myxce/src/frontend/apiserver/backend/clib_test.go

//go:build cgo

package backend

import (
	"os"
	"strings"
	"testing"
	"time"

	"go.uber.org/zap/zaptest/observer"
)

// clib_test.go exercises RegisterToBackend/buildClusterConfig (clib.go), the
// unified candidate-list connect pipeline that replaces the former
// dual-mechanism implementation (component-failover-adoption, task 4.2,
// boundary: ApiserverBackendBridge). Kept //go:build cgo since these
// functions only exist in the CGO build -- this file itself does not
// `import "C"` (a _test.go file can never do that directly, golang/go#13938):
// it only calls the plain-Go-signature functions defined in clib.go, same
// split candidates.go/candidates_test.go already established in this
// package.
//
// Test strategy:
//   - Req 1.5 (no-Peers and single-entry-Peers converge on the identical
//     validated xce_cluster_config_t shape) and Req 1.1-1.3 (multi-candidate
//     passthrough) are exercised via buildClusterConfig() directly -- pure
//     struct-building + validation, no network I/O, so these run fast and
//     avoid any socket/leak concerns from a real (failing) connect attempt.
//   - "a single xce_backend_init() call carries the full candidate count"
//     is exercised by calling the real RegisterToBackend() against closed
//     local ports (connect() returns ECONNREFUSED near-instantly, same
//     "no backoff loop inside xcecli_start()'s per-master connect scan"
//     property scheduler's test_scheduler_connect.c and worker's task 2.3
//     test both rely on) and inspecting gClientCfg.nmasters afterward via
//     currentClientConfigNMasters() -- this is deliberately the ONLY test in
//     this file that reaches xce_backend_init()/a real connect() call, to
//     keep the (unavoidable, pre-existing -- see status report CONCERNS)
//     per-call xcecli_ctx_init() allocation this path leaks on a failed
//     connect bounded to one occurrence.
//   - "no-peers and single-entry-peers produce the identical connect log
//     sequence" is exercised behaviorally via captureLog() (handler_test.go,
//     this package), asserting on the log MESSAGE sequence (not field
//     values) RegisterToBackend emits for each source.
//   - The "source config structs are package-level, not stack-local"
//     regression guard is a source-text scan of clib.go, same genre as
//     scheduler's/worker's structural regression tests (test_scheduler_connect.c,
//     test_controller_connector.c) for the identical hazard.

// testPortBase is deliberately unlikely to have a real listener bound in
// CI/dev sandboxes -- connect() against a closed local port fails fast, so
// RegisterToBackend() calls below return quickly without needing a live
// controller (same precedent as scheduler's TEST_PORT_BASE).
const testPortBase = 28581

// ════════════════════════════════════════════════════════════════════════
// Req 1.5 -- unified candidate-list pipeline: no branch for "single host"
// ════════════════════════════════════════════════════════════════════════

func TestBuildClusterConfig_NoPeersAndSingleEntryPeers_IdenticalShape(t *testing.T) {
	// (a) No Peers: legacy Address/Port synthesized into a single candidate.
	cfgNoPeers := &BackendConfig{
		Address: "127.0.0.1",
		Port:    testPortBase,
		From:    "xce_apiserver_test",
		Key:     "keepsecret$",
	}
	if err := buildClusterConfig(cfgNoPeers); err != nil {
		t.Fatalf("buildClusterConfig(no peers) failed: %v", err)
	}
	noPeersCandidates := currentClusterCandidates()

	// (b) A single-entry Peers list pointed at the exact same host:port.
	cfgOnePeer := &BackendConfig{
		Peers: []string{"127.0.0.1:28581"},
		From:  "xce_apiserver_test",
		Key:   "keepsecret$",
	}
	if err := buildClusterConfig(cfgOnePeer); err != nil {
		t.Fatalf("buildClusterConfig(one peer) failed: %v", err)
	}
	onePeerCandidates := currentClusterCandidates()

	if len(noPeersCandidates) != 1 {
		t.Fatalf("no-peers: got %d candidates, want 1", len(noPeersCandidates))
	}
	if len(onePeerCandidates) != 1 {
		t.Fatalf("single-entry-peers: got %d candidates, want 1", len(onePeerCandidates))
	}
	if noPeersCandidates[0].Host != onePeerCandidates[0].Host ||
		noPeersCandidates[0].Port != onePeerCandidates[0].Port {
		t.Errorf("no-peers and single-entry-peers candidates diverge: %+v vs %+v",
			noPeersCandidates[0], onePeerCandidates[0])
	}
}

func TestBuildClusterConfig_MultiCandidatePassthrough(t *testing.T) {
	cfg := &BackendConfig{
		Peers: []string{
			"127.0.0.1:28590",
			"127.0.0.1:28591",
			"127.0.0.1:28592",
		},
		From: "xce_apiserver_test",
		Key:  "keepsecret$",
	}
	if err := buildClusterConfig(cfg); err != nil {
		t.Fatalf("buildClusterConfig failed: %v", err)
	}
	candidates := currentClusterCandidates()
	if len(candidates) != 3 {
		t.Fatalf("got %d candidates, want 3", len(candidates))
	}
	for i, want := range []int{28590, 28591, 28592} {
		if candidates[i].Host != "127.0.0.1" || candidates[i].Port != want {
			t.Errorf("candidate %d = %+v, want host=127.0.0.1 port=%d", i, candidates[i], want)
		}
	}
}

func TestBuildClusterConfig_MalformedPeersRejected(t *testing.T) {
	cfg := &BackendConfig{
		Peers: []string{"not-a-valid-entry"},
		From:  "xce_apiserver_test",
		Key:   "keepsecret$",
	}
	err := buildClusterConfig(cfg)
	if err == nil {
		t.Fatal("expected error for malformed --peers entry, got nil")
	}
	if !strings.Contains(err.Error(), "not-a-valid-entry") {
		t.Errorf("error %q does not identify the malformed input", err.Error())
	}
}

// ════════════════════════════════════════════════════════════════════════
// Req 1.1-1.3 -- one xce_backend_init() call carries the full candidate list
// ════════════════════════════════════════════════════════════════════════

func TestRegisterToBackend_MultiCandidate_SingleInitCallCarriesFullCount(t *testing.T) {
	captureLog(t) // RegisterToBackend logs via utils.Log; nil otherwise (utils.InitLogger is never called in unit tests)

	cfg := &BackendConfig{
		Peers: []string{
			"127.0.0.1:28600",
			"127.0.0.1:28601",
			"127.0.0.1:28602",
		},
		From: "xce_apiserver_test",
		Key:  "keepsecret$",
	}

	start := time.Now()
	err := RegisterToBackend(cfg)
	elapsed := time.Since(start)

	if err == nil {
		t.Fatal("expected RegisterToBackend to fail against 3 closed local ports")
	}
	// Generous bound: closed-port ECONNREFUSED is near-instant, well under a
	// few seconds even across 3 candidates -- proves this is NOT the deleted
	// per-candidate Go loop silently retrying/blocking somewhere.
	if elapsed > 5*time.Second {
		t.Errorf("RegisterToBackend took %v against closed ports -- too slow, "+
			"suggests a retry loop is still present", elapsed)
	}

	// The struct xce_backend_init() was actually called with must carry the
	// FULL candidate count (nmasters=3), not the hardcoded 1 the deleted
	// per-peer Go loop always used -- proves ONE call, not three separate
	// single-candidate calls.
	if got := currentClientConfigNMasters(); got != 3 {
		t.Errorf("gClientCfg.nmasters = %d, want 3 (a single call must carry "+
			"the full candidate count)", got)
	}
}

// ════════════════════════════════════════════════════════════════════════
// Req 5.1/5.2 -- no-peers and single-entry-peers produce the identical
// connect log sequence (shape, not values)
// ════════════════════════════════════════════════════════════════════════

func TestRegisterToBackend_NoPeersAndSingleEntryPeers_IdenticalLogSequence(t *testing.T) {
	logs := captureLog(t)

	cfgNoPeers := &BackendConfig{
		Address: "127.0.0.1",
		Port:    testPortBase + 20,
		From:    "xce_apiserver_test",
		Key:     "keepsecret$",
	}
	if err := RegisterToBackend(cfgNoPeers); err == nil {
		t.Fatal("expected RegisterToBackend(no peers) to fail against a closed port")
	}
	noPeersMessages := logMessages(logs) // drains the observer buffer

	cfgOnePeer := &BackendConfig{
		Peers: []string{"127.0.0.1:28601"},
		From:  "xce_apiserver_test",
		Key:   "keepsecret$",
	}
	if err := RegisterToBackend(cfgOnePeer); err == nil {
		t.Fatal("expected RegisterToBackend(single-entry peers) to fail against a closed port")
	}
	onePeerMessages := logMessages(logs)

	if len(noPeersMessages) == 0 {
		t.Fatal("expected at least one log message from RegisterToBackend(no peers)")
	}
	if !equalStringSlices(noPeersMessages, onePeerMessages) {
		t.Errorf("log message sequence diverges between no-peers and "+
			"single-entry-peers:\n  no-peers:    %v\n  single-entry: %v",
			noPeersMessages, onePeerMessages)
	}
}

func logMessages(logs *observer.ObservedLogs) []string {
	entries := logs.TakeAll()
	msgs := make([]string, len(entries))
	for i, e := range entries {
		msgs[i] = e.Message
	}
	return msgs
}

func equalStringSlices(a, b []string) bool {
	if len(a) != len(b) {
		return false
	}
	for i := range a {
		if a[i] != b[i] {
			return false
		}
	}
	return true
}

// ════════════════════════════════════════════════════════════════════════
// Structural regression guard: source config structs are package-level
// (source-text scan of the real, current on-disk clib.go -- same genre as
// scheduler's/worker's structural regression tests for the identical
// use-after-scope hazard).
// ════════════════════════════════════════════════════════════════════════

func TestClibSource_ClusterAndClientConfigsArePackageLevel(t *testing.T) {
	src, err := os.ReadFile("clib.go")
	if err != nil {
		t.Fatalf("could not read clib.go source for scan: %v", err)
	}
	text := string(src)

	if !strings.Contains(text, "var gClusterCfg C.xce_cluster_config_t") {
		t.Error("gClusterCfg must be declared as a package-level var " +
			"(`var gClusterCfg C.xce_cluster_config_t`) -- a function-local " +
			"variable would go out of scope while xce_cluster_config_to_client_config()'s " +
			"aliased pointers (and xcecli_ctx_init()'s retained ctx->conf pointer) " +
			"are still in use, a use-after-scope bug (see tasks.md's Implementation " +
			"Notes for task 2.1)")
	}
	if !strings.Contains(text, "var gClientCfg C.xce_client_config_t") {
		t.Error("gClientCfg must be declared as a package-level var " +
			"(`var gClientCfg C.xce_client_config_t`) -- xcecli_ctx_init() " +
			"stores a raw pointer to it (ctx->conf) dereferenced for the ctx's " +
			"entire runtime life, not only the initial connect")
	}

	// Exactly one xce_backend_init() call site: the former per-peer loop
	// called it once per candidate (a real loop, not a single call), and the
	// former no-Peers branch duplicated it a second time in the same
	// function. A single call site proves both are gone.
	first := strings.Index(text, "C.xce_backend_init(")
	if first < 0 {
		t.Fatal("C.xce_backend_init(...) call site not found in clib.go")
	}
	if strings.Index(text[first+1:], "C.xce_backend_init(") >= 0 {
		t.Error("clib.go must contain exactly ONE C.xce_backend_init() call site -- " +
			"more than one means the deleted per-peer loop or the separate " +
			"no-Peers branch is still present")
	}

	if strings.Contains(text, "func tryConnect(") {
		t.Error("tryConnect() must be deleted entirely -- its per-call " +
			"C.CString(host)/C.CString(from)/C.CString(key) + deferred C.free " +
			"immediately after a single xce_backend_init() call left ctx->conf's " +
			"aliased masters[]/from/key pointers dangling for the ctx's remaining " +
			"runtime life (see status report CONCERNS for the precise finding)")
	}
}