narratio/internal/adapters/subprocess/run.go

package subprocess

import (
	"context"
	"errors"
	"fmt"
	"io"
	"os"
	"os/exec"
	"path/filepath"
	"sort"
	"strings"
	"time"

	"gopkg.in/yaml.v3"
)

// RunRequest defines a subprocess invocation.
type RunRequest struct {
	Executable    string
	Args          []string
	WorkingDir    string
	EnvOverrides  map[string]string
	Timeout       time.Duration
	StdoutLogPath string
	StderrLogPath string
}

// RunResult captures subprocess execution details.
type RunResult struct {
	ExitCode      int
	StartedAt     time.Time
	CompletedAt   time.Time
	Duration      time.Duration
	StdoutLogPath string
	StderrLogPath string
	TimedOut      bool
	Canceled      bool
}

// Run executes a subprocess with context cancellation and optional timeout.
func Run(ctx context.Context, req RunRequest) (RunResult, error) {
	if strings.TrimSpace(req.Executable) == "" {
		return RunResult{}, fmt.Errorf("subprocess executable is required")
	}
	if ctx == nil {
		ctx = context.Background()
	}

	runCtx := ctx
	cancel := func() {}
	if req.Timeout > 0 {
		runCtx, cancel = context.WithTimeout(ctx, req.Timeout)
	}
	defer cancel()

	logs, err := openLogWriters(req.StdoutLogPath, req.StderrLogPath)
	if err != nil {
		return RunResult{}, err
	}
	defer logs.Close()

	cmd := exec.CommandContext(runCtx, req.Executable, req.Args...)
	cmd.Dir = req.WorkingDir
	cmd.Env = mergeEnv(os.Environ(), req.EnvOverrides)
	cmd.Stdout = logs.Stdout
	cmd.Stderr = logs.Stderr

	started := time.Now().UTC()
	result := RunResult{
		ExitCode:      -1,
		StartedAt:     started,
		StdoutLogPath: req.StdoutLogPath,
		StderrLogPath: req.StderrLogPath,
	}

	if err := cmd.Start(); err != nil {
		result.CompletedAt = time.Now().UTC()
		result.Duration = result.CompletedAt.Sub(result.StartedAt)
		return result, fmt.Errorf("start command %q with args %v: %w", req.Executable, req.Args, err)
	}

	waitErr := cmd.Wait()
	result.CompletedAt = time.Now().UTC()
	result.Duration = result.CompletedAt.Sub(result.StartedAt)
	if cmd.ProcessState != nil {
		result.ExitCode = cmd.ProcessState.ExitCode()
	}

	ctxErr := runCtx.Err()
	if errors.Is(ctxErr, context.DeadlineExceeded) {
		result.TimedOut = true
	}
	if errors.Is(ctxErr, context.Canceled) && !result.TimedOut {
		result.Canceled = true
	}

	if waitErr == nil {
		return result, nil
	}

	stderrTail := readRedactedTail(req.StderrLogPath, req.EnvOverrides, 2048)
	diagnostics := buildDiagnostics(req, result, stderrTail)

	if result.TimedOut {
		return result, fmt.Errorf("command timed out after %s (%s)", req.Timeout, diagnostics)
	}
	if result.Canceled {
		return result, fmt.Errorf("command canceled (%s)", diagnostics)
	}
	if exitErr, ok := waitErr.(*exec.ExitError); ok {
		return result, fmt.Errorf("command failed with exit code %d (%s): %w", exitErr.ExitCode(), diagnostics, waitErr)
	}

	return result, fmt.Errorf("command failed to run (%s): %w", diagnostics, waitErr)
}

// WriteYAMLAtomic marshals value as YAML and atomically writes it to path.
func WriteYAMLAtomic(path string, value any, perm os.FileMode) error {
	data, err := yaml.Marshal(value)
	if err != nil {
		return fmt.Errorf("marshal yaml for %q: %w", path, err)
	}
	if err := WriteFileAtomic(path, data, perm); err != nil {
		return fmt.Errorf("write yaml %q: %w", path, err)
	}
	return nil
}

// WriteFileAtomic writes bytes via same-directory temp file + atomic rename.
func WriteFileAtomic(path string, data []byte, perm os.FileMode) error {
	if strings.TrimSpace(path) == "" {
		return fmt.Errorf("write file: path is required")
	}

	dir := filepath.Dir(path)
	if err := os.MkdirAll(dir, 0o755); err != nil {
		return fmt.Errorf("create parent directory %q: %w", dir, err)
	}

	base := filepath.Base(path)
	tmp, err := os.CreateTemp(dir, "."+base+".tmp-*")
	if err != nil {
		return fmt.Errorf("create temp file: %w", err)
	}
	tmpPath := tmp.Name()
	removeTmp := true
	defer func() {
		if removeTmp {
			_ = os.Remove(tmpPath)
		}
	}()

	if _, err := tmp.Write(data); err != nil {
		_ = tmp.Close()
		return fmt.Errorf("write temp file: %w", err)
	}
	if err := tmp.Sync(); err != nil {
		_ = tmp.Close()
		return fmt.Errorf("sync temp file: %w", err)
	}
	if err := tmp.Close(); err != nil {
		return fmt.Errorf("close temp file: %w", err)
	}
	if err := os.Chmod(tmpPath, perm); err != nil {
		return fmt.Errorf("chmod temp file: %w", err)
	}
	if err := os.Rename(tmpPath, path); err != nil {
		return fmt.Errorf("rename temp file: %w", err)
	}
	removeTmp = false
	return nil
}

type logWriters struct {
	files  []*os.File
	Stdout io.Writer
	Stderr io.Writer
}

func (l *logWriters) Close() {
	for _, f := range l.files {
		_ = f.Close()
	}
}

func openLogWriters(stdoutPath, stderrPath string) (*logWriters, error) {
	cleanStdout := cleanLogPath(stdoutPath)
	cleanStderr := cleanLogPath(stderrPath)

	// Keep stdout/stderr on the same file descriptor when both paths target
	// the same file to avoid descriptor aliasing surprises across runtimes.
	if cleanStdout != "" && cleanStdout == cleanStderr {
		f, err := openLogFile(cleanStdout)
		if err != nil {
			return nil, fmt.Errorf("open shared stdout/stderr log %q: %w", cleanStdout, err)
		}
		return &logWriters{
			files:  []*os.File{f},
			Stdout: f,
			Stderr: f,
		}, nil
	}

	stdoutFile, stdoutWriter, err := logWriter(cleanStdout)
	if err != nil {
		return nil, fmt.Errorf("open stdout log: %w", err)
	}
	stderrFile, stderrWriter, err := logWriter(cleanStderr)
	if err != nil {
		closeFile(stdoutFile)
		return nil, fmt.Errorf("open stderr log: %w", err)
	}

	files := make([]*os.File, 0, 2)
	if stdoutFile != nil {
		files = append(files, stdoutFile)
	}
	if stderrFile != nil {
		files = append(files, stderrFile)
	}
	return &logWriters{
		files:  files,
		Stdout: stdoutWriter,
		Stderr: stderrWriter,
	}, nil
}

func cleanLogPath(path string) string {
	trimmed := strings.TrimSpace(path)
	if trimmed == "" {
		return ""
	}
	return filepath.Clean(trimmed)
}

func logWriter(path string) (*os.File, io.Writer, error) {
	if strings.TrimSpace(path) == "" {
		return nil, io.Discard, nil
	}
	f, err := openLogFile(path)
	if err != nil {
		return nil, nil, err
	}
	return f, f, nil
}

func openLogFile(path string) (*os.File, error) {
	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
		return nil, fmt.Errorf("create log directory for %q: %w", path, err)
	}
	f, err := os.Create(path)
	if err != nil {
		return nil, fmt.Errorf("open log file %q: %w", path, err)
	}
	return f, nil
}

func closeFile(f *os.File) {
	if f != nil {
		_ = f.Close()
	}
}

func buildDiagnostics(req RunRequest, result RunResult, stderrTail string) string {
	details := fmt.Sprintf(
		"executable=%q args=%v cwd=%q timeout=%s exit_code=%d timed_out=%t canceled=%t stdout_log=%q stderr_log=%q",
		req.Executable,
		req.Args,
		req.WorkingDir,
		req.Timeout,
		result.ExitCode,
		result.TimedOut,
		result.Canceled,
		req.StdoutLogPath,
		req.StderrLogPath,
	)
	if strings.TrimSpace(stderrTail) == "" {
		if hint := fdDiagnosticsHint(result.ExitCode, ""); hint != "" {
			return details + fmt.Sprintf(" hint=%q", hint)
		}
		return details
	}
	details = details + fmt.Sprintf(" stderr_tail=%q", stderrTail)
	if hint := fdDiagnosticsHint(result.ExitCode, stderrTail); hint != "" {
		details = details + fmt.Sprintf(" hint=%q", hint)
	}
	return details
}

func fdDiagnosticsHint(exitCode int, stderrTail string) string {
	lowerTail := strings.ToLower(stderrTail)
	if strings.Contains(lowerTail, "bad file descriptor") || strings.Contains(lowerTail, "errno 9") {
		return "stderr stream appears invalid in child process (fd 2); check wrapper/subprocess environment differences versus direct shell execution"
	}
	if exitCode == 120 {
		return "exit code 120 may indicate interpreter shutdown stream-flush failures (commonly invalid stdout/stderr descriptors in Python processes)"
	}
	return ""
}

func readRedactedTail(path string, envOverrides map[string]string, maxBytes int64) string {
	if strings.TrimSpace(path) == "" || maxBytes <= 0 {
		return ""
	}
	f, err := os.Open(path)
	if err != nil {
		return ""
	}
	defer f.Close()

	info, err := f.Stat()
	if err != nil {
		return ""
	}
	size := info.Size()
	start := int64(0)
	if size > maxBytes {
		start = size - maxBytes
	}
	if _, err := f.Seek(start, io.SeekStart); err != nil {
		return ""
	}
	data, err := io.ReadAll(f)
	if err != nil {
		return ""
	}
	tail := strings.TrimSpace(string(data))
	if tail == "" {
		return ""
	}
	return redactSensitiveTail(tail, envOverrides)
}

func redactSensitiveTail(tail string, envOverrides map[string]string) string {
	out := tail
	for k, v := range envOverrides {
		if strings.TrimSpace(v) == "" {
			continue
		}
		if looksSensitiveEnvKey(k) {
			out = strings.ReplaceAll(out, v, "<redacted>")
		}
	}
	return out
}

func looksSensitiveEnvKey(key string) bool {
	k := strings.ToUpper(strings.TrimSpace(key))
	return strings.Contains(k, "KEY") ||
		strings.Contains(k, "TOKEN") ||
		strings.Contains(k, "SECRET") ||
		strings.Contains(k, "PASSWORD")
}

func mergeEnv(base []string, overrides map[string]string) []string {
	if len(overrides) == 0 {
		return base
	}

	kv := make(map[string]string, len(base)+len(overrides))
	for _, item := range base {
		k, v, ok := strings.Cut(item, "=")
		if !ok {
			continue
		}
		kv[k] = v
	}
	for k, v := range overrides {
		kv[k] = v
	}

	keys := make([]string, 0, len(kv))
	for k := range kv {
		keys = append(keys, k)
	}
	sort.Strings(keys)

	out := make([]string, 0, len(keys))
	for _, k := range keys {
		out = append(out, k+"="+kv[k])
	}
	return out
}