0
0
mirror of https://github.com/go-gitea/gitea.git synced 2026-06-02 14:39:17 +02:00
gitea/models/actions/task.go
Zettat123 0359746abe
feat(actions)!: improve support for reusable workflows (#37478)
## Summary

This PR improves reusable workflow support for Gitea Actions. The
parsing of the called workflow now happens on Gitea side, not on the
runner. When the caller becomes ready, Gitea fetches the called workflow
source, parses it, and inserts each child job into the database as a
`ActionRunJob` linked to the caller via `ParentCallJobID`. As a result,
every callee job is dispatched as its own task and its logs surface as
an independent job entry in the UI, rather than being inlined into the
caller's "Set up job" step.

This PR supports two kinds of `uses` : 
- same-repo call: `uses: ./.gitea/workflows/foo.yaml`
- cross-repo call: `uses: OWNER/REPO/.gitea/workflows/foo.yaml@REF`

## **⚠️ BREAKING ⚠️**
External reusable workflows (`uses:
https://other-gitea-instance/OWNER/REPO/.gitea/workflows/test.yaml@REF`)
are no longer supported. To keep using them, clone the repositories to
the local instance.

## Main changes

### Execution model

- Each caller job carries `IsReusableCaller=true` and won't be fetched
by runners.
- `ParentCallJobID` can link a called job to its caller.
- Caller status is derived from its direct children.


### Workflow syntax

- `jobparser` now supports parsing `on: workflow_call` trigger with
`inputs:`, `outputs:`, and `secrets:` declarations.
- **Max nesting depth**: capped at `MaxReusableCallLevels = 9`, which
means a top-level caller may have at most 9 nested callers below it.
- **Cycle prevention**: at expansion time, `checkCallerChain` walks the
caller's ancestor chain via `ParentCallJobID` and rejects if the same
`uses:` string appears anywhere upstream (`reusable workflow call cycle
detected`). This catches both direct (`A -> A`) and indirect (`A -> B ->
A`) cycles.

### Cross-repo access

- To share reusable workflows from private repos, use `Collaborative
Owners` introduced by #32562

### Rerun semantics

- `expandRerunJobIDs` partitions the latest attempt's jobs into:
- a **rerun set**: jobs being rerun + downstream siblings within the
same scope.
- an **ancestor set**: reusable callers whose only *some* descendants
are being rerun (the caller itself is not).
- Cloning behavior for callers in `execRerunPlan`:
- **Caller is fully rerun** (caller's `AttemptJobID` in `rerunSet`):
none of its descendants are cloned. The caller is cloned with
`IsCallerExpanded=false`, and re-expansion (which reinserts the children
fresh) happens later when the resolver brings the caller to `Waiting`
again.
- **Caller is in ancestor set** (only some descendants rerun): the
caller is pass-through (`Status` will be updated by its fresh children).
Its non-rerun descendants are also pass-through clones (point
`SourceTaskID` at the original task). Their `ParentCallJobID` is
remapped to the new attempt's caller row.

### UI

- Job list in `RepoActionView.vue` is now tree-shaped: callers indent
their children. Callers default to collapsed.
- New caller detail page using `WorkflowGraph` to show direct children
only; the run summary's `WorkflowGraph` shows top-level callers and
their immediate descendants.

### Known trade-offs

- **Caller expansion runs inside the enclosing write transaction.**
`expandReusableWorkflowCaller` performs a git read of the called
workflow while holding the row locks that update the caller and insert
its children. This is intentional: the caller-row update and child-row
inserts must commit atomically. None of the call sites is hot (each
caller is expanded once per attempt), so the trade-off is acceptable.

- **A malformed `if:` expression on a job leaves it `Blocked`
silently.** `evaluateJobIf` now runs server-side as part of resolver
passes; deterministic expression errors (typos, undefined context
fields) are logged but do not surface in the UI. This is the same
behavior the resolver already had for concurrency-expression errors.
Distinguishing transient DB errors from user-authored expression errors
and writing the latter back as `StatusFailure` is a follow-up.


#### Screenshots

<img width="1600" alt="image"
src="https://github.com/user-attachments/assets/bfaa9b7a-07e9-4127-8de9-a81f86e82828"
/>

<img width="1600" alt="image"
src="https://github.com/user-attachments/assets/8af109b3-ef28-4b53-aaad-d4632b923224"
/>


## References

-
https://docs.github.com/en/actions/how-tos/reuse-automations/reuse-workflows
-
https://docs.github.com/en/actions/reference/workflows-and-actions/reusing-workflow-configurations

---

Replace #36388

---------

Signed-off-by: Zettat123 <zettat123@gmail.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
2026-05-30 08:31:14 +02:00

550 lines
15 KiB
Go

// Copyright 2022 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package actions
import (
"context"
"crypto/subtle"
"errors"
"fmt"
"strings"
"time"
runnerv1 "gitea.dev/actions-proto-go/runner/v1"
auth_model "gitea.dev/models/auth"
"gitea.dev/models/db"
"gitea.dev/models/unit"
"gitea.dev/modules/actions/jobparser"
"gitea.dev/modules/log"
"gitea.dev/modules/setting"
"gitea.dev/modules/timeutil"
"gitea.dev/modules/util"
lru "github.com/hashicorp/golang-lru/v2"
"google.golang.org/protobuf/types/known/timestamppb"
"xorm.io/builder"
)
// ActionTask represents a distribution of job
type ActionTask struct {
ID int64
JobID int64
Job *ActionRunJob `xorm:"-"`
Steps []*ActionTaskStep `xorm:"-"`
Attempt int64
RunnerID int64 `xorm:"index"`
Status Status `xorm:"index"`
Started timeutil.TimeStamp `xorm:"index"`
Stopped timeutil.TimeStamp `xorm:"index(stopped_log_expired)"`
RepoID int64 `xorm:"index"`
OwnerID int64 `xorm:"index"`
CommitSHA string `xorm:"index"`
IsForkPullRequest bool
Token string `xorm:"-"`
TokenHash string `xorm:"UNIQUE"` // sha256 of token
TokenSalt string
TokenLastEight string `xorm:"index token_last_eight"`
LogFilename string // file name of log
LogInStorage bool // read log from database or from storage
LogLength int64 // lines count
LogSize int64 // blob size
LogIndexes LogIndexes `xorm:"LONGBLOB"` // line number to offset
LogExpired bool `xorm:"index(stopped_log_expired)"` // files that are too old will be deleted
Created timeutil.TimeStamp `xorm:"created"`
Updated timeutil.TimeStamp `xorm:"updated index"`
}
var successfulTokenTaskCache *lru.Cache[string, any]
func init() {
db.RegisterModel(new(ActionTask), func() error {
if setting.SuccessfulTokensCacheSize > 0 {
var err error
successfulTokenTaskCache, err = lru.New[string, any](setting.SuccessfulTokensCacheSize)
if err != nil {
return fmt.Errorf("unable to allocate Task cache: %v", err)
}
} else {
successfulTokenTaskCache = nil
}
return nil
})
}
func (task *ActionTask) Duration() time.Duration {
return calculateDuration(task.Started, task.Stopped, task.Status, task.Updated)
}
func (task *ActionTask) IsStopped() bool {
return task.Stopped > 0
}
func (task *ActionTask) GetRunLink() string {
if task.Job == nil || task.Job.Run == nil {
return ""
}
return task.Job.Run.Link()
}
func (task *ActionTask) GetCommitLink() string {
if task.Job == nil || task.Job.Run == nil || task.Job.Run.Repo == nil {
return ""
}
return task.Job.Run.Repo.CommitLink(task.CommitSHA)
}
func (task *ActionTask) GetRepoName() string {
if task.Job == nil || task.Job.Run == nil || task.Job.Run.Repo == nil {
return ""
}
return task.Job.Run.Repo.FullName()
}
func (task *ActionTask) GetRepoLink() string {
if task.Job == nil || task.Job.Run == nil || task.Job.Run.Repo == nil {
return ""
}
return task.Job.Run.Repo.Link()
}
func (task *ActionTask) LoadJob(ctx context.Context) error {
if task.Job == nil {
job, err := GetRunJobByRepoAndID(ctx, task.RepoID, task.JobID)
if err != nil {
return err
}
task.Job = job
}
return nil
}
// LoadAttributes load Job Steps if not loaded
func (task *ActionTask) LoadAttributes(ctx context.Context) error {
if err := task.LoadJob(ctx); err != nil {
return err
}
if err := task.Job.LoadAttributes(ctx); err != nil {
return err
}
if task.Steps == nil { // be careful, an empty slice (not nil) also means loaded
steps, err := GetTaskStepsByTaskID(ctx, task.ID)
if err != nil {
return err
}
task.Steps = steps
}
return nil
}
func (task *ActionTask) GenerateAndFillToken() {
task.Token, task.TokenSalt, task.TokenHash, task.TokenLastEight = generateSaltedToken()
}
func GetTaskByID(ctx context.Context, id int64) (*ActionTask, error) {
var task ActionTask
has, err := db.GetEngine(ctx).Where("id=?", id).Get(&task)
if err != nil {
return nil, err
} else if !has {
return nil, fmt.Errorf("task with id %d: %w", id, util.ErrNotExist)
}
return &task, nil
}
func GetRunningTaskByToken(ctx context.Context, token string) (*ActionTask, error) {
errNotExist := fmt.Errorf("task with token %q: %w", token, util.ErrNotExist)
if token == "" {
return nil, errNotExist
}
// A token is defined as being SHA1 sum these are 40 hexadecimal bytes long
if len(token) != 40 {
return nil, errNotExist
}
for _, x := range []byte(token) {
if x < '0' || (x > '9' && x < 'a') || x > 'f' {
return nil, errNotExist
}
}
lastEight := token[len(token)-8:]
if id := getTaskIDFromCache(token); id > 0 {
task := &ActionTask{
TokenLastEight: lastEight,
}
// Re-get the task from the db in case it has been deleted in the intervening period
has, err := db.GetEngine(ctx).ID(id).Get(task)
if err != nil {
return nil, err
}
if has {
return task, nil
}
successfulTokenTaskCache.Remove(token)
}
var tasks []*ActionTask
// Cancelling tasks are still authenticating — post-run cleanup steps need API access (artifact uploads, cache saves, etc.) before the runner finalizes the task.
err := db.GetEngine(ctx).Where("token_last_eight = ? AND status IN (?, ?)", lastEight, StatusRunning, StatusCancelling).Find(&tasks)
if err != nil {
return nil, err
} else if len(tasks) == 0 {
return nil, errNotExist
}
for _, t := range tasks {
tempHash := auth_model.HashToken(token, t.TokenSalt)
if subtle.ConstantTimeCompare([]byte(t.TokenHash), []byte(tempHash)) == 1 {
if successfulTokenTaskCache != nil {
successfulTokenTaskCache.Add(token, t.ID)
}
return t, nil
}
}
return nil, errNotExist
}
func makeTaskStepDisplayName(step *jobparser.Step, limit int) (name string) {
if step.Name != "" {
name = step.Name // the step has an explicit name
} else {
// for unnamed step, its "String()" method tries to get a display name by its "name", "uses",
// "run" or "id" (last fallback), we add the "Run " prefix for unnamed steps for better display
// for multi-line "run" scripts, only use the first line to match GitHub's behavior
// https://github.com/actions/runner/blob/66800900843747f37591b077091dd2c8cf2c1796/src/Runner.Worker/Handlers/ScriptHandler.cs#L45-L58
runStr, _, _ := strings.Cut(strings.TrimSpace(step.Run), "\n")
name = "Run " + util.IfZero(strings.TrimSpace(runStr), step.String())
}
return util.EllipsisDisplayString(name, limit) // database column has a length limit
}
func CreateTaskForRunner(ctx context.Context, runner *ActionRunner) (*ActionTask, bool, error) {
ctx, committer, err := db.TxContext(ctx)
if err != nil {
return nil, false, err
}
defer committer.Close()
e := db.GetEngine(ctx)
jobCond := builder.NewCond()
if runner.RepoID != 0 {
jobCond = builder.Eq{"repo_id": runner.RepoID}
} else if runner.OwnerID != 0 {
jobCond = builder.In("repo_id", builder.Select("`repository`.id").From("repository").
Join("INNER", "repo_unit", "`repository`.id = `repo_unit`.repo_id").
Where(builder.Eq{"`repository`.owner_id": runner.OwnerID, "`repo_unit`.type": unit.TypeActions}))
}
if jobCond.IsValid() {
jobCond = builder.In("run_id", builder.Select("id").From("action_run").Where(jobCond))
}
var jobs []*ActionRunJob
if err := e.Where("task_id=? AND status=? AND is_reusable_caller=?", 0, StatusWaiting, false).And(jobCond).Asc("updated", "id").Find(&jobs); err != nil {
return nil, false, err
}
// TODO: a more efficient way to filter labels
var job *ActionRunJob
log.Trace("runner labels: %v", runner.AgentLabels)
for _, v := range jobs {
if runner.CanMatchLabels(v.RunsOn) {
job = v
break
}
}
if job == nil {
return nil, false, nil
}
if err := job.LoadAttributes(ctx); err != nil {
return nil, false, err
}
now := timeutil.TimeStampNow()
job.Started = now
job.Status = StatusRunning
task := &ActionTask{
JobID: job.ID,
Attempt: job.Attempt,
RunnerID: runner.ID,
Started: now,
Status: StatusRunning,
RepoID: job.RepoID,
OwnerID: job.OwnerID,
CommitSHA: job.CommitSHA,
IsForkPullRequest: job.IsForkPullRequest,
}
task.GenerateAndFillToken()
workflowJob, err := job.ParseJob()
if err != nil {
return nil, false, fmt.Errorf("load job %d: %w", job.ID, err)
}
if _, err := e.Insert(task); err != nil {
return nil, false, err
}
task.LogFilename = logFileName(job.Run.Repo.FullName(), task.ID)
if err := UpdateTask(ctx, task, "log_filename"); err != nil {
return nil, false, err
}
if len(workflowJob.Steps) > 0 {
steps := make([]*ActionTaskStep, len(workflowJob.Steps))
for i, v := range workflowJob.Steps {
steps[i] = &ActionTaskStep{
Name: makeTaskStepDisplayName(v, 255),
TaskID: task.ID,
Index: int64(i),
RepoID: task.RepoID,
Status: StatusWaiting,
}
}
if _, err := e.Insert(steps); err != nil {
return nil, false, err
}
task.Steps = steps
}
job.TaskID = task.ID
if n, err := UpdateRunJob(ctx, job, builder.Eq{"task_id": 0}); err != nil {
return nil, false, err
} else if n != 1 {
return nil, false, nil
}
task.Job = job
if err := committer.Commit(); err != nil {
return nil, false, err
}
return task, true, nil
}
func UpdateTask(ctx context.Context, task *ActionTask, cols ...string) error {
sess := db.GetEngine(ctx).ID(task.ID)
if len(cols) > 0 {
sess.Cols(cols...)
}
_, err := sess.Update(task)
// Automatically delete the ephemeral runner if the task is done
if err == nil && task.Status.IsDone() && util.SliceContainsString(cols, "status") {
return DeleteEphemeralRunner(ctx, task.RunnerID)
}
return err
}
// UpdateTaskByState updates the task by the state.
// It will always update the task if the state is not final, even there is no change.
// So it will update ActionTask.Updated to avoid the task being judged as a zombie task.
func UpdateTaskByState(ctx context.Context, runnerID int64, state *runnerv1.TaskState) (*ActionTask, error) {
stepStates := map[int64]*runnerv1.StepState{}
for _, v := range state.Steps {
stepStates[v.Id] = v
}
return db.WithTx2(ctx, func(ctx context.Context) (*ActionTask, error) {
e := db.GetEngine(ctx)
task := &ActionTask{}
if has, err := e.ID(state.Id).Get(task); err != nil {
return nil, err
} else if !has {
return nil, util.ErrNotExist
} else if runnerID != task.RunnerID {
return nil, errors.New("invalid runner for task")
}
if task.Status.IsDone() {
// the state is final, do nothing
return task, nil
}
// state.Result is not unspecified means the task is finished
if state.Result != runnerv1.Result_RESULT_UNSPECIFIED {
if task.Status == StatusCancelling {
// The runner may report SUCCESS/FAILURE for the cleanup phase; preserve user intent.
task.Status = StatusCancelled
} else {
task.Status = StatusFromResult(state.Result)
}
task.Stopped = timeutil.TimeStamp(state.StoppedAt.AsTime().Unix())
if err := UpdateTask(ctx, task, "status", "stopped"); err != nil {
return nil, err
}
if _, err := UpdateRunJob(ctx, &ActionRunJob{
ID: task.JobID,
RepoID: task.RepoID,
Status: task.Status,
Stopped: task.Stopped,
}, nil, "status", "stopped"); err != nil {
return nil, err
}
} else {
// Force update ActionTask.Updated to avoid the task being judged as a zombie task
task.Updated = timeutil.TimeStampNow()
if err := UpdateTask(ctx, task, "updated"); err != nil {
return nil, err
}
}
if err := task.LoadAttributes(ctx); err != nil {
return nil, err
}
for _, step := range task.Steps {
var result runnerv1.Result
if v, ok := stepStates[step.Index]; ok {
result = v.Result
step.LogIndex = v.LogIndex
step.LogLength = v.LogLength
step.Started = convertTimestamp(v.StartedAt)
step.Stopped = convertTimestamp(v.StoppedAt)
}
if result != runnerv1.Result_RESULT_UNSPECIFIED {
step.Status = StatusFromResult(result)
} else if step.Started != 0 {
step.Status = StatusRunning
}
if _, err := e.ID(step.ID).Update(step); err != nil {
return nil, err
}
}
return task, nil
})
}
func StopTask(ctx context.Context, taskID int64, status Status) error {
if !status.IsDone() && status != StatusCancelling {
return fmt.Errorf("cannot stop task with status %v", status)
}
e := db.GetEngine(ctx)
task := &ActionTask{}
if has, err := e.ID(taskID).Get(task); err != nil {
return err
} else if !has {
return util.ErrNotExist
}
if task.Status.IsDone() {
return nil
}
now := timeutil.TimeStampNow()
if status == StatusCancelling {
runner, err := GetRunnerByID(ctx, task.RunnerID)
if err != nil {
if !errors.Is(err, util.ErrNotExist) {
return err
}
status = StatusCancelled
} else if !runner.HasCancellingSupport {
status = StatusCancelled
}
}
if status == StatusCancelling {
task.Status = StatusCancelling
if _, err := UpdateRunJob(ctx, &ActionRunJob{
ID: task.JobID,
RepoID: task.RepoID,
Status: StatusCancelling,
}, nil, "status"); err != nil {
return err
}
return UpdateTask(ctx, task, "status")
}
task.Status = status
task.Stopped = now
if _, err := UpdateRunJob(ctx, &ActionRunJob{
ID: task.JobID,
RepoID: task.RepoID,
Status: task.Status,
Stopped: task.Stopped,
}, nil); err != nil {
return err
}
if err := UpdateTask(ctx, task, "status", "stopped"); err != nil {
return err
}
if err := task.LoadAttributes(ctx); err != nil {
return err
}
for _, step := range task.Steps {
if !step.Status.IsDone() {
step.Status = status
if step.Started == 0 {
step.Started = now
}
step.Stopped = now
}
if _, err := e.ID(step.ID).Update(step); err != nil {
return err
}
}
return nil
}
func FindOldTasksToExpire(ctx context.Context, olderThan timeutil.TimeStamp, limit int) ([]*ActionTask, error) {
e := db.GetEngine(ctx)
tasks := make([]*ActionTask, 0, limit)
// Check "stopped > 0" to avoid deleting tasks that are still running
return tasks, e.Where("stopped > 0 AND stopped < ? AND log_expired = ?", olderThan, false).
Limit(limit).
Find(&tasks)
}
func convertTimestamp(timestamp *timestamppb.Timestamp) timeutil.TimeStamp {
if timestamp.GetSeconds() == 0 && timestamp.GetNanos() == 0 {
return timeutil.TimeStamp(0)
}
return timeutil.TimeStamp(timestamp.AsTime().Unix())
}
func logFileName(repoFullName string, taskID int64) string {
ret := fmt.Sprintf("%s/%02x/%d.log", repoFullName, taskID%256, taskID)
if setting.Actions.LogCompression.IsZstd() {
ret += ".zst"
}
return ret
}
func getTaskIDFromCache(token string) int64 {
if successfulTokenTaskCache == nil {
return 0
}
tInterface, ok := successfulTokenTaskCache.Get(token)
if !ok {
return 0
}
t, ok := tInterface.(int64)
if !ok {
return 0
}
return t
}