Gather Run Summary even when tasks fail (#4524)

cli/integration_tests/run_summary/error.t

@@ -0,0 +1,114 @@
+Setup
+  $ . ${TESTDIR}/../_helpers/setup.sh
+  $ . ${TESTDIR}/../_helpers/setup_monorepo.sh $(pwd)
+
+  $ rm -rf .turbo/runs
+
+Use --filter because otherwise we'll get nondeterministic execution summary depending on
+whether the other workspaces had finished executing their task yet. We don't care to validate
+that behavior in this test.
+  $ ${TURBO} run maybefails --filter=my-app --summarize > /dev/null 2>&1
+  [1]
+
+  $ source "$TESTDIR/../_helpers/run-summary-utils.sh"
+  $ SUMMARY=$(/bin/ls .turbo/runs/*.json | head -n1)
+
+Validate that there was a failed task and exitCode is 1 (which is what we get from npm for the failed task)
+  $ cat $SUMMARY | jq '.execution'
+  {
+    "command": "turbo run maybefails --filter=my-app",
+    "repoPath": "",
+    "success": 0,
+    "failed": 1,
+    "cached": 0,
+    "attempted": 1,
+    "startTime": [0-9]+, (re)
+    "endTime": [0-9]+, (re)
+    "exitCode": 1
+  }
+
+Validate that we got a full task summary for the failed task with an error in .execution
+  $ echo $(getSummaryTaskId $SUMMARY "my-app#maybefails") | jq
+  {
+    "taskId": "my-app#maybefails",
+    "task": "maybefails",
+    "package": "my-app",
+    "hash": "c11710f9379cb7ea",
+    "inputs": {
+      "package.json": "6bcf57fd6ff30d1a6f40ad8d8d08e8b940fc7e3b"
+    },
+    "hashOfExternalDependencies": "ccab0b28617f1f56",
+    "cache": {
+      "local": false,
+      "remote": false
+    },
+    "command": "exit 4",
+    "cliArguments": [],
+    "outputs": null,
+    "excludedOutputs": null,
+    "logFile": "apps/my-app/.turbo/turbo-maybefails.log",
+    "directory": "apps/my-app",
+    "dependencies": [],
+    "dependents": [],
+    "resolvedTaskDefinition": {
+      "outputs": [],
+      "cache": true,
+      "dependsOn": [],
+      "inputs": [],
+      "outputMode": "full",
+      "env": [],
+      "persistent": false
+    },
+    "expandedOutputs": [],
+    "framework": "",
+    "environmentVariables": {
+      "configured": [],
+      "inferred": [],
+      "global": [
+        "SOME_ENV_VAR=",
+        "VERCEL_ANALYTICS_ID="
+      ]
+    },
+    "execution": {
+      "startTime": [0-9]+, (re)
+      "endTime": [0-9]+, (re)
+      "error": "command .* npm run maybefails exited \(1\)", (re)
+      "exitCode": 1
+    }
+  }
+
+With --continue flag. We won't validate the whole thing, just execution
+Use  --force flag so we can be deterministic about cache hits
+Don't use --filter here, so we can validate that both tasks attempted to run
+  $ rm -rf .turbo/runs
+  $ ${TURBO} run maybefails --summarize --force --continue > /dev/null  2>&1
+  [1]
+
+  $ source "$TESTDIR/../_helpers/run-summary-utils.sh"
+  $ SUMMARY=$(/bin/ls .turbo/runs/*.json | head -n1)
+
+success should be 1, and attempted should be 2
+  $ cat $SUMMARY | jq '.execution'
+  {
+    "command": "turbo run maybefails --continue",
+    "repoPath": "",
+    "success": 1,
+    "failed": 1,
+    "cached": 0,
+    "attempted": 2,
+    "startTime": [0-9]+, (re)
+    "endTime": [0-9]+, (re)
+    "exitCode": 1
+  }
+
+  $ cat $SUMMARY | jq '.tasks | length'
+  2
+
+# exitCode is 1, because npm will report all errors with exitCode 1
+  $ getSummaryTaskId $SUMMARY "my-app#maybefails" | jq '.execution'
+  {
+    "startTime": [0-9]+, (re)
+    "endTime": [0-9]+, (re)
+    "error": "command .* npm run maybefails exited \(1\)", (re)
+    "exitCode": 1
+  }

cli/integration_tests/run_summary/run_summary.t → cli/integration_tests/run_summary/monorepo.t

@@ -59,7 +59,6 @@ Setup
   {
     "startTime": [0-9]+, (re)
     "endTime": [0-9]+, (re)
-    "error": null,
     "exitCode": 0
   }
   $ echo $FIRST_APP_BUILD | jq '.cliArguments'
@@ -89,7 +88,6 @@ Setup
   {
     "startTime": [0-9]+, (re)
     "endTime": [0-9]+, (re)
-    "error": null,
     "exitCode": 0
   }
 

cli/integration_tests/single_package/run-summary.t → cli/integration_tests/run_summary/single-pacakge.t

@@ -70,7 +70,6 @@ Check
   {
     "startTime": [0-9]+, (re)
     "endTime": [0-9]+, (re)
-    "error": null,
     "exitCode": 0
   }
   $ echo $TASK_SUMMARY | jq '.cliArguments'

cli/internal/graph/graph.go

@@ -52,7 +52,7 @@ func (g *CompleteGraph) GetPackageTaskVisitor(
 	taskGraph *dag.AcyclicGraph,
 	getArgs func(taskID string) []string,
 	logger hclog.Logger,
-	visitor func(ctx gocontext.Context, packageTask *nodes.PackageTask, taskSummary *runsummary.TaskSummary) error,
+	execFunc func(ctx gocontext.Context, packageTask *nodes.PackageTask, taskSummary *runsummary.TaskSummary) error,
 ) func(taskID string) error {
 	return func(taskID string) error {
 		packageName, taskName := util.GetPackageTaskFromId(taskID)
@@ -140,7 +140,7 @@ func (g *CompleteGraph) GetPackageTaskVisitor(
 			summary.Dependents = descendents
 		}
 
-		return visitor(ctx, packageTask, summary)
+		return execFunc(ctx, packageTask, summary)
 	}
 }
 

cli/internal/run/real_run.go

@@ -98,9 +98,6 @@ func RealRun(
 	taskSummaries := []*runsummary.TaskSummary{}
 	execFunc := func(ctx gocontext.Context, packageTask *nodes.PackageTask, taskSummary *runsummary.TaskSummary) error {
 		taskExecutionSummary, err := ec.exec(ctx, packageTask)
-		if err != nil {
-			return err
-		}
 
 		// taskExecutionSummary will be nil if the task never executed
 		// (i.e. if the workspace didn't implement the script corresponding to the task)
@@ -117,6 +114,11 @@ func RealRun(
 			mu.Unlock()
 		}
 
+		// Return the error when there is one
+		if err != nil {
+			return err
+		}
+
 		return nil
 	}
 

cli/internal/runsummary/execution_summary.go

@@ -26,7 +26,7 @@ type executionEvent struct {
 	// Its current status
 	Status executionEventName
 	// Error, only populated for failure statuses
-	Err error
+	Err string
 
 	exitCode *int
 }
@@ -68,7 +68,7 @@ func (en executionEventName) toString() string {
 type TaskExecutionSummary struct {
 	startAt  time.Time          // set once
 	status   executionEventName // current status, updated during execution
-	err      error              // only populated for failure statuses
+	err      string             // only populated for failure statuses
 	Duration time.Duration      // updated during the task execution
 	exitCode *int               // pointer so we can distinguish between 0 and unknown.
 }
@@ -81,10 +81,10 @@ func (ts *TaskExecutionSummary) endTime() time.Time {
 // We'll use an anonmyous, private struct for this, so it's not confusingly duplicated
 func (ts *TaskExecutionSummary) MarshalJSON() ([]byte, error) {
 	serializable := struct {
-		Start    int64 `json:"startTime"`
-		End      int64 `json:"endTime"`
-		Err      error `json:"error"`
-		ExitCode *int  `json:"exitCode"`
+		Start    int64  `json:"startTime"`
+		End      int64  `json:"endTime"`
+		Err      string `json:"error,omitempty"`
+		ExitCode *int   `json:"exitCode"`
 	}{
 		Start:    ts.startAt.UnixMilli(),
 		End:      ts.endTime().UnixMilli(),
@@ -197,8 +197,9 @@ func (es *executionSummary) run(taskID string) (func(outcome executionEventName,
 			// when we assign it to the taskExecutionSummary.
 			exitCode: exitCode,
 		}
+
 		if err != nil {
-			result.Err = fmt.Errorf("running %v failed: %w", taskID, err)
+			result.Err = err.Error()
 		}
 
 		// Ignore the return value here