| // Copyright 2020 The Go Authors. All rights reserved. |
| // Use of this source code is governed by a BSD-style |
| // license that can be found in the LICENSE file. |
| |
| package workflow_test |
| |
| import ( |
| "context" |
| "errors" |
| "fmt" |
| "reflect" |
| "strings" |
| "sync/atomic" |
| "testing" |
| "time" |
| |
| "github.com/google/go-cmp/cmp" |
| "github.com/google/go-cmp/cmp/cmpopts" |
| "github.com/google/uuid" |
| wf "golang.org/x/build/internal/workflow" |
| ) |
| |
| func TestTrivial(t *testing.T) { |
| echo := func(ctx context.Context, arg string) (string, error) { |
| return arg, nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| wf.Task1(wd, "echo", echo, wf.Const("hello world")) |
| greeting := wf.Task1(wd, "echo", echo, wf.Const("hello world")) |
| wf.Output(wd, "greeting", greeting) |
| |
| w := startWorkflow(t, wd, nil) |
| outputs := runWorkflow(t, w, nil) |
| if got, want := outputs["greeting"], "hello world"; got != want { |
| t.Errorf("greeting = %q, want %q", got, want) |
| } |
| } |
| |
| func TestDependency(t *testing.T) { |
| var actionRan, checkRan bool |
| action := func(ctx context.Context) error { |
| actionRan = true |
| return nil |
| } |
| checkAction := func(ctx context.Context) (string, error) { |
| if !actionRan { |
| return "", fmt.Errorf("prior action didn't run") |
| } |
| checkRan = true |
| return "", nil |
| } |
| hi := func(ctx context.Context) (string, error) { |
| if !actionRan || !checkRan { |
| return "", fmt.Errorf("either action (%v) or checkAction (%v) didn't run", actionRan, checkRan) |
| } |
| return "hello world", nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| firstDep := wf.Action0(wd, "first action", action) |
| secondDep := wf.Task0(wd, "check action", checkAction, wf.After(firstDep)) |
| wf.Output(wd, "greeting", wf.Task0(wd, "say hi", hi, wf.After(secondDep))) |
| |
| w := startWorkflow(t, wd, nil) |
| outputs := runWorkflow(t, w, nil) |
| if got, want := outputs["greeting"], "hello world"; got != want { |
| t.Errorf("greeting = %q, want %q", got, want) |
| } |
| } |
| |
| func TestDependencyError(t *testing.T) { |
| action := func(ctx context.Context) error { |
| return fmt.Errorf("hardcoded error") |
| } |
| task := func(ctx context.Context) (string, error) { |
| return "", fmt.Errorf("unexpected error") |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| dep := wf.Action0(wd, "failing action", action) |
| wf.Output(wd, "output", wf.Task0(wd, "task", task, wf.After(dep))) |
| w := startWorkflow(t, wd, nil) |
| if got, want := runToFailure(t, w, nil, "failing action"), "hardcoded error"; got != want { |
| t.Errorf("got error %q, want %q", got, want) |
| } |
| } |
| |
| func TestSub(t *testing.T) { |
| hi := func(ctx context.Context) (string, error) { |
| return "hi", nil |
| } |
| concat := func(ctx context.Context, s1, s2 string) (string, error) { |
| return s1 + " " + s2, nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| sub1 := wd.Sub("sub1") |
| g1 := wf.Task0(sub1, "Greeting", hi) |
| sub2 := wd.Sub("sub2") |
| g2 := wf.Task0(sub2, "Greeting", hi) |
| wf.Output(wd, "result", wf.Task2(wd, "Concatenate", concat, g1, g2)) |
| |
| w := startWorkflow(t, wd, nil) |
| outputs := runWorkflow(t, w, nil) |
| if got, want := outputs["result"], "hi hi"; got != want { |
| t.Errorf("result = %q, want %q", got, want) |
| } |
| } |
| |
| func TestSplitJoin(t *testing.T) { |
| echo := func(ctx context.Context, arg string) (string, error) { |
| return arg, nil |
| } |
| appendInt := func(ctx context.Context, s string, i int) (string, error) { |
| return fmt.Sprintf("%v%v", s, i), nil |
| } |
| join := func(ctx context.Context, s []string) (string, error) { |
| return strings.Join(s, ","), nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| in := wf.Task1(wd, "echo", echo, wf.Const("string #")) |
| add1 := wf.Task2(wd, "add 1", appendInt, in, wf.Const(1)) |
| add2 := wf.Task2(wd, "add 2", appendInt, in, wf.Const(2)) |
| both := wf.Slice(add1, add2) |
| out := wf.Task1(wd, "join", join, both) |
| wf.Output(wd, "strings", out) |
| |
| w := startWorkflow(t, wd, nil) |
| outputs := runWorkflow(t, w, nil) |
| if got, want := outputs["strings"], "string #1,string #2"; got != want { |
| t.Errorf("joined output = %q, want %q", got, want) |
| } |
| } |
| |
| func TestParallelism(t *testing.T) { |
| // block1 and block2 block until they're both running. |
| chan1, chan2 := make(chan bool, 1), make(chan bool, 1) |
| block1 := func(ctx context.Context) (string, error) { |
| chan1 <- true |
| select { |
| case <-chan2: |
| case <-ctx.Done(): |
| } |
| return "", ctx.Err() |
| } |
| block2 := func(ctx context.Context) (string, error) { |
| chan2 <- true |
| select { |
| case <-chan1: |
| case <-ctx.Done(): |
| } |
| return "", ctx.Err() |
| } |
| wd := wf.New(wf.ACL{}) |
| out1 := wf.Task0(wd, "block #1", block1) |
| out2 := wf.Task0(wd, "block #2", block2) |
| wf.Output(wd, "out1", out1) |
| wf.Output(wd, "out2", out2) |
| |
| w := startWorkflow(t, wd, nil) |
| runWorkflow(t, w, nil) |
| } |
| |
| func TestParameters(t *testing.T) { |
| echo := func(ctx context.Context, arg string) (string, error) { |
| return arg, nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| param1 := wf.Param(wd, wf.ParamDef[string]{Name: "param1"}) |
| param2 := wf.Param(wd, wf.ParamDef[string]{Name: "param2"}) |
| out1 := wf.Task1(wd, "echo 1", echo, param1) |
| out2 := wf.Task1(wd, "echo 2", echo, param2) |
| wf.Output(wd, "out1", out1) |
| wf.Output(wd, "out2", out2) |
| |
| w := startWorkflow(t, wd, map[string]interface{}{"param1": "#1", "param2": "#2"}) |
| outputs := runWorkflow(t, w, nil) |
| if want := map[string]interface{}{"out1": "#1", "out2": "#2"}; !reflect.DeepEqual(outputs, want) { |
| t.Errorf("outputs = %#v, want %#v", outputs, want) |
| } |
| |
| t.Run("CountMismatch", func(t *testing.T) { |
| _, err := wf.Start(wd, map[string]interface{}{"param1": "#1"}) |
| if err == nil { |
| t.Errorf("wf.Start didn't return an error despite a parameter count mismatch") |
| } |
| }) |
| t.Run("NameMismatch", func(t *testing.T) { |
| _, err := wf.Start(wd, map[string]interface{}{"paramA": "#1", "paramB": "#2"}) |
| if err == nil { |
| t.Errorf("wf.Start didn't return an error despite a parameter name mismatch") |
| } |
| }) |
| t.Run("TypeMismatch", func(t *testing.T) { |
| _, err := wf.Start(wd, map[string]interface{}{"param1": "#1", "param2": 42}) |
| if err == nil { |
| t.Errorf("wf.Start didn't return an error despite a parameter type mismatch") |
| } |
| }) |
| } |
| |
| // Test that passing wf.Parameter{...} directly to Definition.Task would be a build-time error. |
| // Parameters need to be registered via the Definition.Parameter method. |
| func TestParameterValue(t *testing.T) { |
| var p interface{} = wf.ParamDef[int]{} |
| if _, ok := p.(wf.Value[int]); ok { |
| t.Errorf("Parameter unexpectedly implements Value; it intentionally tries not to reduce possible API misuse") |
| } |
| } |
| |
| func TestExpansion(t *testing.T) { |
| first := func(_ context.Context) (string, error) { |
| return "hey", nil |
| } |
| second := func(_ context.Context) (string, error) { |
| return "there", nil |
| } |
| third := func(_ context.Context) (string, error) { |
| return "friend", nil |
| } |
| join := func(_ context.Context, args []string) (string, error) { |
| return strings.Join(args, " "), nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| v1 := wf.Task0(wd, "first", first) |
| v2 := wf.Task0(wd, "second", second) |
| wf.Output(wd, "second", v2) |
| joined := wf.Expand1(wd, "add a task", func(wd *wf.Definition, arg string) (wf.Value[string], error) { |
| v3 := wf.Task0(wd, "third", third) |
| // v1 is resolved before the expansion runs, v2 and v3 are dependencies |
| // created outside and inside the epansion. |
| return wf.Task1(wd, "join", join, wf.Slice(wf.Const(arg), v2, v3)), nil |
| }, v1) |
| wf.Output(wd, "final value", joined) |
| |
| w := startWorkflow(t, wd, nil) |
| outputs := runWorkflow(t, w, nil) |
| if got, want := outputs["final value"], "hey there friend"; got != want { |
| t.Errorf("joined output = %q, want %q", got, want) |
| } |
| } |
| |
| func TestResumeExpansion(t *testing.T) { |
| counter := 0 |
| succeeds := func(ctx *wf.TaskContext) (string, error) { |
| counter++ |
| return "", nil |
| } |
| wd := wf.New(wf.ACL{}) |
| result := wf.Expand0(wd, "expand", func(wd *wf.Definition) (wf.Value[string], error) { |
| return wf.Task0(wd, "succeeds", succeeds), nil |
| }) |
| wf.Output(wd, "result", result) |
| |
| storage := &mapListener{Listener: &verboseListener{t}} |
| w := startWorkflow(t, wd, nil) |
| runWorkflow(t, w, storage) |
| resumed, err := wf.Resume(wd, &wf.WorkflowState{ID: w.ID}, storage.states[w.ID]) |
| if err != nil { |
| t.Fatal(err) |
| } |
| runWorkflow(t, resumed, nil) |
| if counter != 1 { |
| t.Errorf("task ran %v times, wanted 1", counter) |
| } |
| } |
| |
| func TestRetryExpansion(t *testing.T) { |
| counter := 0 |
| wd := wf.New(wf.ACL{}) |
| out := wf.Expand0(wd, "expand", func(wd *wf.Definition) (wf.Value[string], error) { |
| counter++ |
| if counter == 1 { |
| return nil, fmt.Errorf("first try fail") |
| } |
| return wf.Task0(wd, "hi", func(_ context.Context) (string, error) { |
| return "", nil |
| }), nil |
| }) |
| wf.Output(wd, "out", out) |
| |
| w := startWorkflow(t, wd, nil) |
| retry := func(string) { |
| go func() { |
| w.RetryTask(context.Background(), "expand") |
| }() |
| } |
| listener := &errorListener{ |
| taskName: "expand", |
| callback: retry, |
| Listener: &verboseListener{t}, |
| } |
| runWorkflow(t, w, listener) |
| if counter != 2 { |
| t.Errorf("task ran %v times, wanted 2", counter) |
| } |
| } |
| |
| func TestManualRetry(t *testing.T) { |
| counter := 0 |
| needsRetry := func(ctx *wf.TaskContext) (string, error) { |
| ctx.DisableRetries() |
| counter++ |
| if counter == 1 { |
| return "", fmt.Errorf("counter %v too low", counter) |
| } |
| return "hi", nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| wf.Output(wd, "result", wf.Task0(wd, "needs retry", needsRetry)) |
| |
| w := startWorkflow(t, wd, nil) |
| |
| retry := func(string) { |
| go func() { |
| w.RetryTask(context.Background(), "needs retry") |
| }() |
| } |
| listener := &errorListener{ |
| taskName: "needs retry", |
| callback: retry, |
| Listener: &verboseListener{t}, |
| } |
| runWorkflow(t, w, listener) |
| if counter != 2 { |
| t.Errorf("task ran %v times, wanted 2", counter) |
| } |
| } |
| |
| // Test that manual retry works on tasks that come from different expansions. |
| // |
| // This is similar to how the Go minor release workflow plans builders for |
| // both releases. It previously failed due to expansions racing with with other, |
| // leading to "unknown task" errors when retrying. See go.dev/issue/70249. |
| func TestManualRetryMultipleExpansions(t *testing.T) { |
| // Create two sub-workflows, each one with an expansion that adds one work task. |
| // The work tasks fail on the first try, and require being successfully restarted |
| // for the workflow to complete. |
| var counters, retried [2]int |
| wd := wf.New(wf.ACL{}) |
| sub1 := wd.Sub("sub1") |
| sub2 := wd.Sub("sub2") |
| for i, wd := range []*wf.Definition{sub1, sub2} { |
| out := wf.Expand0(wd, fmt.Sprintf("expand %d", i+1), func(wd *wf.Definition) (wf.Value[string], error) { |
| return wf.Task0(wd, fmt.Sprintf("work %d", i+1), func(ctx *wf.TaskContext) (string, error) { |
| ctx.DisableRetries() |
| counters[i]++ |
| if counters[i] == 1 { |
| return "", fmt.Errorf("first try fail") |
| } |
| return "", nil |
| }), nil |
| }) |
| wf.Output(wd, "out", out) |
| } |
| |
| w := startWorkflow(t, wd, nil) |
| listener := &errorListener{ |
| taskName: "work 1", |
| callback: func(string) { |
| go func() { |
| retried[0]++ |
| err := w.RetryTask(context.Background(), "work 1") |
| if err != nil { |
| t.Errorf(`RetryTask("work 1") failed: %v`, err) |
| } |
| }() |
| }, |
| Listener: &errorListener{ |
| taskName: "work 2", |
| callback: func(string) { |
| go func() { |
| retried[1]++ |
| err := w.RetryTask(context.Background(), "work 2") |
| if err != nil { |
| t.Errorf(`RetryTask("work 2") failed: %v`, err) |
| } |
| }() |
| }, |
| Listener: &verboseListener{t}, |
| }, |
| } |
| runWorkflow(t, w, listener) |
| if counters[0] != 2 { |
| t.Errorf("sub1 task ran %v times, wanted 2", counters[0]) |
| } |
| if retried[0] != 1 { |
| t.Errorf("sub1 task was retried %v times, wanted 1", retried[0]) |
| } |
| if counters[1] != 2 { |
| t.Errorf("sub2 task ran %v times, wanted 2", counters[1]) |
| } |
| if retried[1] != 1 { |
| t.Errorf("sub2 task was retried %v times, wanted 1", retried[1]) |
| } |
| } |
| |
| func TestAutomaticRetry(t *testing.T) { |
| counter := 0 |
| needsRetry := func(ctx *wf.TaskContext) (string, error) { |
| if counter < 2 { |
| counter++ |
| return "", fmt.Errorf("counter %v too low", counter) |
| } |
| return "hi", nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| wf.Output(wd, "result", wf.Task0(wd, "needs retry", needsRetry)) |
| |
| w := startWorkflow(t, wd, nil) |
| outputs := runWorkflow(t, w, nil) |
| if got, want := outputs["result"], "hi"; got != want { |
| t.Errorf("result = %q, want %q", got, want) |
| } |
| if counter != 2 { |
| t.Errorf("counter = %v, want 2", counter) |
| } |
| } |
| |
| func TestAutomaticRetryDisabled(t *testing.T) { |
| counter := 0 |
| noRetry := func(ctx *wf.TaskContext) (string, error) { |
| ctx.DisableRetries() |
| counter++ |
| return "", fmt.Errorf("do not pass go") |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| wf.Output(wd, "result", wf.Task0(wd, "no retry", noRetry)) |
| |
| w := startWorkflow(t, wd, nil) |
| if got, want := runToFailure(t, w, nil, "no retry"), "do not pass go"; got != want { |
| t.Errorf("got error %q, want %q", got, want) |
| } |
| if counter != 1 { |
| t.Errorf("task with retries disabled ran %v times, wanted 1", counter) |
| } |
| } |
| |
| func TestWatchdog(t *testing.T) { |
| t.Run("success", func(t *testing.T) { |
| testWatchdog(t, true) |
| }) |
| t.Run("failure", func(t *testing.T) { |
| testWatchdog(t, false) |
| }) |
| } |
| |
| func testWatchdog(t *testing.T, success bool) { |
| defer func(r int, d time.Duration) { |
| wf.MaxRetries = r |
| wf.WatchdogDelay = d |
| }(wf.MaxRetries, wf.WatchdogDelay) |
| wf.MaxRetries = 1 |
| wf.WatchdogDelay = 750 * time.Millisecond |
| |
| maybeLog := func(ctx *wf.TaskContext) (string, error) { |
| select { |
| case <-ctx.Done(): |
| return "", ctx.Err() |
| case <-time.After(500 * time.Millisecond): |
| } |
| if success { |
| ctx.Printf("*snore*") |
| } |
| select { |
| case <-ctx.Done(): |
| return "", ctx.Err() |
| case <-time.After(500 * time.Millisecond): |
| } |
| return "huh? what?", nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| wf.Output(wd, "result", wf.Task0(wd, "sleepy", maybeLog)) |
| |
| w := startWorkflow(t, wd, nil) |
| if success { |
| runWorkflow(t, w, nil) |
| } else { |
| if got, want := runToFailure(t, w, nil, "sleepy"), "assumed hung"; !strings.Contains(got, want) { |
| t.Errorf("got error %q, want %q", got, want) |
| } |
| } |
| } |
| |
| func TestLogging(t *testing.T) { |
| log := func(ctx *wf.TaskContext, arg string) (string, error) { |
| ctx.Printf("logging argument: %v", arg) |
| return arg, nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| out := wf.Task1(wd, "log", log, wf.Const("hey there")) |
| wf.Output(wd, "out", out) |
| |
| logger := &capturingLogger{} |
| listener := &logTestListener{ |
| Listener: &verboseListener{t}, |
| logger: logger, |
| } |
| w := startWorkflow(t, wd, nil) |
| runWorkflow(t, w, listener) |
| if want := []string{"logging argument: hey there"}; !reflect.DeepEqual(logger.lines, want) { |
| t.Errorf("unexpected logging result: got %v, want %v", logger.lines, want) |
| } |
| } |
| |
| type logTestListener struct { |
| wf.Listener |
| logger wf.Logger |
| } |
| |
| func (l *logTestListener) Logger(_ uuid.UUID, _ string) wf.Logger { |
| return l.logger |
| } |
| |
| type capturingLogger struct { |
| lines []string |
| } |
| |
| func (l *capturingLogger) Printf(format string, v ...interface{}) { |
| l.lines = append(l.lines, fmt.Sprintf(format, v...)) |
| } |
| |
| func TestResume(t *testing.T) { |
| // We expect runOnlyOnce to only run once. |
| var runs int64 |
| runOnlyOnce := func(ctx context.Context) (string, error) { |
| atomic.AddInt64(&runs, 1) |
| return "ran", nil |
| } |
| // blockOnce blocks the first time it's called, so that the workflow can be |
| // canceled at its step. |
| block := true |
| blocked := make(chan bool, 1) |
| maybeBlock := func(ctx *wf.TaskContext, _ string) (string, error) { |
| ctx.DisableRetries() |
| if block { |
| blocked <- true |
| <-ctx.Done() |
| return "blocked", ctx.Err() |
| } |
| return "not blocked", nil |
| } |
| wd := wf.New(wf.ACL{}) |
| v1 := wf.Task0(wd, "run once", runOnlyOnce) |
| v2 := wf.Task1(wd, "block", maybeBlock, v1) |
| wf.Output(wd, "output", v2) |
| |
| // Cancel the workflow once we've entered maybeBlock. |
| ctx, cancel := context.WithCancel(context.Background()) |
| go func() { |
| <-blocked |
| cancel() |
| }() |
| w, err := wf.Start(wd, nil) |
| if err != nil { |
| t.Fatal(err) |
| } |
| storage := &mapListener{Listener: &verboseListener{t}} |
| _, err = w.Run(ctx, storage) |
| if !errors.Is(err, context.Canceled) { |
| t.Fatalf("canceled workflow returned error %v, wanted Canceled", err) |
| } |
| storage.assertState(t, w, map[string]*wf.TaskState{ |
| "run once": {Name: "run once", Started: true, Finished: true, Result: "ran"}, |
| "block": {Name: "block", Started: true, Finished: true, Error: "context canceled"}, // We cancelled the workflow before it could save its state. |
| }) |
| |
| block = false |
| wfState := &wf.WorkflowState{ID: w.ID, Params: nil} |
| taskStates := storage.states[w.ID] |
| taskStates["block"] = &wf.TaskState{Name: "block"} |
| w2, err := wf.Resume(wd, wfState, taskStates) |
| if err != nil { |
| t.Fatal(err) |
| } |
| out := runWorkflow(t, w2, storage) |
| if got, want := out["output"], "not blocked"; got != want { |
| t.Errorf("output from maybeBlock was %q, wanted %q", got, want) |
| } |
| if runs != 1 { |
| t.Errorf("runOnlyOnce ran %v times, wanted 1", runs) |
| } |
| storage.assertState(t, w, map[string]*wf.TaskState{ |
| "run once": {Name: "run once", Started: true, Finished: true, Result: "ran"}, |
| "block": {Name: "block", Started: true, Finished: true, Result: "not blocked"}, |
| }) |
| } |
| |
| type badResult struct { |
| unexported string |
| } |
| |
| func TestBadMarshaling(t *testing.T) { |
| greet := func(_ context.Context) (badResult, error) { |
| return badResult{"hi"}, nil |
| } |
| |
| wd := wf.New(wf.ACL{}) |
| wf.Output(wd, "greeting", wf.Task0(wd, "greet", greet)) |
| w := startWorkflow(t, wd, nil) |
| if got, want := runToFailure(t, w, nil, "greet"), "JSON marshaling"; !strings.Contains(got, want) { |
| t.Errorf("got error %q, want %q", got, want) |
| } |
| } |
| |
| type mapListener struct { |
| wf.Listener |
| states map[uuid.UUID]map[string]*wf.TaskState |
| } |
| |
| func (l *mapListener) TaskStateChanged(workflowID uuid.UUID, taskID string, state *wf.TaskState) error { |
| if l.states == nil { |
| l.states = map[uuid.UUID]map[string]*wf.TaskState{} |
| } |
| if l.states[workflowID] == nil { |
| l.states[workflowID] = map[string]*wf.TaskState{} |
| } |
| l.states[workflowID][taskID] = state |
| return l.Listener.TaskStateChanged(workflowID, taskID, state) |
| } |
| |
| func (l *mapListener) assertState(t *testing.T, w *wf.Workflow, want map[string]*wf.TaskState) { |
| t.Helper() |
| if diff := cmp.Diff(l.states[w.ID], want, cmpopts.IgnoreFields(wf.TaskState{}, "SerializedResult")); diff != "" { |
| t.Errorf("task state didn't match expectations: %v", diff) |
| } |
| } |
| |
| func startWorkflow(t *testing.T, wd *wf.Definition, params map[string]interface{}) *wf.Workflow { |
| t.Helper() |
| w, err := wf.Start(wd, params) |
| if err != nil { |
| t.Fatal(err) |
| } |
| return w |
| } |
| |
| func runWorkflow(t *testing.T, w *wf.Workflow, listener wf.Listener) map[string]interface{} { |
| ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second) |
| defer cancel() |
| t.Helper() |
| if listener == nil { |
| listener = &verboseListener{t} |
| } |
| outputs, err := w.Run(ctx, listener) |
| if err != nil { |
| t.Fatalf("w.Run() = _, %v, wanted no error", err) |
| } |
| return outputs |
| } |
| |
| type verboseListener struct{ t *testing.T } |
| |
| func (l *verboseListener) WorkflowStalled(workflowID uuid.UUID) error { |
| l.t.Logf("workflow %q: stalled", workflowID.String()) |
| return nil |
| } |
| |
| func (l *verboseListener) TaskStateChanged(_ uuid.UUID, _ string, st *wf.TaskState) error { |
| switch { |
| case !st.Started: |
| // Task creation is uninteresting. |
| case !st.Finished: |
| l.t.Logf("task %-10v: started", st.Name) |
| case st.Error != "": |
| l.t.Logf("task %-10v: error: %v", st.Name, st.Error) |
| default: |
| l.t.Logf("task %-10v: done: %v", st.Name, st.Result) |
| } |
| return nil |
| } |
| |
| func (l *verboseListener) Logger(_ uuid.UUID, task string) wf.Logger { |
| return &testLogger{t: l.t, task: task} |
| } |
| |
| type testLogger struct { |
| t *testing.T |
| task string |
| } |
| |
| func (l *testLogger) Printf(format string, v ...interface{}) { |
| l.t.Logf("task %-10v: LOG: %s", l.task, fmt.Sprintf(format, v...)) |
| } |
| |
| func runToFailure(t *testing.T, w *wf.Workflow, listener wf.Listener, task string) string { |
| ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second) |
| defer cancel() |
| t.Helper() |
| if listener == nil { |
| listener = &verboseListener{t} |
| } |
| var message string |
| listener = &errorListener{ |
| taskName: task, |
| callback: func(m string) { |
| message = m |
| // Allow other tasks to run before shutting down the workflow. |
| time.AfterFunc(50*time.Millisecond, cancel) |
| }, |
| Listener: listener, |
| } |
| _, err := w.Run(ctx, listener) |
| if err == nil { |
| t.Fatalf("workflow unexpectedly succeeded") |
| } |
| return message |
| } |
| |
| type errorListener struct { |
| taskName string |
| callback func(string) |
| wf.Listener |
| } |
| |
| func (l *errorListener) TaskStateChanged(id uuid.UUID, taskID string, st *wf.TaskState) error { |
| if st.Name == l.taskName && st.Finished && st.Error != "" { |
| l.callback(st.Error) |
| } |
| l.Listener.TaskStateChanged(id, taskID, st) |
| return nil |
| } |