diff --git a/.github/workflows/bc-test-from-source.yml b/.github/workflows/bc-test-from-source.yml index 5fbcf002..f01bceb6 100644 --- a/.github/workflows/bc-test-from-source.yml +++ b/.github/workflows/bc-test-from-source.yml @@ -320,6 +320,20 @@ on: required: false type: string default: "auto" + test_runner_codeunit_id: + description: >- + Run the test suite through this test runner codeunit, for example + 130451 "Test Runner - Isol. Disabled" or a test app's own codeunit + with Subtype = TestRunner (BcContainerHelper's -testRunnerCodeunitId). + Blank (default) uses 130450 "Test Runner - Isol. Codeunit", + BcContainerHelper's default. Only the websocket runner + runs tests through a runner codeunit, so when this is set every + codeunit goes to it: test_runner=auto uses websocket for the whole + run, and test_runner=altool fails. Results come back in the same JUnit + format (issue #193). + required: false + type: string + default: "" altool_transport: description: >- How the altool leg of test_runner=auto reaches the TestRunnerHub. @@ -833,7 +847,7 @@ jobs: # - test_runner=auto on an older version → skip the install; the # run step's live probe will route to the websocket runner - name: Install AL dotnet tool (altool test runner) - if: inputs.test_runner != 'websocket' + if: inputs.test_runner != 'websocket' && inputs.test_runner_codeunit_id == '' run: | set -e BC_MAJOR=$(echo "${{ inputs.bc_version }}" | cut -d. -f1) @@ -1322,6 +1336,18 @@ jobs: # "altool" without server support fails inside the script with # a clear error instead of silently degrading. EFFECTIVE_RUNNER="${{ inputs.test_runner }}" + # A runner codeunit only exists on the websocket path; the hub runs + # tests without one (issue #193). + RUNNER_CODEUNIT_ARGS=() + if [ -n "${{ inputs.test_runner_codeunit_id }}" ]; then + RUNNER_CODEUNIT_ARGS=(--test-runner-codeunit "${{ inputs.test_runner_codeunit_id }}") + if [ "$EFFECTIVE_RUNNER" = "altool" ]; then + echo "ERROR: test_runner_codeunit_id needs the websocket runner; test_runner=altool runs tests without a runner codeunit" + exit 1 + fi + EFFECTIVE_RUNNER="websocket" + echo "test_runner_codeunit_id=${{ inputs.test_runner_codeunit_id }}: using the websocket runner for every codeunit" + fi if [ "$EFFECTIVE_RUNNER" = "auto" ]; then if [ -x "$HOME/.dotnet/tools/al" ] \ && python3 scripts/run-tests-altool.py --probe --port "$BC_DEV_PORT"; then @@ -1406,6 +1432,7 @@ jobs: --base-url "http://localhost:${BC_ODATA_PORT}/BC" \ --dev-url "http://localhost:${BC_DEV_PORT}/BC/dev" \ "${COMPANY_ARGS[@]}" \ + "${RUNNER_CODEUNIT_ARGS[@]}" \ --timeout ${{ inputs.test_timeout_minutes }} 2>&1 | tee "$tmplog" fi RUN_TESTS_RC=${PIPESTATUS[0]} diff --git a/.github/workflows/bc-test-prebuilt.yml b/.github/workflows/bc-test-prebuilt.yml index 949f3dba..d2ef3d7d 100644 --- a/.github/workflows/bc-test-prebuilt.yml +++ b/.github/workflows/bc-test-prebuilt.yml @@ -124,6 +124,17 @@ on: "50000..50100,130450,200000..210000" mixed required: true type: string + test_runner_codeunit_id: + description: >- + Run the test suite through this test runner codeunit, for example + 130451 "Test Runner - Isol. Disabled" or a test app's own codeunit + with Subtype = TestRunner (BcContainerHelper's -testRunnerCodeunitId). + Blank (default) uses 130450 "Test Runner - Isol. Codeunit", + BcContainerHelper's default. Results come back in the same JUnit + format (issue #193). + required: false + type: string + default: "" extra_apps: description: >- Newline-separated names of Microsoft apps from the BC artifact to @@ -479,6 +490,10 @@ jobs: if [ -n "${TEST_COMPANY:-}" ]; then COMPANY_ARGS=(--company "$TEST_COMPANY") fi + RUNNER_CODEUNIT_ARGS=() + if [ -n "${{ inputs.test_runner_codeunit_id }}" ]; then + RUNNER_CODEUNIT_ARGS=(--test-runner-codeunit "${{ inputs.test_runner_codeunit_id }}") + fi # Per-app JUnit XML — see bc-test-from-source.yml for the rationale. mkdir -p ../build for app in $TEST_APP_FILES; do @@ -494,6 +509,7 @@ jobs: --base-url "http://localhost:${BC_ODATA_PORT}/BC" \ --dev-url "http://localhost:${BC_DEV_PORT}/BC/dev" \ "${COMPANY_ARGS[@]}" \ + "${RUNNER_CODEUNIT_ARGS[@]}" \ --timeout 30 2>&1 | tee "$tmplog" # `tee` always exits 0; capture run-tests.sh's real exit code. RUN_TESTS_RC=${PIPESTATUS[0]} diff --git a/CLAUDE.md b/CLAUDE.md index 9651e0cb..17f9b061 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -81,6 +81,19 @@ moved. part safe to re-post. `--timeout` is scaled by the number of parts, because it used to be a per-part budget and is now a whole-run one. +**The runner codeunit can be chosen** (issue #193). The suite runs through page +130455 with runner 130450 "Test Runner - Isol. Codeunit", BcContainerHelper's +default. `--test-runner-codeunit ` on `run-tests.sh`, and the +`test_runner_codeunit_id` input on both reusable workflows, set the suite's +"Test Runner Id" instead (field `TestRunnerId` on table 99903, Test Runner +Extension 3.3.0.0). That covers 130451 and a test app's own `Subtype = +TestRunner` codeunit (BcContainerHelper's `-testRunnerCodeunitId`). The hub has +no runner codeunit, so with the input set `bc-test-from-source.yml` sends every +codeunit to the websocket runner, and `test_runner=altool` fails. A runner id +that does not exist fails the run with BC's "An object with that ID does not +exist". The extension cannot check it earlier: it has no platform dependency, +so `AllObj` is not visible to it. + **EXPERIMENTAL altool runner (BC 28+ only):** `scripts/run-tests-altool.py` runs tests through the AL dotnet tool's native `al runtests` command (Microsoft.Dynamics.BusinessCentral.Development.Tools, 18.x prerelease — stable 17.x has no `runtests`), which drives the NST's built-in SignalR hub at `/dev/TestRunnerHub`. No TestRunnerExtension, no OData suite, no WebSocket emulation — the server pushes per-method results (status, output, duration) over the hub. Requires the server to advertise Dev API 7.0 (`GET /BC/dev/metadata`), which only exists in BC 28.0+. Caveats: tests do NOT run under an AL test runner codeunit (no AI tests, no test-runner setup/teardown events, isolation from `RequiredTestIsolation`, default Codeunit) — so Microsoft BCApps suites may behave differently than under `run-tests.sh`; the test app must already be published+installed (the script doesn't publish). The reusable workflow's `test_runner` input defaults to `auto`: after BC is healthy it probes `GET /dev/metadata` (via `run-tests-altool.py --probe`, exit 0 = Dev API ≥ 7.0) and uses the altool runner when supported, falling back to the websocket runner otherwise — so 27.x legs and consumers on older versions keep working unchanged. `websocket` forces the legacy flow; `altool` forces the hub and fails hard when unsupported (the regression-detection mode). `altool_version` pins the dotnet tool. Auth comes from `BC_SERVER_USERNAME`/`BC_SERVER_PASSWORD` env vars, which the script sets from `--auth`. The script's stdout deliberately prints the same `N total, P passed, F failed` and `Test codeunits: ...` lines the workflow parser greps — keep that contract if you touch either side. ### Editing the startup hook diff --git a/extensions/TestRunnerExtension/TestRunnerExtension.app b/extensions/TestRunnerExtension/TestRunnerExtension.app index ef5e89c0..99da2180 100644 Binary files a/extensions/TestRunnerExtension/TestRunnerExtension.app and b/extensions/TestRunnerExtension/TestRunnerExtension.app differ diff --git a/extensions/TestRunnerExtension/app.json b/extensions/TestRunnerExtension/app.json index 833f63ec..cca59c26 100644 --- a/extensions/TestRunnerExtension/app.json +++ b/extensions/TestRunnerExtension/app.json @@ -2,7 +2,7 @@ "id": "bbbbbbbb-bbbb-bbbb-bbbb-bbbbbbbbbbbb", "name": "Test Runner Extension", "publisher": "ALDirectCompile", - "version": "3.2.0.0", + "version": "3.3.0.0", "runtime": "16.0", "target": "OnPrem", "dependencies": [ diff --git a/extensions/TestRunnerExtension/src/RunnerTable.al b/extensions/TestRunnerExtension/src/RunnerTable.al index 87f0856f..8b4613d2 100644 --- a/extensions/TestRunnerExtension/src/RunnerTable.al +++ b/extensions/TestRunnerExtension/src/RunnerTable.al @@ -115,6 +115,18 @@ table 99903 "Codeunit Run Request" Caption = 'Last Execution (UTC)'; DataClassification = SystemMetadata; } + + /// + /// The test runner codeunit SetupSuite and SetupSuiteAppend give the DEFAULT suite + /// (issue #193). 0 means 130450 "Test Runner - Isol. Codeunit", BcContainerHelper's + /// default. Set it to run the suite through another runner, such as a test app's own + /// codeunit with Subtype = TestRunner (BcContainerHelper's -testRunnerCodeunitId). + /// + field(7; TestRunnerId; Integer) + { + Caption = 'Test Runner Id'; + DataClassification = SystemMetadata; + } } keys @@ -179,6 +191,7 @@ page 99902 "Codeunit Run Requests" field(Id; Rec.Id) { Editable = false; } field(CodeunitId; Rec.CodeunitId) { } field(CodeunitIds; Rec.CodeunitIds) { } + field(TestRunnerId; Rec.TestRunnerId) { } field(Status; Rec.Status) { Editable = false; } field(LastResult; Rec.LastResult) { Editable = false; } field(LastExecutionUTC; Rec.LastExecutionUTC) { Editable = false; } @@ -222,7 +235,7 @@ page 99902 "Codeunit Run Requests" // 130450 = TestIsolation=Codeunit (fresh session per codeunit) ← CORRECT // 130451 = TestIsolation=Disabled (no isolation) ← WRONG, causes cascading failures SuiteRunner.InitSuite('DEFAULT'); - OverrideSuiteRunner('DEFAULT', 130450); + OverrideSuiteRunner('DEFAULT', SuiteRunnerId()); exit(AddRequestedCodeunits(SuiteRunner, 'Suite ready') > 0); end; @@ -249,7 +262,7 @@ page 99902 "Codeunit Run Requests" exit(false); SuiteRunner.InitSuiteKeep('DEFAULT'); - OverrideSuiteRunner('DEFAULT', 130450); + OverrideSuiteRunner('DEFAULT', SuiteRunnerId()); exit(AddRequestedCodeunits(SuiteRunner, 'Suite extended') > 0); end; @@ -334,6 +347,14 @@ page 99902 "Codeunit Run Requests" exit(true); end; + /// TestRunnerId, or 130450 when it is 0. + local procedure SuiteRunnerId(): Integer + begin + if Rec.TestRunnerId = 0 then + exit(130450); + exit(Rec.TestRunnerId); + end; + local procedure OverrideSuiteRunner(SuiteName: Code[10]; RunnerId: Integer) var ALTestSuite: Record "AL Test Suite"; diff --git a/scripts/run-tests.sh b/scripts/run-tests.sh index 978c18b9..5fb21425 100755 --- a/scripts/run-tests.sh +++ b/scripts/run-tests.sh @@ -55,6 +55,7 @@ REPO_DIR="$(cd "$SCRIPT_DIR/.." && pwd)" TEST_RUNNER_APP="" JUNIT_OUTPUT="" RENEW_CLIENT_CONTEXT=0 +TEST_RUNNER_CODEUNIT="" show_help() { cat <<'HELPEOF' @@ -87,6 +88,12 @@ Options: the whole suite, so SingleInstance codeunit state carries across test codeunits the way it does on a real tier. Turn this on only to reproduce the pre-#65 behaviour. + --test-runner-codeunit + Run the suite through this test runner codeunit, for + example 130451 or a test app's own codeunit with + Subtype = TestRunner (BcContainerHelper's + -testRunnerCodeunitId). Default: 130450 "Test Runner - + Isol. Codeunit", BcContainerHelper's default. -h, --help Show this help Examples: @@ -115,12 +122,19 @@ while [[ $# -gt 0 ]]; do --disabled-tests) DISABLED_TESTS_DIR="$2"; shift 2;; --junit-output) JUNIT_OUTPUT="$2"; shift 2;; --renew-client-context) RENEW_CLIENT_CONTEXT=1; shift;; + --test-runner-codeunit) TEST_RUNNER_CODEUNIT="$2"; shift 2;; --host|--test-runner|--suite-name|--codeunit-timeout|--extension-id|--sql-password) shift 2;; *) echo "Unknown option: $1 (try --help)"; exit 1;; esac done +if [ -n "$TEST_RUNNER_CODEUNIT" ] && ! [[ "$TEST_RUNNER_CODEUNIT" =~ ^[1-9][0-9]*$ ]]; then + echo "ERROR: --test-runner-codeunit must be a codeunit id, got '$TEST_RUNNER_CODEUNIT'" + exit 1 +fi + echo "=== BC Test Runner ===" +[ -n "$TEST_RUNNER_CODEUNIT" ] && echo "Test runner codeunit: $TEST_RUNNER_CODEUNIT" # --- Helper --- py3() { env -u PYTHONHOME -u PYTHONPATH python3 "$@"; } @@ -440,11 +454,17 @@ REQUEST_ID="" setup_suite_chunk() { local chunk="$1" label="$2" action="${3:-setupSuite}" echo -n "Setting up test suite${label}... " - local create_body create_http + local create_body create_http request_json + # TestRunnerId arrived in Test Runner Extension 3.3.0.0 (issue #193). Only send it when + # asked for, so the default run still works against an older image. + request_json="{\"CodeunitIds\": \"$chunk\"}" + if [ -n "$TEST_RUNNER_CODEUNIT" ]; then + request_json="{\"CodeunitIds\": \"$chunk\", \"TestRunnerId\": $TEST_RUNNER_CODEUNIT}" + fi create_body=$(mktemp) create_http=$(curl -s -o "$create_body" -w "%{http_code}" --max-time 15 -u "$AUTH" -X POST \ -H "Content-Type: application/json" \ - -d "{\"CodeunitIds\": \"$chunk\"}" \ + -d "$request_json" \ "${API_BASE}/codeunitRunRequests" 2>/dev/null) REQUEST_ID=$(py3 -c "import sys,json; print(json.load(sys.stdin)['Id'])" < "$create_body" 2>/dev/null || true) @@ -453,7 +473,7 @@ setup_suite_chunk() { echo "" echo "ERROR: POST ${API_BASE}/codeunitRunRequests" echo " HTTP code: $create_http" - echo " Request body: {\"CodeunitIds\": \"$chunk\"}" + echo " Request body: $request_json" echo " CodeunitIds length: ${#chunk} chars (BC's field limit is 2048)" echo " Response body:" sed 's/^/ /' "$create_body" @@ -466,6 +486,10 @@ setup_suite_chunk() { echo " POST in this BC version (the auto-detection picked an endpoint that" echo " accepts GET on this path but not POST)." echo " - The request body schema doesn't match the page's bound action." + if [ -n "$TEST_RUNNER_CODEUNIT" ]; then + echo " - The bc-runner image is older than this script: TestRunnerId" + echo " (--test-runner-codeunit) needs Test Runner Extension 3.3.0.0." + fi echo " - Application_StringExceededLength above means the part is still" echo " too long — lower BC_CODEUNIT_IDS_MAX_CHARS (currently" echo " ${CODEUNIT_IDS_MAX_CHARS})."