From b71adfe1ad5657d641dfdbe5769858ab645ef216 Mon Sep 17 00:00:00 2001
From: Giteabot <teabot@gitea.io>
Date: Wed, 5 Aug 2026 23:45:42 -0700
Subject: [PATCH] fix(actions): write an action task report in one transaction
 (#38792) (#38794)

Backport #38792 by @silverwind

`UpdateTaskByState` wrote the task, its job and its steps in separate
statements. An interruption in between left the task finished with a
running job, so the run stayed in progress, and the "state is final"
early return made every retry, cancel and cleanup a no-op.

Fixes https://github.com/go-gitea/gitea/issues/38790

Co-authored-by: silverwind <me@silverwind.io>
---
 models/actions/task.go      |  6 +++++-
 models/actions/task_test.go | 33 +++++++++++++++++++++++++++++++++
 2 files changed, 38 insertions(+), 1 deletion(-)

diff --git a/models/actions/task.go b/models/actions/task.go
index 5a764090222..2a4304710ce 100644
--- a/models/actions/task.go
+++ b/models/actions/task.go
@@ -478,7 +478,7 @@ func UpdateTaskByState(ctx context.Context, runnerID int64, state *runnerv1.Task
 		return nil, err
 	}
 	task := &ActionTask{}
-	err = globallock.LockAndDo(ctx, fmt.Sprintf("UpdateTaskByState-run-%d", runID), func(ctx context.Context) error {
+	applyState := func(ctx context.Context) error {
 		if has, err := db.GetEngine(ctx).ID(taskID).Get(task); err != nil {
 			return err
 		} else if !has {
@@ -543,6 +543,10 @@ func UpdateTaskByState(ctx context.Context, runnerID int64, state *runnerv1.Task
 			}
 		}
 		return nil
+	}
+	err = globallock.LockAndDo(ctx, fmt.Sprintf("UpdateTaskByState-run-%d", runID), func(ctx context.Context) error {
+		// A half-written report leaves the task done with a running job, which no retry repairs.
+		return db.WithTx(ctx, applyState)
 	})
 	return task, err
 }
diff --git a/models/actions/task_test.go b/models/actions/task_test.go
index 139de0c9ec3..a48dbf0ebe0 100644
--- a/models/actions/task_test.go
+++ b/models/actions/task_test.go
@@ -4,7 +4,10 @@
 package actions
 
 import (
+	"context"
+	"errors"
 	"strings"
+	"sync/atomic"
 	"testing"
 
 	runnerv1 "gitea.dev/actions-proto-go/runner/v1"
@@ -17,6 +20,7 @@ import (
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"google.golang.org/protobuf/types/known/timestamppb"
+	"xorm.io/xorm/contexts"
 )
 
 func TestActionTask_GetRunJobLink(t *testing.T) {
@@ -357,6 +361,35 @@ func TestCreateTaskForRunnerPagination(t *testing.T) {
 	assert.Equal(t, task.ID, claimed.TaskID)
 }
 
+type failFirstStepWrite struct{ fired atomic.Bool }
+
+func (h *failFirstStepWrite) BeforeProcess(c *contexts.ContextHook) (context.Context, error) {
+	if !h.fired.Load() && strings.HasPrefix(c.SQL, "UPDATE") && strings.Contains(c.SQL, "action_task_step") {
+		h.fired.Store(true)
+		return nil, errors.New("interrupted")
+	}
+	return c.Ctx, nil
+}
+
+func (*failFirstStepWrite) AfterProcess(*contexts.ContextHook) error { return nil }
+
+// TestUpdateTaskByStateIsAtomic checks that an interrupted report writes nothing: a surviving task or
+// job write would hit the "state is final" early return, which no retry or cleanup repairs.
+func TestUpdateTaskByStateIsAtomic(t *testing.T) {
+	require.NoError(t, unittest.PrepareTestDatabase())
+	task, job := newRunningTaskForCancelling(t, "atomic-report-job", true)
+	require.NoError(t, db.Insert(t.Context(), &ActionTaskStep{TaskID: task.ID, RepoID: task.RepoID, Status: StatusRunning}))
+	unittest.GetXORMEngine().AddHook(&failFirstStepWrite{})
+	finalState := &runnerv1.TaskState{Id: task.ID, Result: runnerv1.Result_RESULT_SUCCESS, StoppedAt: timestamppb.Now()}
+	_, err := UpdateTaskByState(t.Context(), task.RunnerID, finalState)
+	require.Error(t, err)
+	assert.Equal(t, StatusRunning, unittest.AssertExistsAndLoadBean(t, &ActionTask{ID: task.ID}).Status)
+	assert.Equal(t, StatusRunning, unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: job.ID}).Status)
+	_, err = UpdateTaskByState(t.Context(), task.RunnerID, finalState)
+	require.NoError(t, err)
+	assert.Equal(t, StatusSuccess, unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: job.ID}).Status)
+}
+
 // newRunningTaskForCancelling inserts a running run/job/task assigned to a fresh runner,
 // which is the state every cancellation test starts from.
 func newRunningTaskForCancelling(t *testing.T, name string, hasCancellingSupport bool) (*ActionTask, *ActionRunJob) {