felhom-controller/controller/internal/stacks/infra_tunnel_test.go

package stacks

import (
	"io"
	"log"
	"os"
	"path/filepath"
	"runtime"
	"strings"
	"testing"

	"gitea.dooplex.hu/admin/felhom-controller/internal/config"
	"gitea.dooplex.hu/admin/felhom-controller/internal/infra"
)

// R-753 (`09` §3 decision 63, Part A): the base stack moves cloudflared onto its own network at a fixed address and
// makes traefik trust forwarded headers from that address only. These tests drive EnsureBaseStack's pieces against a
// STUB docker on PATH (never the real one — R-650) and a recorded compose seam, and assert the consequence on disk and
// in the commands run.

// stubDocker writes a fake `docker` into a temp dir on PATH. State lives in files under state/:
//
//	running-<name>   → `docker inspect --format {{.State.Running}} <name>` prints true
//	nets-<name>      → the networks `containerOnNetwork` sees (one per line)
//	net-<network>    → `docker network inspect --format … <network>` prints the file (else exit 1)
//
// `docker network create … --subnet S … <network>` writes net-<network> = S unless state/create-fails exists.
// Every call is appended to state/calls.
func stubDocker(t *testing.T) (state string) {
	t.Helper()
	if runtime.GOOS == "windows" {
		t.Skip("the stub docker is a shell script")
	}
	dir := t.TempDir()
	state = filepath.Join(dir, "state")
	if err := os.MkdirAll(state, 0o755); err != nil {
		t.Fatal(err)
	}
	script := `#!/bin/sh
S="` + state + `"
echo "$*" >> "$S/calls"
case "$1" in
inspect)
  last=""; for a in "$@"; do last="$a"; done
  case "$*" in
  *State.Running*) if [ -f "$S/running-$last" ]; then echo true; else echo false; fi; exit 0;;
  *NetworkSettings.Networks*) [ -f "$S/nets-$last" ] && cat "$S/nets-$last"; exit 0;;
  esac;;
network)
  last=""; for a in "$@"; do last="$a"; done
  case "$2" in
  inspect) if [ -f "$S/net-$last" ]; then cat "$S/net-$last"; exit 0; fi; echo "Error: no such network: $last" >&2; exit 1;;
  create)
    if [ -f "$S/create-fails" ]; then echo "Error response from daemon: Pool overlaps with other one on this address space" >&2; exit 1; fi
    sub=""; prev=""; for a in "$@"; do [ "$prev" = "--subnet" ] && sub="$a"; prev="$a"; done
    if [ -n "$sub" ]; then echo "$sub" > "$S/net-$last"; else echo "auto" > "$S/net-$last"; fi; exit 0;;
  connect) exit 0;;
  esac;;
esac
exit 0
`
	if err := os.WriteFile(filepath.Join(dir, "docker"), []byte(script), 0o755); err != nil {
		t.Fatal(err)
	}
	t.Setenv("PATH", dir+string(os.PathListSeparator)+os.Getenv("PATH"))
	return state
}

func touch(t *testing.T, path, content string) {
	t.Helper()
	if err := os.WriteFile(path, []byte(content), 0o644); err != nil {
		t.Fatal(err)
	}
}

type composeCall struct {
	dir  string
	args string
}

func tunnelTestManager(t *testing.T, email string) (*Manager, *[]composeCall) {
	t.Helper()
	cfg := &config.Config{}
	cfg.Customer.Email = email
	cfg.Infrastructure.CFTunnelToken = "tok-test"
	m := &Manager{cfg: cfg, logger: log.New(io.Discard, "", 0)}
	var calls []composeCall
	m.composeExecFn = func(dir string, env map[string]string, args ...string) (string, error) {
		calls = append(calls, composeCall{dir: dir, args: strings.Join(args, " ")})
		return "", nil
	}
	return m, &calls
}

// The network is created with its FIXED subnet — the address traefik trusts must be one docker cannot hand elsewhere.
func TestEnsureTunnelNetwork_CreatesFixedSubnet(t *testing.T) {
	state := stubDocker(t)
	m, _ := tunnelTestManager(t, "")
	if err := m.ensureTunnelNetwork(); err != nil {
		t.Fatalf("ensureTunnelNetwork: %v", err)
	}
	got, _ := os.ReadFile(filepath.Join(state, "net-"+infra.TunnelNetwork))
	if strings.TrimSpace(string(got)) != infra.TunnelSubnet {
		t.Fatalf("network created with subnet %q, want %s", got, infra.TunnelSubnet)
	}
	calls, _ := os.ReadFile(filepath.Join(state, "calls"))
	if !strings.Contains(string(calls), "--gateway "+infra.TunnelGateway) || !strings.Contains(string(calls), "--ip-range "+infra.TunnelIPRange) {
		t.Fatalf("create did not fix the gateway: %s", calls)
	}
}

// A network of that name with another subnet is NOT trusted and NOT touched: an error, and the caller keeps the old shape.
func TestEnsureTunnelNetwork_WrongSubnetIsAnError(t *testing.T) {
	state := stubDocker(t)
	touch(t, filepath.Join(state, "net-"+infra.TunnelNetwork), "172.30.0.0/16\n")
	m, _ := tunnelTestManager(t, "")
	err := m.ensureTunnelNetwork()
	if err == nil || !strings.Contains(err.Error(), "172.30.0.0/16") {
		t.Fatalf("want an error naming the wrong subnet, got %v", err)
	}
	calls, _ := os.ReadFile(filepath.Join(state, "calls"))
	if strings.Contains(string(calls), "network create") || strings.Contains(string(calls), " rm ") {
		t.Fatalf("a wrong-subnet network must be left alone; calls: %s", calls)
	}
}

// THE CONSEQUENCE on a box that already runs traefik (every installed box): the new release rewrites traefik.yml with the
// tunnel trust and RECREATES traefik (static config is read only at start). A second tick with nothing changed does nothing.
func TestEnsureTraefik_ReconcilesARunningTraefik(t *testing.T) {
	state := stubDocker(t)
	touch(t, filepath.Join(state, "running-traefik"), "")
	m, calls := tunnelTestManager(t, "owner@example.com")
	dir := t.TempDir()
	old, err := infra.RenderTraefik(infra.TraefikData{ACMEEmail: "owner@example.com"})
	if err != nil {
		t.Fatal(err)
	}
	// the pre-R-753 file: no trust, no middleware
	oldYML := strings.Replace(old["traefik.yml"].Content, "    http:\n      middlewares:\n        - "+infra.ForwardedMiddleware+"@file\n", "    http:\n", 1)
	touch(t, filepath.Join(dir, "traefik.yml"), oldYML)
	touch(t, filepath.Join(dir, "docker-compose.yml"), old["docker-compose.yml"].Content)

	if err := m.ensureTraefik(dir, true); err != nil {
		t.Fatalf("ensureTraefik: %v", err)
	}
	yml, _ := os.ReadFile(filepath.Join(dir, "traefik.yml"))
	if !strings.Contains(string(yml), `- "`+infra.TunnelAddr+`/32"`) {
		t.Fatalf("traefik.yml was not rewritten with the tunnel trust:\n%s", yml)
	}
	cmp, _ := os.ReadFile(filepath.Join(dir, "docker-compose.yml"))
	if !strings.Contains(string(cmp), "ipv4_address: "+infra.TunnelTraefikAddr) {
		t.Fatalf("traefik's compose does not join %s:\n%s", infra.TunnelNetwork, cmp)
	}
	if len(*calls) != 1 || (*calls)[0].args != "up -d --force-recreate" {
		t.Fatalf("want ONE `up -d --force-recreate`, got %+v", *calls)
	}
	if err := m.ensureTraefik(dir, true); err != nil {
		t.Fatalf("second ensureTraefik: %v", err)
	}
	if len(*calls) != 1 {
		t.Fatalf("an unchanged config must not recreate traefik again; calls %+v", *calls)
	}
}

// A rewrite that would DROP the certificate resolver the running file has is refused (no e-mail in the config).
func TestEnsureTraefik_RefusesToDropTheCertResolver(t *testing.T) {
	state := stubDocker(t)
	touch(t, filepath.Join(state, "running-traefik"), "")
	m, calls := tunnelTestManager(t, "") // no customer e-mail → the render has no resolver
	dir := t.TempDir()
	withACME, _ := infra.RenderTraefik(infra.TraefikData{ACMEEmail: "owner@example.com"})
	touch(t, filepath.Join(dir, "traefik.yml"), withACME["traefik.yml"].Content)
	if err := m.ensureTraefik(dir, true); err != nil {
		t.Fatalf("ensureTraefik: %v", err)
	}
	yml, _ := os.ReadFile(filepath.Join(dir, "traefik.yml"))
	if string(yml) != withACME["traefik.yml"].Content || len(*calls) != 0 {
		t.Fatalf("the running file with a resolver must be left alone; calls %+v", *calls)
	}
}

// cloudflared moves to the tunnel network at the fixed address — and is recreated by compose — only when asked.
func TestEnsureCloudflared_MovesToTheTunnelNetwork(t *testing.T) {
	state := stubDocker(t)
	touch(t, filepath.Join(state, "running-cloudflared"), "")
	m, calls := tunnelTestManager(t, "")
	dir := t.TempDir()
	old, _ := infra.RenderCloudflared(infra.CloudflaredData{CFTunnelToken: "tok-test"})
	touch(t, filepath.Join(dir, "docker-compose.yml"), old["docker-compose.yml"].Content)

	if err := m.ensureCloudflared(dir, false); err != nil || len(*calls) != 0 {
		t.Fatalf("unchanged old shape must do nothing; err %v calls %+v", err, *calls)
	}
	if err := m.ensureCloudflared(dir, true); err != nil {
		t.Fatalf("ensureCloudflared: %v", err)
	}
	cmp, _ := os.ReadFile(filepath.Join(dir, "docker-compose.yml"))
	if !strings.Contains(string(cmp), "ipv4_address: "+infra.TunnelAddr) || strings.Contains(string(cmp), "traefik-public") {
		t.Fatalf("cloudflared must be ALONE on %s at %s:\n%s", infra.TunnelNetwork, infra.TunnelAddr, cmp)
	}
	if len(*calls) != 1 || (*calls)[0].args != "up -d" {
		t.Fatalf("want one `up -d`, got %+v", *calls)
	}
}

// The whole bring-up, in order: network → the header clean-up file → traefik (recreated with the trust) → cloudflared
// moved only because traefik is on the tunnel network. And when the network cannot be made, NOTHING moves and traefik
// keeps trusting nobody.
func TestEnsureBaseStack_TunnelOrder(t *testing.T) {
	for _, tc := range []struct {
		name       string
		createFail bool
	}{{"network made", false}, {"network refused", true}} {
		t.Run(tc.name, func(t *testing.T) {
			state := stubDocker(t)
			touch(t, filepath.Join(state, "net-traefik-public"), "172.18.0.0/16\n")
			for _, c := range []string{"traefik", "cloudflared", "filebrowser", "felhom-controller"} {
				touch(t, filepath.Join(state, "running-"+c), "")
			}
			if tc.createFail {
				touch(t, filepath.Join(state, "create-fails"), "")
				touch(t, filepath.Join(state, "nets-traefik"), "traefik-public\n")
			} else {
				touch(t, filepath.Join(state, "nets-traefik"), "traefik-public\n"+infra.TunnelNetwork+"\n")
			}
			touch(t, filepath.Join(state, "nets-felhom-controller"), "traefik-public\n")
			m, calls := tunnelTestManager(t, "owner@example.com")
			m.cfg.Paths.StacksDir = t.TempDir()
			_ = m.EnsureBaseStack()

			traefikDir := filepath.Join(m.cfg.Paths.StacksDir, "traefik")
			if _, err := os.Stat(filepath.Join(traefikDir, "dynamic", "forwarded.yml")); err != nil {
				t.Fatalf("the forwarded-header file must be written in either case: %v", err)
			}
			yml, _ := os.ReadFile(filepath.Join(traefikDir, "traefik.yml"))
			cf, _ := os.ReadFile(filepath.Join(m.cfg.Paths.StacksDir, "cloudflared", "docker-compose.yml"))
			trusts := strings.Contains(string(yml), "trustedIPs")
			moved := strings.Contains(string(cf), "ipv4_address: "+infra.TunnelAddr)
			if tc.createFail {
				if trusts || moved {
					t.Fatalf("no network → no trust and no move; trust %v moved %v", trusts, moved)
				}
				return
			}
			if !trusts || !moved {
				t.Fatalf("network made → traefik trusts the tunnel and cloudflared moved; trust %v moved %v", trusts, moved)
			}
			var order []string
			for _, c := range *calls {
				order = append(order, filepath.Base(c.dir)+":"+c.args)
			}
			if len(order) < 2 || order[0] != "traefik:up -d --force-recreate" || order[1] != "cloudflared:up -d" {
				t.Fatalf("traefik must be recreated BEFORE cloudflared moves; compose calls %v", order)
			}
		})
	}
}