fix(apiserver): restart carries only the options a job gets (bug-1968)

RestartJobByID rebuilt the new job from the old job's row:
- it accepted asked_hosts and copied asked hosts, I/O paths and out_dir,
  which no job gets (refused at submit since bug-1936); restart now
  refuses asked_hosts with 400 and no longer carries those over;
- it read the JSONB columns chkpnt_info/extend_data as strings, but rows
  keep models.Jsonb, so env/limits were lost and a checkpointed job never
  restarted from its checkpoint; new rowJSON() reads both;
- it read "res_req" (the column is "resreq"), losing the resource
  requirement, and never copied exec_home.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

src/frontend/apiserver/controllers/jobs_handler_test.go

@@ -1522,6 +1522,131 @@ func TestJobHandler_RestartNamedJob(t *testing.T) {
 	}
 }
 
+// bug-1968: restart, like submit (bug-1936), refuses an option the restarted
+// job would never get: asked hosts (-m) have no field in the controller's
+// submit data. Nothing is submitted.
+func TestJobHandler_RestartRejectsAskedHosts(t *testing.T) {
+	f := newHandlerFixture(t)
+	ord := handlerTestOrdinaryBase + 50<<15
+	f.seedRestartJob(ord, "sleep 1")
+
+	before := len(f.submitted)
+	w := f.do(http.MethodPost, "/xce/v1/jobs/"+strconv.FormatUint(ord, 10)+"/restart",
+		map[string]interface{}{"asked_hosts": "hostA"})
+	if w.Code != http.StatusBadRequest {
+		t.Errorf("status = %d, want 400; body: %s", w.Code, w.Body.String())
+	}
+	if len(f.submitted) != before {
+		t.Errorf("%d submits, want none", len(f.submitted)-before)
+	}
+}
+
+// bug-1968: the restarted job is rebuilt from the old job's row. Options no
+// job can get (asked hosts, begin/term time, I/O paths, out_dir; refused at
+// submit since bug-1936) must not be carried over, while the options the job
+// does get -- extended data (env, limits), cwd and exec home -- must be,
+// whether or not the old job was checkpointed.
+func TestJobHandler_RestartCarriesOnlyHonouredOptions(t *testing.T) {
+	f := newHandlerFixture(t)
+	ord := handlerTestOrdinaryBase + 51<<15
+	exdata := `{"sub_env_vars":{"A":"1"},"limits":{"run_limit":60}}`
+	f.seedJob(models.JobInfo{
+		Unique_JobID: int64(ord),
+		JobID:        int64(backend.ExtractArrayBaseID(ord)),
+		JobName:      "r",
+		Command:      "sleep 1",
+		JobStatus:    models.JOBSTAT["DONE"],
+		CWD:          "/data/run",
+		ExecHome:     "/home/u",
+		ResReq:       "rusage[mem=100]",
+		ExtendData:   models.Jsonb{String: exdata},
+		AskedHosts:   "hostA",
+		BeginTime:    models.CustomTime{Time: time.Unix(1700000000, 0)},
+		TermTime:     models.CustomTime{Time: time.Unix(1800000000, 0)},
+		InFile:       "/in",
+		OutFile:      "/out",
+		ErrFile:      "/err",
+		OutDir:       "/outdir",
+	})
+
+	before := len(f.submitted)
+	w := f.restart(strconv.FormatUint(ord, 10))
+	if w.Code != http.StatusOK && w.Code != http.StatusCreated {
+		t.Fatalf("status = %d; body: %s", w.Code, w.Body.String())
+	}
+	if len(f.submitted) != before+1 {
+		t.Fatalf("%d submits, want 1", len(f.submitted)-before)
+	}
+	job := f.submitted[before].Job
+
+	dropped := map[string]bool{
+		"asked_hosts": job.AskedHosts != "",
+		"begin_time":  !job.BeginTime.Time.IsZero(),
+		"term_time":   !job.TermTime.Time.IsZero(),
+		"in_file":     job.InFile != "",
+		"out_file":    job.OutFile != "",
+		"err_file":    job.ErrFile != "",
+		"out_dir":     job.OutDir != "",
+	}
+	for field, kept := range dropped {
+		if kept {
+			t.Errorf("restarted job kept %s, which no job gets", field)
+		}
+	}
+	if job.ExtendData.String != exdata {
+		t.Errorf("ExtendData = %q, want %q", job.ExtendData.String, exdata)
+	}
+	if job.CWD != "/data/run" {
+		t.Errorf("CWD = %q, want /data/run", job.CWD)
+	}
+	if job.ExecHome != "/home/u" {
+		t.Errorf("ExecHome = %q, want /home/u", job.ExecHome)
+	}
+	if job.ResReq != "rusage[mem=100]" {
+		t.Errorf("ResReq = %q, want rusage[mem=100]", job.ResReq)
+	}
+}
+
+// bug-1968: a checkpointed job restarts from its checkpoint: the restart
+// command replaces the command and the checkpointed pid reaches the job as
+// XCE_JOB_CHECKPOINT_PID next to the old job's environment. The stored
+// chkpnt_info/extend_data are models.Jsonb, which the restart never read.
+func TestJobHandler_RestartFromCheckpoint(t *testing.T) {
+	f := newHandlerFixture(t)
+	ord := handlerTestOrdinaryBase + 52<<15
+	f.seedJob(models.JobInfo{
+		Unique_JobID: int64(ord),
+		JobID:        int64(backend.ExtractArrayBaseID(ord)),
+		JobName:      "r",
+		Command:      "sleep 1",
+		JobStatus:    models.JOBSTAT["EXIT"],
+		ChkpntInfo:   models.Jsonb{String: `{"restart_command":"criu restore","ckpt_pid":4321}`},
+		ExtendData:   models.Jsonb{String: `{"sub_env_vars":{"A":"1"}}`},
+	})
+
+	before := len(f.submitted)
+	w := f.restart(strconv.FormatUint(ord, 10))
+	if w.Code != http.StatusOK && w.Code != http.StatusCreated {
+		t.Fatalf("status = %d; body: %s", w.Code, w.Body.String())
+	}
+	if len(f.submitted) != before+1 {
+		t.Fatalf("%d submits, want 1", len(f.submitted)-before)
+	}
+	job := f.submitted[before].Job
+	if job.Command != "criu restore" {
+		t.Errorf("Command = %q, want the checkpoint's restart command", job.Command)
+	}
+	var ex struct {
+		Env map[string]string `json:"sub_env_vars"`
+	}
+	if err := json.Unmarshal([]byte(job.ExtendData.String), &ex); err != nil {
+		t.Fatalf("ExtendData %q: %v", job.ExtendData.String, err)
+	}
+	if ex.Env["A"] != "1" || ex.Env["XCE_JOB_CHECKPOINT_PID"] != "4321" {
+		t.Errorf("sub_env_vars = %v, want A=1 and XCE_JOB_CHECKPOINT_PID=4321", ex.Env)
+	}
+}
+
 // bug-1936: the job handed to the backend carries the cwd and the extended
 // data under the key names and units the worker's parse_exdata()/
 // parse_limits() read (run_limit in seconds, memory limits in MB,