fix(xdebug-trace): pid-named temp dir, remove dirs of dead processes

cmd/buggregator/main.go

@@ -125,7 +125,7 @@ func main() {
 		// The module is optional: without a dir it stays off, the rest runs.
 		traceDir := filepath.Join(cfg.Storage.Path, "xdebug-traces")
 		if cfg.Storage.Mode != "filesystem" {
-			tmp, err := os.MkdirTemp("", "buggregator-xdebug-traces-")
+			tmp, err := xdebugtrace.TempDir("")
 			if err != nil {
 				slog.Error("xdebug-trace: cannot create temp dir; module disabled", "err", err)
 				traceDir = ""

modules/xdebugtrace/process_unix.go

@@ -0,0 +1,15 @@
+//go:build !windows
+
+package xdebugtrace
+
+import (
+	"errors"
+	"syscall"
+)
+
+// processAlive reports whether a process with pid exists (EPERM: it does,
+// owned by another user).
+func processAlive(pid int) bool {
+	err := syscall.Kill(pid, 0)
+	return err == nil || errors.Is(err, syscall.EPERM)
+}

modules/xdebugtrace/process_windows.go

@@ -0,0 +1,21 @@
+//go:build windows
+
+package xdebugtrace
+
+import (
+	"errors"
+	"os"
+	"syscall"
+)
+
+// processAlive reports whether a process with pid exists: FindProcess
+// opens a handle on Windows and fails for a pid that is gone. Access
+// denied (an elevated process seen from a plain one) means it exists.
+func processAlive(pid int) bool {
+	p, err := os.FindProcess(pid)
+	if err != nil {
+		return errors.Is(err, syscall.ERROR_ACCESS_DENIED)
+	}
+	p.Release()
+	return true
+}

modules/xdebugtrace/tempdir.go

@@ -0,0 +1,45 @@
+package xdebugtrace
+
+import (
+	"log/slog"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+)
+
+const tempPrefix = "buggregator-xdebug-traces-"
+
+// TempDir makes the trace dir for memory storage mode: <base>/
+// buggregator-xdebug-traces-<pid>. Dirs of processes that are gone (a
+// crash, kill -9 skips the cleanup at exit) are removed first, so stale
+// traces never pile up. base "" = os.TempDir().
+func TempDir(base string) (string, error) {
+	if base == "" {
+		base = os.TempDir()
+	}
+	self := os.Getpid()
+	if entries, err := os.ReadDir(base); err == nil {
+		for _, e := range entries {
+			name := e.Name()
+			if !e.IsDir() || !strings.HasPrefix(name, tempPrefix) {
+				continue
+			}
+			pid, err := strconv.Atoi(strings.TrimPrefix(name, tempPrefix))
+			if err != nil || pid <= 0 || (pid != self && processAlive(pid)) {
+				continue
+			}
+			// Our own pid here is a leftover of an earlier process. Two
+			// containers sharing one /tmp volume are both pid 1: give each
+			// its own TMPDIR.
+			if err := os.RemoveAll(filepath.Join(base, name)); err == nil {
+				slog.Info("xdebug-trace: removed stale temp dir", "dir", name)
+			}
+		}
+	}
+	dir := filepath.Join(base, tempPrefix+strconv.Itoa(self))
+	if err := os.Mkdir(dir, 0o700); err != nil {
+		return "", err
+	}
+	return dir, nil
+}

modules/xdebugtrace/tempdir_test.go

@@ -0,0 +1,60 @@
+package xdebugtrace
+
+import (
+	"os"
+	"os/exec"
+	"path/filepath"
+	"sort"
+	"strconv"
+	"testing"
+)
+
+func TestTempDir_RemovesDeadProcessDirs(t *testing.T) {
+	base := t.TempDir()
+
+	// A pid that is surely gone: a finished child.
+	cmd := exec.Command(os.Args[0], "-test.run=^$")
+	if err := cmd.Run(); err != nil {
+		t.Fatal(err)
+	}
+	dead := cmd.Process.Pid
+	alive := os.Getppid() // the go test runner
+
+	mk := func(name string) {
+		if err := os.MkdirAll(filepath.Join(base, name, "sub"), 0o755); err != nil {
+			t.Fatal(err)
+		}
+	}
+	mk(tempPrefix + strconv.Itoa(dead))
+	mk(tempPrefix + strconv.Itoa(alive))
+	mk(tempPrefix + strconv.Itoa(os.Getpid())) // own leftover
+	mk(tempPrefix + "abc")
+	mk("other-dir")
+
+	dir, err := TempDir(base)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if dir != filepath.Join(base, tempPrefix+strconv.Itoa(os.Getpid())) {
+		t.Errorf("dir = %s", dir)
+	}
+	if _, err := os.Stat(filepath.Join(dir, "sub")); !os.IsNotExist(err) {
+		t.Error("own leftover dir not emptied")
+	}
+	entries, _ := os.ReadDir(base)
+	var got []string
+	for _, e := range entries {
+		got = append(got, e.Name())
+	}
+	sort.Strings(got)
+	want := []string{tempPrefix + "abc", tempPrefix + strconv.Itoa(alive), tempPrefix + strconv.Itoa(os.Getpid()), "other-dir"}
+	sort.Strings(want)
+	if len(got) != len(want) {
+		t.Fatalf("left %v, want %v", got, want)
+	}
+	for i := range want {
+		if got[i] != want[i] {
+			t.Fatalf("left %v, want %v", got, want)
+		}
+	}
+}