test(syncfs): lock in FS family + fd kind classification

Audit of syncfs(2) confirmed the existing tracing is correct: single fd
arg (fd=args[0], KindFd), FamilyFS like its fsync/fdatasync/
sync_file_range siblings, and an int 0/-1 return that stays Unclassified
(plain ret_event). No code or generated artifacts changed.

Add lock-in tests so a stray reclassification trips CI:
- TestClassifySyncFamilyFdSyscallsByName: enter -> KindFd for the
  fsync/fdatasync/syncfs/sync_file_range group.
- TestClassifyExitSyncfs: exit -> KindRet.
- sync-family FamilyFS assertions in TestClassifySyncallFamily.
- syncfs added to the ret-UNCLASSIFIED list.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

1 files changed, 4 insertions, 0 deletions

diff --git a/internal/generate/retclassify_test.go b/internal/generate/retclassify_test.go
index 70ade1a..148c1e6 100644
--- a/internal/generate/retclassify_test.go
+++ b/internal/generate/retclassify_test.go
@@ -44,6 +44,10 @@ func TestClassifyRetUnclassified(t *testing.T) {
 		"openat", "close", "rename", "unlink", "fcntl", "dup", "dup2", "dup3",
 		"mkdir", "rmdir", "chmod", "chown", "chdir", "stat", "lseek",
 		"truncate", "fallocate", "mmap", "fsync", "flock", "recvmmsg", "sendmmsg",
+		// syncfs(2) returns int 0/-1 (no byte count); it commits the filesystem
+		// containing args[0]'s fd and transfers no bytes, so its exit must stay
+		// UNCLASSIFIED (plain ret_event), like its fsync/fdatasync siblings.
+		"syncfs",
 		// gettimeofday(2) returns int 0/-1 (no byte count); its exit carries a
 		// plain ret_event and must stay UNCLASSIFIED, not a read/write transfer.
 		"gettimeofday",