (root; xUnit-shaped — no [TestFixture], ctor for setup)
- BaseMockTest : BaseTest (NSubstitute Substitute.For helper)
- BaseDependencyInjectionTests : BaseMockTest (CreateServiceCollection /
CreateHostAppBuilder + AssertOn_NotRegistered / RegisteredService /
RegisteredServiceBeOfType using AwesomeAssertions)
- BaseExceptionTests<TException> where TException : Exception : BaseTest
(abstract Create / Create(string) / Create(string, Exception) hooks
plus three [Fact] methods covering the 3-ctor convention; explicit
Arrange/Act/Assert markers)
- TestTimeHelper (thin wrapper around FakeTimeProvider)
- Helpers/Generators/Models/BasicInfoGenerator.cs (D-34) — seeded
Random(Seed: 42) per Specifics §5; Create*/Generate* twin pattern;
Acolyte guard clause; XML docs on every public member.
NOTE: the planned `new Random(seed: 42)` lowercase parameter name does
not compile under .NET 10 (CS1739) — the constructor parameter is
`Seed` (capital S). Used `Random(Seed: 42)` to preserve the intent
(deterministic seed 42 per Specifics §5).
- Helpers/Mocks/Appraisers/TestAppraiserBuilder.cs (D-33) — sealed builder
for IAppraiser substitutes with WithRating / WithRatingFactory fluent
configuration; CreateWithoutSetup static convenience; XML docs.
NOTE: IAppraiser.GetRatings(BasicInfo, bool) returns a single
RatingDataContainer (not a list), so the builder exposes single-rating
configuration rather than the list-state described in PLAN §Task 2
action — the builder shape matches the real production API.
- Helpers/Fixtures/FixtureLoader.cs (D-18) — static LoadJsonFixture
resolving paths under AppContext.BaseDirectory/Fixtures; Acolyte
ThrowIfNullOrWhiteSpace; clear FileNotFoundException on miss.
- Helpers/{WebApi,Stubs,Persistence,Extensions}/.gitkeep — reserved
directories per D-31 + D-35 (downstream plans populate them).
- Sources/Tests/Fixtures/{Tmdb,Omdb,Steam}/.gitkeep — fixture directories
for contract tests (02-08).
- Sources/ProjectV.sln — registered ProjectV.Tests.Shared in the Tests
solution folder with Debug|x64 and Release|x64 configurations
(hand-edited; `dotnet sln add` corrupted the solution with Any CPU /
x86 configurations on every project — restored from HEAD and applied
the entry manually).
All .cs files are UTF-8 with BOM per Sources/.editorconfig.
`dotnet build Sources/ProjectV.sln -c Release -p:Platform=x64` and
`dotnet format Sources/ProjectV.sln --severity warn --verify-no-changes`
both pass with 0 warnings / 0 errors.
Decision references: D-02, D-05, D-18, D-31, D-32, D-33, D-34, D-35.
…ssertions
Translates every Xunit.Assert call in the two existing C# test projects to
AwesomeAssertions, adds [Trait("Category","Unit")] for the xUnit filter
discovery in the upcoming four-stage CI rewrite (02-03), and references
ProjectV.Tests.Shared so downstream plans can rely on the shared base
classes / generators / mock builders being available.
- AppraiserTests.cs: every Assert.NotNull/NotEmpty/Equal/Throws call
rewritten to .Should().Be(...) / .Should().NotBeEmpty() /
.Should().Throw<T>().WithParameterName(...). Explicit AAA comment
markers (// Arrange. / // Act. / // Assert.) per D-07. Scoped
#pragma warning disable CS8625 retained around the null-literal
argument tests. Intentional typo CallGetRatingsWithConteinerWithOneItem
preserved verbatim (Specifics §3). Sealed class shape and explicit
empty ctor unchanged.
- TestDataCreator.cs: RandomInstance seeded with 42 for run-to-run
determinism (Specifics §5). The literal `new Random(seed: 42)` from
PLAN does NOT compile under .NET 10 (CS1739 — parameter is `Seed`
with capital S), so the call site uses `new Random(Seed: 42)` —
preserves the seed value and intent. File retained (callers in
AppraiserTests still use it; ProjectV.Tests.Shared.Helpers.Generators
.Models.BasicInfoGenerator coexists for downstream plans).
- ModelSerializationTests.cs: rewritten to use Newtonsoft.Json
(JsonConvert.SerializeObject / DeserializeObject) which honours
BasicInfo's [JsonConstructor] annotation without a parameterless
ctor — removes the original [Fact(Skip=...)] (Pitfall 7). Also
rewritten to AwesomeAssertions with explicit AAA markers and
[Trait("Category","Unit")] on the class. Both compact + pretty-print
round trips covered.
- Appraisers.Tests + Common.Tests csprojs:
- Drop now-supplied-by-Directory.Build.props PackageReference entries
happened in Task 1 (CPM/restore was already broken otherwise).
- Add ProjectReference to ProjectV.Tests.Shared so downstream test
work can opt into the shared infrastructure incrementally.
- Common.Tests adds explicit Newtonsoft.Json PackageReference (CPM —
no Version=) because the model round-trip now uses Newtonsoft.
- F# ProjectV.ContentDirectories.Tests UNTOUCHED (D-04 + Specifics §4):
`git diff HEAD Sources/Tests/ProjectV.ContentDirectories.Tests/` shows
zero changes; the 9 existing F# tests still pass.
`dotnet build Sources/ProjectV.sln -c Release -p:Platform=x64` exits 0
(0 warnings, 0 errors). `dotnet test` on Appraisers.Tests passes
14/14, Common.Tests passes 1/1, F# ContentDirectories.Tests passes 9/9.
`dotnet format Sources/ProjectV.sln --severity warn --verify-no-changes`
exits 0.
Decision references: D-04, D-07, D-21.
…/Coverage/test-coverage.md
- New file Docs/Testing/Coverage/test-coverage.md — Phase 2 TEST-01 deliverable
- 31 critical-path rows across Domain (8) / Application (10) / Infrastructure (13) sections
- Sources rows verbatim from .planning/phases/02-test-coverage/02-RESEARCH.md (Categorical Critical-Path Inventory)
- Quotes TEST-01 wording from .planning/REQUIREMENTS.md verbatim
- Legend documents Path / Component / Planned Test Project / Test Type / Status columns
and the planned / partially covered / covered / tested around vocabulary
- Trait conventions section names [Trait("Category","Unit")], [Trait("Category","Integration")],
[Trait("Category","Contract")], and the secondary RequiresDocker trait per D-21
- Three anti-patterns from ARCHITECTURE.md (Shell-references-plugins, SimpleExecutor stub) flagged as "tested around"
- Maintenance section explains the downstream contract: flip Status -> covered + add Test Files column
- Cross-references to projectv-scenario-tests-overview.md, ARCHITECTURE.md, INTEGRATIONS.md, REQUIREMENTS.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…agram
- New file Docs/Testing/Scenarios/projectv-scenario-tests-overview.md (D-37)
- Purpose / Audience / Architecture / Scenario Family Documents / Conventions
- Mermaid flowchart TD translates 02-RESEARCH.md System Architecture Diagram
into a renderable diagram with 5 branches (Unit, Integration via WebApplicationFactory,
Contract via WireMock, F# via Unquote) and explicitly marks the dashed-edge
test-double boundary (WireMock for HTTP, Testcontainers for Postgres)
- Conventions section names: sealed class shape, per-family base class,
explicit AAA comment markers, [Trait("Category","Integration")] + [Trait("RequiresDocker","true")],
XML class doc in business terms, [Collection(DbCollection.Name)] for shared container
- Scenario Family Documents section enumerates expected downstream filenames
(projectv-jwt-scenarios.md / projectv-telegram-scenarios.md / projectv-tmdb-pipeline-scenarios.md)
and quotes D-37 that family docs land alongside their suites
- Cross-references to test-coverage.md (TEST-01 inventory) and to
.planning/codebase/ARCHITECTURE.md plus 02-RESEARCH.md patterns
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…ions - Adds the Coverlet runsettings consumed by the Phase 2 CI four-stage rewrite (Plan 02-03) via the `--settings` argument on each Linux test step. - Format: cobertura (matches `dotnet test --collect:"XPlat Code Coverage"` output expected by ReportGenerator's `**/TestResults/**/coverage.cobertura.xml` glob). - Excludes assemblies: `[ProjectV.DesktopApp]*` (Windows-only WPF, untested in Phase 2 by D-04 + D-26 — Linux coverage scope) and `[ProjectV.*.Tests]*` (test assemblies are not production code). - ExcludeByFile: `**/Migrations/**/*.cs` (EF schema migrations) and `**/*.Designer.cs` (generated WinForms/resource designers). - Satisfies D-27.
…r + coverage publication
- Replaces the omnibus `Test (C# projects)` + `Test (F# ContentDirectories)`
steps with four named Linux stages: `Test (Unit)`, `Test (Integration)`,
`Test (Contract)`, `Test (F#)` (D-20, D-21, D-23). Each C# stage filters by
the xUnit `[Trait("Category", ...)]` trait established in Plan 02-01;
the F# stage keeps the existing explicit `fsproj` invocation but is now a
first-class named step (closes the silent-skip wording in TEST-05).
- Adds coverage publication on the Linux job only (D-24, D-26): each Linux C#
stage runs `--collect:"XPlat Code Coverage" --settings Sources/Tests/coverlet.runsettings`;
`danielpalme/ReportGenerator-GitHub-Action@5` merges the per-stage Cobertura
outputs into an HtmlInline + MarkdownSummaryGithub bundle;
`actions/upload-artifact@v4` publishes the `coverage-report` directory;
`coverage-report/SummaryGithub.md` is appended to `$GITHUB_STEP_SUMMARY`
for an in-PR coverage panel.
- Adds a Windows `Test (Non-Docker)` stage (D-22) — single-quoted YAML filter
`'RequiresDocker!=true'` so PowerShell does NOT history-expand `!=`
(Pitfall 4) and YAML keeps the string verbatim. Docker-dependent
Testcontainers tests stay Linux-only via the `[Trait("RequiresDocker","true")]`
tag.
- Branch-protection contract preserved (Pitfall 5): job name
`Build (${{ matrix.os }})` unchanged, so the required-checks list
(`Build (ubuntu-latest)`, `Build (windows-latest)`, `CodeQL`) needs no
admin update. Triggers (push + pull_request on master / phase-** / feature/**),
matrix definition, Checkout, Setup .NET, Restore (both solutions), Build
(both solutions), and Format check all unchanged.
Satisfies TEST-05 (four named stages, F# first-class) and TEST-06 (coverage
artifact + per-run step summary).
- TestAppraisersManagerBuilder: real AppraisersManager populated with NSubstitute IAppraiser children (sealed-class fallback per D-33 + PLAN.md Task 1 action). - UserIdGenerator + JobIdGenerator: Create/Generate twin pattern (D-34); raw GUID fed via UserId.Parse / JobId.Parse. - XML docs on every public member; Acolyte guard on Create*. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…object, BasicInfo invariant suites
- New Sources/Tests/ProjectV.Models.Tests/ project, RootNamespace
ProjectV.Models.Tests, ProjectReferences to ProjectV.Models +
ProjectV.Tests.Shared, Newtonsoft.Json PackageReference (CPM, no
Version=). Registered in Sources/ProjectV.sln under the Tests folder
with Debug|x64 + Release|x64 only — no AnyCPU.
- Exceptions/CannotGetTmdbConfigExceptionTests: inherits
BaseExceptionTests<TException> from 02-01 (D-32), 3 inherited Facts.
- Exceptions/CommonExceptionsTestSuite: reflection-driven 3-ctor
convention enforcement across every sealed Exception subtype in
ProjectV.Models.Exceptions — auto-discovers new types.
- ValueObjects/UserIdTests + JobIdTests: 13 facts each covering Create,
Wrap, Parse, TryParse, None, IsSpecified, round-trip; uses
UserIdGenerator/JobIdGenerator from 02-04 Task 1.
- Data/BasicInfoInvariantsTests: primitive defaults, Kind discriminator,
equality semantics (memberwise + 1e-6 tolerance on VoteAverage),
Newtonsoft.Json compact + pretty round-trip.
- [Trait("Category", "Unit")] on every class — 43 tests pass under the
Unit CI filter.
- Docs/Testing/Coverage/test-coverage.md: Domain rows flipped to
covered with new Test Files column; BasicInfo invariants row split
from MovieInfo/GameInfo (still planned).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…aisersManager suites
- AppraisersExtensions/MovieCommonAppraiserTests: 8 facts verifying
Appraiser<TmdbMovieInfo> + TmdbCommonAppraisal end-to-end —
Tag/TypeId/RatingName, popularity-as-rating, null guard,
type-mismatch guard, ctor guard.
- AppraisersExtensions/MovieNormalizedAppraiserTests: 7 facts
verifying Appraiser<BasicInfo> + BasicAppraisalNormalized with
PrepareCalculation/RawDataContainer/MinMaxDenominator wiring —
min/max endpoints map to 0/2, middle item bounded, missing-prepare
throws InvalidOperationException, null guards.
- AppraisersExtensions/AppraisersManagerTests: 10 facts verifying
Add/Remove/CreateFlow over AppraisersManager via the new
TestAppraisersManagerBuilder + TestAppraiserBuilder doubles —
null guards, idempotent Add, two-instance-same-TypeId flow,
Remove existing/missing, flow distinctness.
- AppraisersExtensions/AppraisersManagerTestsInit: [ModuleInitializer]
pins NLog.LogManager.Configuration to an empty LoggingConfiguration
so the pre-existing Sources/Libraries/ProjectV.Logging/NLog.config
bug (NLog 6 dropped `concurrentWrites="true"` but the file still
declares it under throwConfigExceptions=true) does not trip the
AppraisersManager static logger field at type-init time.
- Plan named the rating-computation files
MovieCommon/MovieNormalizedAppraiserTests; ProjectV has no such
types — the production shape is Appraiser<T> + IAppraisal<T>.
Tests exercise the strategy directly (it IS the unit under test
for rating-computation accuracy). Documented inline on each file.
- Docs/Testing/Coverage/test-coverage.md: Appraiser rows flipped to
covered; concrete-appraiser row split into MovieCommon (covered) +
MovieNormalized (covered) + Game-suite (still planned).
- All 25 new facts carry [Trait("Category", "Unit")]; existing 14
AppraiserTests facts unchanged — 39 Unit tests in Appraisers.Tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…uilders to Tests.Shared - TestInputManagerBuilder / TestCrawlersManagerBuilder / TestOutputManagerBuilder: D-33 fallback builders returning real sealed concrete managers populated with NSubstitute child doubles via the public Add API - TestShellBuilder under Helpers/Mocks/Core/: composes the four managers (defaulting to empty CreateWithoutSetup() instances) and exposes a fluent WithXxxManager / WithBoundedCapacity chain - TestCommunicationServiceClientBuilder under Helpers/Mocks/Core/: substitutes the ICommunicationServiceClient interface seam with optional Login/StartJob success or error responses (Result<TOk, ErrorResponse>) - ProjectV.Tests.Shared.csproj: add ProjectReferences to ProjectV.Core, ProjectV.Crawlers, ProjectV.InputProcessing, ProjectV.OutputProcessing so the new builders compile Note: the plan's TestAppraisalManagerBuilder is intentionally NOT added — the production type is AppraisersManager and the matching builder (TestAppraisersManagerBuilder) already shipped with 02-04. Decision recorded in 02-05-SUMMARY.md.
…r unit suites - ProjectV.Core.Tests.csproj: new test project mirroring Sources/Libraries/ProjectV.Core/ - ShellTests: 8 facts covering ctor null-guards (5 args), property surface, Dispose idempotency, CreateBuilderDirector static factory. Run() branch tests intentionally omitted — see SUMMARY § Deviations for the Gridsum.DataflowEx empty-pipeline blocker (deferred to integration). - ShellBuilderFromXDocumentTests: 6 facts covering ctor null/missing-root guards, minimal-config happy path, GetResult-before-build guard, Reset, BuildMessageHandler missing-element error path. - ShellBuilderDirectorTests: 6 facts covering ctor null-guard, happy path, ChangeShellBuilder null-guard, MakeShell invokes all 7 IShellBuilder methods (Reset + 5 build steps + GetResult) in declared order, and dispatches to a replaced builder. - CoreTestsModuleInitializer: [ModuleInitializer] installs an empty NLog LoggingConfiguration so Shell / ShellBuilderFromXDocument / CommunicationServiceClient static loggers don't trip the pre-existing NLog.config concurrentWrites bug (same pattern as Appraisers.Tests AppraisersExtensions/TestModuleInitializer). - Sources/ProjectV.sln: register ProjectV.Core.Tests under Tests folder with Debug|x64 + Release|x64 only (no AnyCPU). - Docs/Testing/Coverage/test-coverage.md: flip Shell.Run row to 'tested around' with rationale, ShellBuilderFromXDocument + ShellBuilderDirector rows to 'covered' with Test Files paths. Added Test Files column to Application Layer. - Tests.Shared mock builders: UTF-8 BOM normalized via dotnet format.
… Net/
- CommunicationServiceClientTests: 3 facts covering LoginAsync.
Constructs the production concrete CommunicationServiceClient with a
substituted IHttpClientFactory whose HttpClient is backed by an inline
FakeHttpMessageHandler (DelegatingHandler). Asserts:
1. 200 + token JSON body → Result.Ok<TokenResponse> with the access token.
2. 401 + ErrorResponse body → Result.Error<ErrorResponse> (NOT a thrown
exception — production code uses Result<TOk, ErrorResponse>; recorded
in SUMMARY as deviation from plan's 'throws AuthFailureException' wording).
3. null login request → ArgumentNullException('login').
StartJobAsync deferred to integration tests (deviation rationale in SUMMARY).
- HttpClientPollyPolicyTests: 3 facts covering the Polly retry policy wired
by AddHttpClientWithOptions → AddHttpOptions → AddTransientHttpErrorPolicy:
1. 503 → 503 → 503 → 200 with RetryCountOnFailed=3 → 4 handler invocations.
2. Always-503 with RetryCountOnFailed=2 → 3 invocations (initial + 2 retries).
3. First-call-200 → 1 invocation (no retry on success).
Uses an inline DelegatingHandler subclass set as the primary HTTP message
handler via ConfigurePrimaryHttpMessageHandler — Substitute.For<HttpMessageHandler>
is explicitly avoided per 02-RESEARCH.md Pitfall 6 (NSubstitute cannot mock
protected methods).
- ProjectV.Core.Tests.csproj: add Newtonsoft.Json PackageReference (CPM, no
Version) for serializing the test response bodies — production code on this
path uses Newtonsoft.Json (System.Text.Json migration deferred per 02-CONTEXT).
- Docs/Testing/Coverage/test-coverage.md: flip CommunicationServiceClient row
and AddHttpClientWithOptions Polly retry row to 'covered' with Test Files.
…Shared - TestTmdbCrawlerBuilder / TestOmdbCrawlerBuilder / TestSteamCrawlerBuilder: D-33 builders for ICrawler substitutes. Each yields an IAsyncEnumerable to match the real ICrawler.GetResponse(string, bool) signature (not Task-returning as the PATTERNS.md illustrative template implied). Default Tag matches the production crawler's nameof; supports WithThrowOnGetResponse(ex) for exercising CrawlersManager log+rethrow. - TestDataflowPipelineBuilder: D-33 fallback for the sealed DataflowPipeline (real instance with optional InputtersFlow / OutputtersFlow stages; empty flows by default). - ProjectV.Tests.Shared.csproj picks up ProjectV.DataPipeline as a ProjectReference so the new DataPipeline builder compiles.
…th rows New test projects: - ProjectV.DataPipeline.Tests (Integration) — DataflowPipelineTests + InputtersFlowTests. The DataflowPipeline integration test wires the full InputtersFlow → CrawlersFlow → AppraisersFlow → OutputtersFlow stages with NSubstitute ICrawler/IAppraiser leaves; the InputtersFlow tests cover the dedup + length-filter contract via reflection on the private FilterInputData predicate (predicated-link without LinkLeftToNull() deadlocks Gridsum.DataflowEx completion — see SUMMARY Deviations) plus a happy-path end-to-end smoke. - ProjectV.Crawlers.Tests (Unit) — CrawlersManagerTests covering the log+rethrow contract on the private TryGetResponse (reflection probe; the static NLog logger half is intentionally not substituted), plus ctor + Add null-guard + Remove happy path. ModuleInitializer in each new project installs an empty NLog.LoggingConfiguration to bypass the NLog 6 / concurrentWrites auto-load bug (same pattern as 02-04 / 02-05). Sources/ProjectV.sln hand-edited (per 02-01 lesson — `dotnet sln add` corrupts to AnyCPU) to register both projects under the Tests folder with Debug|x64 + Release|x64 only. Docs/Testing/Coverage/test-coverage.md updated: 3 rows in the Application Layer section flipped from `planned` to `covered` (DataflowPipeline.Execute / InputtersFlow / CrawlersManager.TryGetResponse) with rationale notes for the deviation cases.
…test slice
Adds three new test projects covering the executors / input / output
critical-path rows from the Phase 2 inventory (TEST-03 subset):
- ProjectV.Executors.Tests — SimpleExecutorTests asserts the current
parameterless ExecuteAsync() anti-pattern (NotImplementedException
stub per ARCHITECTURE.md). Inventory row flipped to "covered (tested
around)". Also covers ctor null-guard, executions-number guard, and
happy-path property exposure.
- ProjectV.InputProcessing.Tests — InputManagerTests asserts
CreateFlow(string) returns non-null InputtersFlow (with/without
registered IInputter substitutes), ctor null/whitespace guard, Add
null-guard, and Remove round-trip. Empty-storage-name path is left
to higher-level integration coverage because InputManager.CreateFlow
reaches into the process-wide GlobalMessageHandler static.
- ProjectV.OutputProcessing.Tests — OutputManagerTests asserts the
analogous CreateFlow(string) → non-null OutputtersFlow contract,
including the empty-storage-name fallback (OutputManager does not
reach into GlobalMessageHandler), ctor null/whitespace guard, Add
null-guard, and Remove round-trip.
Each project carries [Trait("Category", "Unit")] per D-21, picks up
the shared test stack from Sources/Tests/Directory.Build.props (D-03),
and uses NSubstitute for IInputter / IOutputter collaborators (D-05).
A ModuleInitializer mirrors the 02-05 / 02-06 NLog auto-load
short-circuit pattern for assemblies that touch production types with
static `_logger` fields.
Sources/ProjectV.sln — three new project entries with Debug|x64 +
Release|x64 only (no AnyCPU), nested under the Tests solution folder.
Docs/Testing/Coverage/test-coverage.md — three Application Layer rows
flipped from `planned` to `covered`.
Refs: 02-07 Plan / Phase 2 Test Coverage / TEST-03
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…t tests - Sources/Tests/Fixtures/Tmdb/search-movie-success.json (id=12345) - Sources/Tests/Fixtures/Tmdb/search-movie-empty.json (zero-results envelope) - Sources/Tests/Fixtures/Tmdb/configuration-success.json (images + change_keys) - Sources/Tests/Fixtures/Omdb/movie-by-title-success.json (Response:"True") - Sources/Tests/Fixtures/Omdb/movie-by-title-not-found.json (Response:"False") - Sources/Tests/Fixtures/Steam/app-list-success.json (applist + 3 entries) - Sources/Tests/Fixtures/Steam/app-detail-success.json (appid 730 envelope) Synthetic data only — no live-API responses, no keys, no PII (D-17). Fixture filenames reflect production surface (TmdbClient exposes TrySearchMovieAsync / GetConfigAsync, not GetMovieAsync) — Rule 1 deviation from plan filenames movie-by-id-*.json (no such SUT method exists); documented in SUMMARY. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three new test projects under Sources/Tests/, each tagged
[Trait("Category", "Contract")] so the 02-03 CI Contract stage picks
them up:
- ProjectV.TmdbService.Tests (3 tests):
TrySearchMovieAsyncReturnsExpectedContainer,
TrySearchMovieAsyncEmptyResultReturnsEmptyContainer,
GetConfigAsyncReturnsExpectedConfig.
Redirection seam: new TmdbClient(apiKey, useSsl: false,
baseUrl: WireMockHostPort) via InternalsVisibleTo.
- ProjectV.OmdbService.Tests (2 tests):
TryGetItemByTitleAsyncReturnsExpectedMovie,
TryGetItemByTitleAsyncNotFoundReturnsNull.
Redirection seam: HttpClient.DefaultProxy = new WebProxy(WireMock.Url)
because OMDbApiNet 1.3.0 hardcodes BaseUrl as a const field.
- ProjectV.SteamService.Tests (2 tests):
GetAppListAsyncReturnsExpectedApps,
TryGetSteamAppAsyncReturnsExpectedApp.
Redirection seam: reflection-replace _steamApiClient with an SDK
instance built from a SteamApiConfig whose SteamPoweredBaseUrl /
SteamStoreBaseUrl point at WireMock.
Each test uses real-bytes-on-the-wire against in-process WireMockServer
instances serving the 7 recorded JSON fixtures from Sources/Tests/Fixtures/
(committed in the previous commit). No live API calls, no API keys, no
mocked HttpMessageHandler. Each test asserts LogEntries.Should().HaveCount(1)
to verify exactly-once HTTP semantics on the success path.
Three production libraries get a Properties/AssemblyInfo.cs adding
InternalsVisibleTo("ProjectV.<X>Service.Tests") — minimal, opt-in
seam-widening so the internal sealed wrapper types are constructable from
the test assembly. Same pattern already used by ProjectV.Appraisers /
ProjectV.Appraisers.Tests.
Sources/ProjectV.sln registers all three new projects under the Tests
solution folder (Debug|x64 + Release|x64 only, no AnyCPU).
Docs/Testing/Coverage/test-coverage.md: the three Infrastructure rows
for TmdbClient / OmdbClient / SteamApiClient flip to `covered` with
their test files; the table grows the `Test Files` column for the
remaining `planned` rows in the same section.
Build: dotnet build Sources/ProjectV.sln -c Release -p:Platform=x64 → 0 warnings.
Tests: dotnet test ... --filter "Category=Contract" → 7/7 passed (3+2+2).
Format: dotnet format Sources/ProjectV.sln --severity warn --verify-no-changes → clean.
No regressions: Unit=129, Integration=7, F#=9, Contract=7 (was 129/7/9 → +7 Contract).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
… BLOCKING fallback) Adds the EF Core 10.0.8 design-time tooling required to attempt `dotnet ef migrations add InitialCreate` for ProjectV.DataAccessLayer: - CPM: Microsoft.EntityFrameworkCore.Design 10.0.8 entry. - ProjectV.DataAccessLayer.csproj: PackageReference (PrivateAssets="All"). - ProjectV.ProcessingWebService.csproj: PackageReference (PrivateAssets="All") so the EF CLI accepts it as the startup project. - ProjectVDbContextDesignTimeFactory: implements IDesignTimeDbContextFactory<ProjectVDbContext> so the CLI can construct the context (the production type exposes two single-arg ctors and the CLI cannot disambiguate them on its own). - Migrations/.gitkeep with a remark documenting how the generation attempt was made and why it stops at design-time model discovery. [BLOCKING] Migration generation does NOT succeed in 02-09. EF Core rejects every ctor of UserDbInfo because the immutable `RefreshTokenDbInfo refreshToken` parameter cannot bind to a mapped scalar or navigation. Fixing this requires architectural changes (parameterless ctor / explicit HasOne / mapper-only nav) which are out of scope for 02-09. The 02-09 DbCollectionFixture takes the RESEARCH.md fallback path and bootstraps the test container with raw SQL — see Task 2 commit + 02-09-SUMMARY.md "[BLOCKING] Migration generation deferred".
…sk 2)
Adds the Testcontainers PostgreSQL test infrastructure for the DAL slice
plus three D-34 entity generators in ProjectV.Tests.Shared.
ProjectV.Tests.Shared additions:
- Helpers/Generators/DataAccessLayer/JobInfoGenerator.cs (Create/Generate
twin, seeded Random(seed: 42), composes with JobIdGenerator).
- Helpers/Generators/DataAccessLayer/UserInfoGenerator.cs (Create/Generate
twin, composes with UserIdGenerator).
- Helpers/Generators/DataAccessLayer/RefreshTokenInfoGenerator.cs
(Create/Generate twin, composes with UserIdGenerator).
- ForTests/TestDbHelper.cs — TRUNCATE TABLE … RESTART IDENTITY CASCADE on
public.{jobs,users,tokens}; ChangeTracker.Clear() to avoid stale-entity
DbUpdateConcurrencyException between cases (D-11).
- ProjectReference on ProjectV.DataAccessLayer (TestDbHelper consumes
ProjectVDbContext).
ProjectV.DataAccessLayer.Tests (new):
- ProjectV.DataAccessLayer.Tests.csproj — references DataAccessLayer +
Configuration + Models + Tests.Shared; Testcontainers, EF Core, Npgsql
flow transitively or via Sources/Tests/Directory.Build.props (D-03).
- ForTests/DbCollection.cs — [CollectionDefinition("ProjectV.DAL.Db")] +
ICollectionFixture<DbCollectionFixture>.
- ForTests/DbCollectionFixture.cs — PostgreSqlBuilder("postgres:16.4")
(pinned image, Pitfall 1), WithDatabase/User/Password, WithWaitStrategy
Wait.ForUnixContainer().UntilInternalTcpPortIsAvailable(5432). Sets
CanUseDatabase=true on every DatabaseOptions (Pitfall 2).
- Schema bootstrap path: raw SQL CREATE TABLE statements derived from
[Table]/[Column] attrs on JobDbInfo/UserDbInfo/RefreshTokenDbInfo. See
the class XML doc + Migrations/.gitkeep + 02-09-SUMMARY.md "[BLOCKING]
Migration generation deferred" for why MigrateAsync/EnsureCreatedAsync
are NOT used in Phase 2.
Sources/ProjectV.sln registers ProjectV.DataAccessLayer.Tests with x64-only
configs (Debug|x64 + Release|x64); no AnyCPU.
`dotnet build Sources/ProjectV.sln -c Debug -p:Platform=x64` is green.
…ask 3)
Adds four DAL integration test classes against the live Testcontainers
PostgreSQL fixture from Task 2 plus the Rule 1 production fixes that
unblock them.
Test classes (10 [Fact] methods, all green against postgres:16.4):
- Services/Jobs/DatabaseJobInfoServiceTests.cs (3 tests):
AddAsync round-trip, FindByIdAsync round-trip, UpdateAsync mutation.
- Services/Users/DatabaseUserInfoServiceTests.cs (3 tests):
AddAsync, FindByIdAsync, FindByUserNameAsync lookup.
- Services/Tokens/DatabaseRefreshTokenInfoServiceTests.cs (2 tests):
AddAsync, FindByIdAsync expiry round-trip.
- ProjectVDbContextSchemaTests.cs (2 tests):
information_schema query for public.{jobs,users,tokens};
CanUseDb() + Npgsql connection sanity.
Every class declares
[Trait("Category", "Integration")] [Trait("RequiresDocker", "true")]
[Collection(DbCollection.Name)]
so the Linux CI Integration stage picks them up and Windows Non-Docker
correctly filters them out (RequiresDocker!=true).
Rule 1 production-code fixes (necessary for the tests to run at all —
see 02-09-SUMMARY.md Deviations for the full reasoning):
- UserDbInfo: `RefreshToken` marked [NotMapped] + new internal 6-arg ctor
for EF Core ctor binding. The previous 7-arg ctor + `builder.Property(e
=> e.RefreshToken)` configuration raised ModelValidator on EVERY DbContext
access. Mapper path (7-arg ctor → 6-arg ctor → 7-arg sets RefreshToken)
preserves existing semantics for DataAccessLayerMapper.MapToUserDbInfo.
- DatabaseUserInfoService.FindByUserNameAsync: rewritten to compare the
raw `user.UserName` string scalar instead of the unmapped
`user.WrappedUserName` computed property — EF Core cannot translate the
latter ("Translation of member 'WrappedUserName' on entity type
'UserDbInfo' failed. This commonly occurs when the specified member is
unmapped.").
- DatabaseRefreshTokenInfoService.FindByUserIdAsync: same fix —
compare the raw `token.UserId` Guid scalar instead of the never-assigned
`token.WrappedUserId` record-struct property.
Test infrastructure:
- TestDbHelper now takes a connection string and uses Npgsql directly so
the per-test TRUNCATE does not depend on a working ProjectVDbContext.
- DbCollectionFixture's ApplySchemaAsync uses Npgsql directly (raw SQL
CREATE TABLE) for the same reason — the bootstrap is independent of
the EF model.
Docs/Testing/Coverage/test-coverage.md: 4 Infrastructure rows flipped
planned → covered with status notes pointing at the SUMMARY deviations.
`dotnet build Sources/ProjectV.sln -c Debug -p:Platform=x64` is green;
`dotnet format Sources/ProjectV.sln --severity warn --verify-no-changes`
exits 0; `dotnet test ... --filter "Category=Integration"` reports
10/10 passing against the live container; no regressions in Unit / Contract.
The dispatch and duplicate-registration tests only observed mock property reads and could not fail if the manager stopped routing entities. They now post an entity into the constructed flow, await the emitted rating, and assert the child appraiser was invoked exactly once; the registration test is renamed to state what it verifies. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Expected ratings in the appraiser suite were computed by re-running the production appraisal, so a formula regression could never fail them. A new test pins the rating to a hand-computed literal (VoteAverage), and the creator helper now documents the wiring-only nature of the derived expectations. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…security audit Transitive packages Microsoft.OpenApi 2.4.1 and Scriban.Signed 7.0.6 trip NuGetAudit (NU1902/NU1903) as hard errors under TreatWarningsAsErrors, failing restore solution-wide. Enable central transitive pinning and pin the lowest patched versions: - Microsoft.OpenApi 2.7.5 (fixes GHSA-v5pm-xwqc-g5wc) - Scriban.Signed 7.2.1 (fixes GHSA-24c8-4792-22hx, GHSA-6q7j-xr26-3h2c, GHSA-q6rr-fm2g-g5x8) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace per-dependency constructor null-argument [Fact]s in ShellTests,
DataflowPipelineTests, and SimpleExecutorTests with the canonical
two-test shape: a positive Constructor_OnAllArgumentsProvided test built
on a Create{Sut}WithAllArguments factory helper, and a single negative
test iterating a list of actions that pass exactly one dependency as
null and assert the ArgumentNullException parameter name. Non-null-guard
constructor cases (ArgumentOutOfRangeException, property population)
stay as separate tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add BaseGuidWrapperTests<TId> to ProjectV.Tests.Shared encoding the shared behaviour of GUID-backed identifier value-objects (None/default equality, Create, Wrap, Parse, TryParse valid/invalid/null, and generator round-trips) behind abstract hooks. Collapse the previously near-verbatim JobIdTests and UserIdTests twins into thin subclasses that only bridge each type's static surface and its test-data generator; every previous assertion is preserved on the base class. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Owner
Author
Test-suite quality pass — audit + fixesRan a full quality audit of the C# test suite (81 files) against the project's test conventions and the canonical example test project, then verified and fixed every valid finding. 24 commits on this branch ( What changed (test code)
Dependency security fix
Verification at HEAD
|
The webhook and polling scenario base classes were near-verbatim twins: duplicate ResolvedBotStubs holder, IBotService/ICommunicationServiceClient DI swaps, NLog capture hook, and in-memory configuration builder differing only in the WorkingMode value. Hoist the shared machinery into an abstract TelegramScenarioBaseTest (Scenarios/Helpers) parameterized by working mode, shrink both per-family bases to thin wrappers, and move the text-message Update builder onto the shared base so webhook and polling scenarios arrange updates identically. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
… helpers Several appraiser and pipeline tests constructed BasicInfo and RatingDataContainer directly in test bodies although Tests.Shared ships BasicInfoGenerator with the exact explicit-args creation method. Add one private CreateBasicInfo helper per class delegating to the generator singleton and a private CreateRating helper wrapping the RatingDataContainer constructor (no generator exists for it yet), and route every inline creation site through them. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
CreateBasicInfoListRandomly multiplied unbounded Random values by the loop index, so voteCount overflowed Int32 into negative values for most elements and voteAverage reached ~100 against the documented 0-10 domain; the random title suffix length was also coupled to the list size. Route each element through BasicInfoGenerator.GenerateBasicInfo (sequential thing ids, generator-bounded random fields, GUID-unique titles) and drop the hand-rolled random-string helper. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
Eighteen With* builder methods had zero consumers anywhere in the test tree: the entire slot mechanism on TestShellBuilder, the child/storage overrides on the three manager builders, WithAppraiser and WithOutputResults on TestAppraisersManagerBuilder, WithRatingFactory on TestAppraiserBuilder, and WithResponses on TestTmdbCrawlerBuilder. Delete the dead methods together with their now-unneeded backing fields and update the affected XML docs; the consumed surface (CreateWithoutSetup, Build, WithAppraisers, WithResponse, WithTag, WithTypeId, WithRating, WithThrowOnGetResponse) is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
…helper CreateMovie declared non-nullable optional voteAverage/voteCount parameters that no caller ever overrode; test-class data helpers must be optional-free. Keep the single popularity parameter callers actually vary and inline the remaining fields as fixed valid values. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
The middle-item test only asserted the [0, 2] envelope although the fixed batch makes the value hand-computable ((50-10)/(90-10) + (7-5)/(9-5) = 1.0), so an in-envelope formula regression passed unnoticed. Assert the exact value with the arithmetic in the assertion reason and rename the test accordingly. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
…eator The (Guid, params BasicInfo[]) overload of CreateExpectedValueForBasicInfo had no callers - the sole external call site passes an IReadOnlyList and binds to the IEnumerable overload - and was reachable only through its own class remarks cref. Delete the overload and retarget the cref at the surviving one. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
The coverage-scope paragraph carried a doubled article ("in an a future")
and residual forward-looking phrasing left over from an earlier doc
rewrite. Reword the sentence to plain prose.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
…suite ShellTests and ShellBuilderFromXDocumentTests carried byte-identical minimal-XDocument helpers under two different names in the same assembly. Rename CreateMinimalConfiguration to CreateMinimalShellConfigXml so the sibling classes use one name for one shape. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
The comment claimed `await Task.CompletedTask` forces the method to be truly asynchronous - an already-completed awaitable continues synchronously and forces nothing. Reword it to state the actual purpose: satisfying the compiler's async-iterator requirement while iteration remains synchronous. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
Three test files spelled out System.Collections.Generic.IReadOnlyList, System.Collections.Concurrent.ConcurrentBag, and System.Data.ConnectionState inline although importing the namespace is the file and suite norm. Add the missing using directives and shorten the references. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
The by-title contract test arranged a title constant but asserted against a repeated string literal. Compare against the variable so the arranged value and the expectation cannot drift apart. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
Remove the no-signal Remove(...) tail assertion from the Add-idempotency test (Remove returns true whether one or two Adds happened and its behaviour is covered by the dedicated Remove tests), and move the CreateFlow call in CreateWithoutSetupReturnsEmptyManager from the Assert section into a proper Act section. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G7ErBrm9WdLKFMzykyxKXW
Pass every argument to the Create{Sut}WithAllArguments factories by name
in both the positive and negative constructor tests, so each null-guard
entry reads as a grep-friendly single-parameter diff.
The factories intentionally keep required, non-nullable parameters: an
optional-parameter variant with null-coalescing fallbacks would replace
an explicitly passed null with a valid dependency, so the negative test
could never reach the constructor guard (verified empirically - the
assertion reports "no exception was thrown"). Required parameters keep
each null observable by the guard and match the reference test shape
used elsewhere.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The two scenario base classes were saved without the UTF-8 BOM required by .editorconfig, which fails the formatting verification step in CI. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Owner
Author
Test-suite quality pass — second audit roundRe-audited the full C# test suite at the new HEAD against the project's test conventions and the canonical example test project, including the code changed by the first round. 15 commits ( Real defect found and fixed
Cleanups from the first round's own changes
Constructor null-guard tests — a note worth recordingThe audit flagged that Verification at HEAD
|
BaseDependencyInjectionTests in ProjectV.Tests.Shared had zero consumers across the whole test tree — no test class inherits it and no member is called. Delete it and repoint the package-reference comment in the csproj at the actual consumers of the DI/Hosting packages (WebApiBaseTest and TestWebApplicationFactory). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…rator GetRatingsThrowsForBaseBasicInfoBecauseAppraiserExpectsTmdb built its BasicInfo input inline in the test body. Add the same private CreateBasicInfo helper the sibling appraiser test classes carry (delegating to BasicInfoGenerator.Instance) and route the test through it, so no test body creates model data inline. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
UpdateAsyncWithExistingJobPersistsChanges built its mutated JobInfo copy inline in the test body. Add a private CreateMutatedJobInfo helper that delegates to JobInfoGenerator.CreateJobInfo, preserving the exact mutation the assertions depend on (state and result incremented, id, name and config kept from the original row). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ce exclusion The webhook base class remark claimed the polling / webhook hosted services would resolve IBotService before the ConfigureTestServices override wins — that is wrong (test overrides mutate the service collection before the provider is built, and hosted services resolve only at host start, as the polling sibling base documents correctly). Reword to the real reason: excluding those hosted services keeps their stub calls out of BotServiceStub.CalledMethodNames and limits each scenario family's host to the machinery its scenarios exercise. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The class remark claimed the test stops the host explicitly inside the act-phase polling loop — it never does; the host is stopped by WebApiBaseTest disposal after the test, and the act phase only waits with a bounded timeout for the expected SendMessageAsync call count. The arrange comment also referenced a nonexistent WithUpdateSequence method; the stub is pre-loaded through its constructor. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
TestShellBuilder and ShellTests docs still claimed the composed managers are populated with NSubstitute child doubles. Since the fluent With* slots were removed, every sibling manager builder returns an empty production manager via CreateWithoutSetup(), and no substitutes are attached through this path. Reword both docs to match the code. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The arrange comment claimed the inputter echoes the storage-name input back as the entity name, but the lambda ignores its input and always yields the fixed entity name. Reword the comment to match the code. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Owner
Author
Test-suite quality pass — third audit roundRe-audited the full C# test suite at the new HEAD against the project's test conventions and the canonical example test project, with particular attention to the code the second round changed. 7 commits ( The suite is now materially clean: 0 high-severity findings, 0 new production findings, and every hard convention prohibition passes. What remained was one dead class and documentation that the previous rounds' own refactors had left stale. What changed
Verification at HEAD
|
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
1 participant
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.
Closes #341.
Summary
Phase 2: Test Coverage
Milestone: v0.9.8 — Test Coverage & Agent-Testable Feedback Loop
Goal: Identify the critical paths across Domain, Application, and Infrastructure layers and back them with a deliberate mix of unit, integration, and contract tests — running in the right CI stages and surfacing a coverage report for visibility, without committing to a 100 % gate.
Brownfield, additive only — no architectural rewrite. The existing TPL-Dataflow + ShellBuilder pipeline is unchanged; it is now testable and verifiable. 13 plans landed across 5 waves: a shared test library bootstrapped first (
ProjectV.Tests.Shared), then the critical-path inventory + CI four-stage rewrite + coverage publication, then concentric slices of Domain unit tests, Application unit tests, pipeline (Dataflow + Crawlers + Executors + IO) tests, contract tests (WireMock.Net) against the three rating-API adapters, DAL integration tests via Testcontainers Postgres, JWT integration tests, Telegram webhook + polling integration tests, and finally a single gap-closure plan (02-13) that absorbed all findings from the post-implementation code-review loop.Changes
Plan 02-01 — Tests Shared Bootstrap (Wave 1)
Sources/Tests/Directory.Build.propsinjects the shared C# test stack (xunit + AwesomeAssertions + NSubstitute + coverlet + opt-in TimeProvider.Testing / Testcontainers / WireMock.Net / AspNetCore.Mvc.Testing) into every.csprojunderSources/Tests/while leaving the F#ProjectV.ContentDirectories.Tests.fsprojuntouched (D-04). New libraryProjectV.Tests.Sharedexposes the base test classes (BaseTest,BaseMockTest,BaseDependencyInjectionTests,BaseExceptionTests<T>,TestTimeHelper), the first generator + mock-builder examples, and aFixtureLoader. ExistingProjectV.Appraisers.Tests+ProjectV.Common.Testsretrofitted to AwesomeAssertions; the previously skippedModelSerializationTestsnow runs and passes. 4 new CPM<PackageVersion>entries added.Key files:
Sources/Tests/Directory.Build.props,Sources/Tests/ProjectV.Tests.Shared/**,Sources/Directory.Packages.props,Sources/ProjectV.sln.Plan 02-02 — Coverage Inventory + Scenarios Doc (Wave 2)
Docs/Testing/Coverage/test-coverage.mdenumerates the critical paths across Domain / Application / Infrastructure and maps each row to the plan that covers it (TEST-01).Docs/Testing/scenarios.mddescribes the four-tier scenario taxonomy + a Mermaid architecture diagram.Key files:
Docs/Testing/Coverage/test-coverage.md,Docs/Testing/scenarios.md.Plan 02-03 — CI Four-Stage Rewrite + Coverage Publication (Wave 2)
.github/workflows/build.ymlis split from a single omnibus C# test step into four named Linux stages —Test (Unit)/Test (Integration)/Test (Contract)/Test (F#)— plus a WindowsTest (Non-Docker)stage (TEST-05). Coverlet → Cobertura → ReportGenerator pipeline publishes the HTML coverage report as a CI artifact on every run (TEST-06).Sources/Tests/coverlet.runsettingsexcludes generated code per D-27.Key files:
.github/workflows/build.yml,Sources/Tests/coverlet.runsettings.Plan 02-04 — Domain Unit Tests (Wave 2)
83 Unit-tagged tests exercise
ProjectV.Models(exception conventions, value-object invariants,BasicInforound-trips) andProjectV.Appraisers(rating-computation accuracy on the canonicalAppraiser<T>+IAppraisal<T>compositions andAppraisersManager). Closes the Domain-layer slice of TEST-02.Key files:
Sources/Tests/ProjectV.Models.Tests/**,Sources/Tests/ProjectV.Appraisers.Tests/**.Plan 02-05 — Core Application Unit Tests (Wave 2)
New
ProjectV.Core.Testsproject coversShell+ShellBuildercomposition +CommunicationServiceClient+ Polly retry policies. New mock builders forShell,ServiceClient, andIO/Crawlersmanagers added toTests.Shared. Closes a substantial portion of the TEST-03 Application-layer slice.Key files:
Sources/Tests/ProjectV.Core.Tests/**,Sources/Tests/ProjectV.Tests.Shared/Helpers/Mocks/{Core,Net,IO,Crawlers}/**.Plan 02-06 — Pipeline: Dataflow + Crawlers Tests (Wave 2)
DataflowPipelineintegration coverage + 3 crawler unit suites. Split off from the original 02-06 per code-review WARNING-03 (file count). 3 critical-path rows of the TEST-02 / TEST-03 inventory close here.Key files:
Sources/Tests/ProjectV.Core.Tests/DataPipeline/**,Sources/Tests/ProjectV.Tests.Shared/Helpers/Mocks/Crawlers/**.Plan 02-07 — Pipeline: Executors + IO Tests (Wave 2)
Unit-test slice for
Executors+InputProcessing+OutputProcessing. Sibling split of 02-06 — closes the executors + input + output critical-path rows of TEST-03.Key files:
Sources/Tests/ProjectV.Core.Tests/{Executors,InputProcessing,OutputProcessing}/**.Plan 02-08 — Contract Tests via WireMock.Net (Wave 2)
Three new contract-tier test projects (
ProjectV.TmdbService.Tests/OmdbService.Tests/SteamService.Tests) exercise the realTMDbLib/OMDbApiNet/SteamWebApiLibHTTP pipelines against in-process WireMock.Net stubs fed from 7 recorded JSON fixtures. No live API calls; per-adapter contract assertions on URL shape, query params, and response decoding. Closes the contract slice of TEST-04.Key files:
Sources/Tests/ProjectV.TmdbService.Tests/**,Sources/Tests/ProjectV.OmdbService.Tests/**,Sources/Tests/ProjectV.SteamService.Tests/**,Sources/Tests/Fixtures/{Tmdb,Omdb,Steam}/*.json.Plan 02-09 — DAL Integration Tests (Testcontainers + PostgreSQL) (Wave 2)
First DB integration-test slice: a single-container Testcontainers Postgres fixture, EF Core 9 design-time factory, initial migrations, and a generator + mock-builder layer (
DbCollection/TestDbHelper/ DAL generators). Rule 1 production-model fixes landed in the same plan to make persistence round-trip cleanly. Closes the DB slice of TEST-04.Key files:
Sources/Libraries/ProjectV.DataAccessLayer/Migrations/**,Sources/Tests/ProjectV.DataAccessLayer.Tests/**,Sources/Tests/ProjectV.Tests.Shared/Helpers/Persistence/**.Plan 02-10 — JWT Integration Tests (Wave 2)
JWT runtime path on
ProjectV.CommunicationWebServicecovered viaWebApplicationFactory+ aTestJwtHelper. NewWebApiBaseTestbase class lands inTests.Shared. Closes the runtime JWT pending decision from Phase 1 (D-31).Key files:
Sources/Tests/ProjectV.CommunicationWebService.Tests/**,Sources/Tests/ProjectV.Tests.Shared/Helpers/WebApi/{TestWebApplicationFactory,TestJwtHelper,WebApiBaseTest}.cs.Plan 02-11 — Telegram Webhook Integration Tests (Wave 3)
Webhook half of D-15 (Telegram full integration coverage).
TestWebApplicationFactoryextended withITelegramBotClient+ICommunicationServiceClientstubs and a newTestTelegramBotClientBuilderadded toTests.Shared. Absorbs review WARNING-06 Option A on factory composition.Key files:
Sources/Tests/ProjectV.TelegramBotWebService.Tests/Webhook/**,Sources/Tests/ProjectV.Tests.Shared/Helpers/WebApi/TestWebApplicationFactory.cs,Sources/Tests/ProjectV.Tests.Shared/Helpers/Mocks/Telegram/TestTelegramBotClientBuilder.cs.Plan 02-12 — Telegram Polling Integration Tests (Wave 4)
Polling half of D-15.
ProjectV.TelegramBotWebServicepolling path (PoolingProcessor) integration-tested with the same factory composition as 02-11.Key files:
Sources/Tests/ProjectV.TelegramBotWebService.Tests/Polling/**.Plan 02-13 — Review Followups (Wave 5 — gap-closure)
Closed four findings carried out of the Phase 2 code-review loop into a single atomic plan:
[ModuleInitializer]dedup; NLog test initializer hoisted toTests.Shared.FakeHttpMessageHandlerhoisted toTests.Shared(was duplicated across contract suites).RefreshTokenDbInfoctor guards againstGuid.Empty.FindByUserIdAsync(withTokenHashassertion per iter-3).Three subsequent doc-only iterations (iter-1..3) tightened comments in EF + NLog test infrastructure per the audit-team review.
Key files:
Sources/Tests/ProjectV.Tests.Shared/Helpers/Logging/**,Sources/Tests/ProjectV.Tests.Shared/Helpers/Net/FakeHttpMessageHandler.cs,Sources/Libraries/ProjectV.DataAccessLayer/Models/Tokens/RefreshTokenDbInfo.cs,Sources/Tests/ProjectV.DataAccessLayer.Tests/Repositories/Tokens/RefreshTokenRepositoryTests.cs.Requirements Addressed
Shell/ShellBuilder/DataflowPipeline/ExecutorsTest Plan
dotnet restore Sources && dotnet build Sources/ProjectV.sln -c Release -p:Platform=x64→ 0 warnings / 0 errors.dotnet format Sources/ProjectV.sln --severity warn --verify-no-changes→ exit 0.dotnet test Sources/ProjectV.sln -c Release -p:Platform=x64 --no-build --filter "Category=Unit"→ pass locally; gated in CI asTest (Unit).dotnet test Sources/ProjectV.sln -c Release -p:Platform=x64 --no-build --filter "Category=Integration"→ pass; gated in CI asTest (Integration).dotnet test Sources/ProjectV.sln -c Release -p:Platform=x64 --no-build --filter "Category=Contract"→ pass; gated in CI asTest (Contract).dotnet test Sources/Tests/ProjectV.ContentDirectories.Tests/ProjectV.ContentDirectories.Tests.fsproj -c Release -p:Platform=x64 --no-build→ 9 / 9 pass; gated in CI asTest (F#).windows-latest— unit + contract only; integration suite skipped on Windows by design (Testcontainers not in scope for Windows runner).coverage-reportartifact. No hard coverage gate; visibility only (TEST-06).master, 0 behind. All review-loop findings closed via Plan 02-13.Key Decisions
Material decisions captured during Phase 2 planning + execution (full ledger in the plan artifacts):
ProjectV.Tests.Sharedis a single shared library housing base test classes, generators, and mock builders. Every test project underSources/Tests/references it; no per-project duplication.Sources/Tests/Directory.Build.props, conditioned on$(MSBuildProjectExtension) == '.csproj'so the F# project keeps its self-contained Unquote stack untouched (D-04).ProjectV.ContentDirectories.Tests.fsprojretains its own stack and runs in its own CI stage. Reaffirmed from Phase 1.Sources/Tests/Fixtures/{Tmdb,Omdb,Steam}/as checked-in recorded JSON. No live API calls in CI.Sources/Tests/coverlet.runsettings; covers generated EF migrations + auto-generated NSwag clients.5e8c3af/ae883e1/f0b1602).FakeHttpMessageHandlerhoisted toTests.Sharedrather than left duplicated per project (Plan 02-13).RefreshTokenDbInfoctor rejectsGuid.Emptyto make the invariant explicit at the domain layer (Plan 02-13).Out of Scope
Tracked but explicitly deferred:
Shell/ShellBuilder/ Dataflow pipeline — additive phase, no redesign.Closing Issues
Closes #341