Fixed issues with telemetry.

ptlinux/cli.py

@@ -9,7 +9,9 @@ from datetime import datetime
 import json
 from pathlib import Path
 import re
+import signal
 import sys
+import threading
 import time
 from typing import Callable, Sequence
 
@@ -1379,6 +1381,8 @@ def _play_locked(
 	)
 	game_started_ns = time.time_ns()
 	game_started_at = time.monotonic()
+	launch_success_lock = threading.Lock()
+	launch_success_reported = False
 	if telemetry is not None and not dry_run:
 		telemetry.send_event(
 			"game_launch_started",
@@ -1399,6 +1403,36 @@ def _play_locked(
 		if depot_file_catalog is not None
 		else None
 	)
+
+	def observe_game_output(line: str) -> str:
+		"""Timestamp launch success when the supervisor first confirms readiness."""
+
+		nonlocal launch_success_reported
+		live_readiness_signal = _game_readiness_signal(line)
+		if telemetry is None or dry_run or live_readiness_signal is None:
+			return line
+		with launch_success_lock:
+			if launch_success_reported:
+				return line
+			ready_at = time.monotonic()
+			launch_success_reported = telemetry.send_event(
+				"game_launch_success",
+				metadata={
+					"launch_duration_ms": round(
+						(ready_at - launch_requested_at) * 1000
+					),
+					"game_start_duration_ms": round(
+						(ready_at - game_started_at) * 1000
+					),
+					"game_window_ready": True,
+					"readiness_signal": live_readiness_signal,
+					"runtime": selection.kind,
+					"realm": start_option.name,
+				},
+				context={"operation": "launch_game", "stage": "ready"},
+			)
+		return line
+
 	try:
 		outcome = run_game(
 			paths,
@@ -1409,6 +1443,7 @@ def _play_locked(
 			use_gamebox=not no_box,
 			concurrent_prefix_launch=reuse_prepared,
 			check=False,
+			output_line_filter=observe_game_output,
 			pass_fds=game_lease_fds,
 		)
 	except KeyboardInterrupt:
@@ -1449,6 +1484,7 @@ def _play_locked(
 		game_started_ns,
 	)
 	runtime_issues = _runtime_issue_summary(outcome.stderr)
+	game_window_ready = runtime_issues["game_window_ready"] is True
 	if (
 		selection.kind == "proton7-container"
 		and runtime_issues["host_futex_fatal_count"]
@@ -1458,8 +1494,13 @@ def _play_locked(
 		runtime_issues["recommended_one_variable_comparison"] = "proton7-container"
 	elif runtime_issues["dbghelp_symdeferred_assertion_count"]:
 		runtime_issues["recommended_one_variable_comparison"] = "umu"
+	clean_exit = outcome.returncode == 0 and current_game_crash is None
+	launch_succeeded = (
+		clean_exit or game_window_ready or current_game_log is not None
+	)
 	report = {
-		"ok": outcome.returncode == 0 and current_game_crash is None,
+		"ok": clean_exit,
+		"launch_succeeded": launch_succeeded,
 		"dry_run": dry_run,
 		"runtime": asdict(selection),
 		"umu_bootstrap": asdict(umu_bootstrap) if umu_bootstrap else None,
@@ -1481,22 +1522,40 @@ def _play_locked(
 		"runtime_issues": runtime_issues,
 	}
 	game_runtime_seconds = max(0.0, time.monotonic() - game_started_at)
-	clean_exit = outcome.returncode == 0 and current_game_crash is None
 	if telemetry is not None and not dry_run:
-		common_game_metadata = {
+		common_game_metadata: dict[str, object] = {
 			"runtime_seconds": round(game_runtime_seconds),
-			"exit_code": outcome.returncode,
 			"clean_exit": clean_exit,
+			"timed_out": outcome.timed_out,
+			"launch_succeeded": launch_succeeded,
+			"game_window_ready": game_window_ready,
+			"crash_dump": current_game_crash is not None,
 			"session_duration_class": _session_duration_class(game_runtime_seconds),
 			"runtime": selection.kind,
 			"realm": start_option.name,
+			**_game_exit_metadata(outcome.returncode, selection.kind),
 		}
-		if clean_exit:
-			telemetry.send_event(
-				"game_launch_success",
-				metadata=common_game_metadata,
-				context={"operation": "launch_game", "stage": "completed"},
-			)
+		if launch_succeeded:
+			if not launch_success_reported:
+				readiness_signal = _game_readiness_signal(outcome.stderr)
+				if readiness_signal is None:
+					readiness_signal = (
+						"game_log"
+						if current_game_log is not None
+						else "clean_process_exit"
+					)
+				telemetry.send_event(
+					"game_launch_success",
+					metadata={
+						"detection_duration_ms": round(game_runtime_seconds * 1000),
+						"detected_after_exit": True,
+						"game_window_ready": game_window_ready,
+						"readiness_signal": readiness_signal,
+						"runtime": selection.kind,
+						"realm": start_option.name,
+					},
+					context={"operation": "launch_game", "stage": "post_exit"},
+				)
 		else:
 			telemetry.send_event(
 				"game_launch_failed",
@@ -1504,17 +1563,57 @@ def _play_locked(
 				message=(
 					"Game client wrote a crash dump"
 					if current_game_crash is not None
+					else "Game session timed out"
+					if outcome.timed_out
 					else f"Game client exited with status {outcome.returncode}"
 				),
-				metadata={**common_game_metadata, "runtime_issues": runtime_issues},
-				context={"operation": "launch_game", "stage": "running"},
+				metadata={
+					**common_game_metadata,
+					"runtime_issues": runtime_issues,
+					"stdout": outcome.stdout,
+					"stderr": outcome.stderr,
+				},
+				context={"operation": "launch_game", "stage": "game_start"},
+			)
+		if launch_succeeded:
+			exit_message = None
+			if current_game_crash is not None:
+				exit_message = "Game client wrote a crash dump"
+			elif outcome.timed_out:
+				exit_message = "Game session timed out"
+			elif common_game_metadata.get("signal_inferred") is True:
+				exit_message = (
+					"Game session ended with status "
+					f"{common_game_metadata['raw_exit_code']}, consistent with "
+					f"{common_game_metadata['signal_name']} wrapped by UMU"
+				)
+			elif "signal" in common_game_metadata:
+				exit_message = (
+					"Game session was terminated by "
+					f"{common_game_metadata['signal_name']}"
+				)
+			elif outcome.returncode not in {0, None}:
+				exit_message = f"Game session ended with status {outcome.returncode}"
+			telemetry.send_event(
+				"game_exited",
+				severity=(
+					"error"
+					if current_game_crash is not None
+					else "info" if clean_exit else "warning"
+				),
+				message=exit_message,
+				metadata=(
+					common_game_metadata
+					if clean_exit
+					else {
+						**common_game_metadata,
+						"runtime_issues": runtime_issues,
+						"stdout": outcome.stdout,
+						"stderr": outcome.stderr,
+					}
+				),
+				context={"operation": "launch_game", "stage": "exited"},
 			)
-		telemetry.send_event(
-			"game_exited",
-			severity="info" if clean_exit else "warning",
-			metadata=common_game_metadata,
-			context={"operation": "launch_game", "stage": "exited"},
-		)
 	if as_json or dry_run:
 		_emit(report, as_json)
 	else:
@@ -1535,6 +1634,49 @@ def _session_duration_class(duration_seconds: float) -> str:
 	return "normal"
 
 
+def _game_exit_metadata(
+	returncode: int | None,
+	runtime_kind: str,
+) -> dict[str, object]:
+	metadata: dict[str, object] = {"exit_code": returncode}
+	if returncode is None:
+		return metadata
+	if returncode < 0:
+		signal_number = -returncode
+		metadata.update(
+			{
+				"exit_code": 128 + min(signal_number, 127),
+				"raw_exit_code": returncode,
+				"signal": signal_number,
+				"signal_name": _signal_name(signal_number),
+			}
+		)
+		return metadata
+
+	# UMU returns Popen's negative child status from a Python main function.
+	# SystemExit encodes -SIGKILL as 247 before our Python gamebox sees it.
+	wrapped_sigkill = 256 - int(signal.SIGKILL)
+	if runtime_kind in {"umu", "proton7-container"} and returncode == wrapped_sigkill:
+		metadata.update(
+			{
+				"exit_code": 128 + int(signal.SIGKILL),
+				"raw_exit_code": returncode,
+				"signal": int(signal.SIGKILL),
+				"signal_name": signal.SIGKILL.name,
+				"signal_inferred": True,
+				"signal_inference": "umu_python_system_exit",
+			}
+		)
+	return metadata
+
+
+def _signal_name(signal_number: int) -> str:
+	try:
+		return signal.Signals(signal_number).name
+	except ValueError:
+		return f"signal {signal_number}"
+
+
 def _ensure_portable_umu_for_play(
 	paths: AppPaths,
 	settings: Settings,
@@ -1684,6 +1826,7 @@ def _runtime_issue_summary(stderr: str) -> dict[str, object]:
 		}
 	)
 	return {
+		"game_window_ready": _game_window_was_ready(folded),
 		"host_futex_fatal_count": folded.count(futex_message),
 		"gstreamer_wrong_elf_class_count": folded.count(wrong_elf_message),
 		"gstreamer_missing_library_count": len(missing_libraries),
@@ -1696,6 +1839,25 @@ def _runtime_issue_summary(stderr: str) -> dict[str, object]:
 	}
 
 
+def _game_window_was_ready(stderr: str) -> bool:
+	return _game_readiness_signal(stderr) is not None
+
+
+def _game_readiness_signal(output: str) -> str | None:
+	folded = output.casefold()
+	for signal_name, marker in (
+		("window_ready", "game window ready:"),
+		("rendered_frame", "rendered game frame:"),
+		(
+			"window_supervision",
+			"game window resizing is disabled to prevent repeated direct3d device resets",
+		),
+	):
+		if marker in folded:
+			return signal_name
+	return None
+
+
 def _print_game_outcome(
 	outcome: CommandOutcome,
 	current_game_log: dict[str, object] | None,

ptlinux/command.py

@@ -110,6 +110,7 @@ class CommandRunner:
 		output_line_filter: OutputLineFilter | None = None,
 		pass_fds: Sequence[int] = (),
 		telemetry_stage: str | None = None,
+		report_failure_telemetry: bool = True,
 	) -> CommandOutcome:
 		arguments = [str(value) for value in argv]
 		if not arguments:
@@ -179,7 +180,9 @@ class CommandRunner:
 					dry_run=False,
 				)
 		self._record(outcome, cwd, overrides)
-		if outcome.timed_out or outcome.returncode != 0:
+		if report_failure_telemetry and (
+			outcome.timed_out or outcome.returncode != 0
+		):
 			telemetry = current_telemetry()
 			if telemetry is not None:
 				telemetry.report_subprocess_failure(

ptlinux/telemetry.py

@@ -34,6 +34,8 @@ DEFAULT_TELEMETRY_ENDPOINT = "https://pristontale.eu/api/launcher/telemetry"
 TELEMETRY_ENDPOINT_VARIABLE = "PTLINUX_TELEMETRY_ENDPOINT"
 TELEMETRY_KEY_VARIABLE = "PTLINUX_TELEMETRY_KEY"
 TELEMETRY_DEBUG_VARIABLE = "PTLINUX_TELEMETRY_DEBUG"
+TELEMETRY_PROCESS_ROLE_VARIABLE = "PTLINUX_TELEMETRY_PROCESS_ROLE"
+TELEMETRY_OPERATION_PROCESS_ROLE = "operation"
 
 MAX_REQUEST_BYTES = 64 * 1024
 MAX_MESSAGE_BYTES = 10 * 1024
@@ -263,9 +265,14 @@ class TelemetryService:
 		self._session_metadata = dict(session_metadata or {})
 		self._monotonic = monotonic
 		self._started_at = monotonic()
+		self._reports_launcher_lifecycle = (
+			self._environment.get(TELEMETRY_PROCESS_ROLE_VARIABLE, "").strip().casefold()
+			!= TELEMETRY_OPERATION_PROCESS_ROLE
+		)
 		self._enabled = bool(enabled)
 		self._active = False
 		self._accepting = False
+		self._closing = False
 		self._ready = False
 		self._first_run = False
 		self._installation_id: str | None = None
@@ -322,6 +329,7 @@ class TelemetryService:
 				self._base_payload = self._collect_common_payload(first_seen)
 				self._stop.clear()
 				self._accepting = True
+				self._closing = False
 				self._active = True
 				worker = threading.Thread(
 					target=self._worker_loop,
@@ -338,15 +346,16 @@ class TelemetryService:
 				self._debug_log(f"initialization failed: {type(exc).__name__}")
 				return
 
-		self.send_event(
-			"launcher_started",
-			metadata={"cold_start": True, **self._session_metadata},
-		)
-		if self._first_run:
+		if self._reports_launcher_lifecycle:
 			self.send_event(
-				"first_run_started",
-				metadata={"install_type": self._install_type()},
+				"launcher_started",
+				metadata={"cold_start": True, **self._session_metadata},
 			)
+			if self._first_run:
+				self.send_event(
+					"first_run_started",
+					metadata={"install_type": self._install_type()},
+				)
 		self._start_environment_collection()
 
 	def set_enabled(self, enabled: bool) -> None:
@@ -368,6 +377,8 @@ class TelemetryService:
 			if self._ready or not self._active:
 				return
 			self._ready = True
+			if not self._reports_launcher_lifecycle:
+				return
 		duration_ms = round((self._monotonic() - self._started_at) * 1000)
 		self.send_event(
 			"launcher_ready",
@@ -674,11 +685,12 @@ class TelemetryService:
 			self._send_repeated_error(state)
 
 	def shutdown(self, timeout_seconds: float = SHUTDOWN_WAIT_SECONDS) -> None:
-		"""Queue the closing event and give the daemon worker a bounded flush."""
+		"""Finish this process role and give the daemon worker a bounded flush."""
 
 		with self._state_lock:
-			if not self._active:
+			if not self._active or self._closing:
 				return
+			self._closing = True
 		deadline = time.monotonic() + max(0.0, timeout_seconds)
 		environment_worker = self._environment_worker
 		if isinstance(environment_worker, threading.Thread) and (
@@ -692,17 +704,18 @@ class TelemetryService:
 			)
 		self._flush_repeated_errors()
 		runtime_seconds = max(0, round(self._monotonic() - self._started_at))
-		if self._first_run and not self._ready:
+		if self._reports_launcher_lifecycle:
+			if self._first_run and not self._ready:
+				self.send_event(
+					"first_run_failed",
+					severity="error",
+					message="Launcher did not reach the ready state",
+					metadata={"duration_ms": runtime_seconds * 1000},
+				)
 			self.send_event(
-				"first_run_failed",
-				severity="error",
-				message="Launcher did not reach the ready state",
-				metadata={"duration_ms": runtime_seconds * 1000},
+				"launcher_closed",
+				metadata={"runtime_seconds": runtime_seconds, **self._session_metadata},
 			)
-		self.send_event(
-			"launcher_closed",
-			metadata={"runtime_seconds": runtime_seconds, **self._session_metadata},
-		)
 		with self._state_lock:
 			self._accepting = False
 			self._stop.set()
@@ -719,6 +732,7 @@ class TelemetryService:
 		with self._state_lock:
 			self._enabled = False
 			self._accepting = False
+			self._closing = True
 			self._stop.set()
 			self._repeated_errors.clear()
 		while True:
@@ -795,6 +809,11 @@ class TelemetryService:
 					self._base_payload[key] = {**existing, **value}
 				else:
 					self._base_payload[key] = value
+			if not self._reports_launcher_lifecycle:
+				# Dashboard operation subprocesses enrich their own events with the
+				# detected environment, while the dashboard remains the sole source
+				# of launcher lifecycle and automatic compatibility events.
+				return
 
 		runtime = sanitized.get("runtime")
 		dependencies = sanitized.get("dependencies")
@@ -816,14 +835,38 @@ class TelemetryService:
 			context={"operation": "launcher", "stage": "environment_detection"},
 		)
 		missing_required = dependency_details.get("missing_required")
-		if dependency_details.get("healthy") is False:
+		generic_dependency_failures = (
+			[
+				str(value)
+				for value in missing_required
+				if isinstance(value, str) and value
+			]
+			if isinstance(missing_required, list)
+			else []
+		)
+		if (
+			dependency_details.get("container_ready") is False
+			and "container_host" not in generic_dependency_failures
+		):
+			generic_dependency_failures.append("container_host")
+		if (
+			dependency_details.get("healthy") is False
+			and generic_dependency_failures
+		):
 			self.send_event(
 				"dependency_check_failed",
 				severity="warning",
 				message="One or more launcher dependencies are unavailable",
 				metadata={
-					"dependencies": (
-						missing_required if isinstance(missing_required, list) else []
+					"dependencies": generic_dependency_failures,
+					"required_shared_libraries_present": dependency_details.get(
+						"required_shared_libraries_present"
+					),
+					"vulkan_64bit_available": dependency_details.get(
+						"vulkan_64bit_available"
+					),
+					"vulkan_32bit_available": dependency_details.get(
+						"vulkan_32bit_available"
 					),
 				},
 				context={"operation": "launcher", "stage": "dependency_check"},
@@ -846,7 +889,17 @@ class TelemetryService:
 			self.send_event(
 				"vulkan_32bit_check_failed",
 				severity="warning",
-				metadata={"dependency": "vulkan_32bit", "check": "32bit_vulkan"},
+				metadata={
+					"dependency": "vulkan_32bit",
+					"check": "32bit_vulkan",
+					"loader_available": runtime_details.get(
+						"vulkan_32bit_loader_available"
+					),
+					"icd_available": runtime_details.get(
+						"vulkan_32bit_icd_available"
+					),
+					"detection": runtime_details.get("vulkan_32bit_detection"),
+				},
 				context={"operation": "launcher", "stage": "vulkan_check"},
 			)
 

ptlinux/telemetry_environment.py

@@ -29,8 +29,11 @@ _VULKAN_MANIFEST_DIRECTORIES = (
 )
 _LIBRARY_ROOTS = (
 	Path("/lib"),
+	Path("/lib32"),
 	Path("/usr/lib"),
+	Path("/usr/lib32"),
 	Path("/usr/local/lib"),
+	Path("/usr/local/lib32"),
 )
 _PCI_DATABASES = (
 	Path("/usr/share/hwdata/pci.ids"),
@@ -198,13 +201,12 @@ def _vulkan_info(
 		if api_version and _VERSION_PATTERN.fullmatch(api_version):
 			api_versions.append(api_version)
 		library = _short_string(icd.get("library_path"))
-		architecture = _manifest_library_architecture(
+		architectures = _manifest_library_architectures(
 			manifest,
 			library,
 			library_roots,
 		)
-		if architecture is not None:
-			manifest_architectures.add(architecture)
+		manifest_architectures.update(architectures)
 
 	vulkaninfo = shutil.which("vulkaninfo", path=environment.get("PATH", ""))
 	probe = (
@@ -240,6 +242,11 @@ def _vulkan_info(
 		"vulkan_available": available_64bit,
 		"vulkan_64bit": available_64bit,
 		"vulkan_32bit": available_32bit,
+		"vulkan_64bit_loader_available": 64 in loader_architectures,
+		"vulkan_64bit_icd_available": 64 in manifest_architectures,
+		"vulkan_64bit_probe_succeeded": functional_64bit,
+		"vulkan_32bit_loader_available": 32 in loader_architectures,
+		"vulkan_32bit_icd_available": 32 in manifest_architectures,
 		"vulkan_64bit_detection": (
 			"vulkaninfo" if functional_64bit else "loader_and_icd"
 		),
@@ -860,17 +867,18 @@ def _vulkan_loader_architectures(library_roots: Sequence[Path]) -> set[int]:
 	return architectures
 
 
-def _manifest_library_architecture(
+def _manifest_library_architectures(
 	manifest: Path,
 	library: str | None,
 	library_roots: Sequence[Path],
-) -> int | None:
+) -> set[int]:
+	architectures: set[int] = set()
 	if library:
 		path = Path(library)
 		if path.is_absolute():
 			architecture = _elf_architecture(path)
 			if architecture is not None:
-				return architecture
+				architectures.add(architecture)
 		else:
 			candidates = [manifest.parent / path]
 			for root in library_roots:
@@ -878,13 +886,15 @@ def _manifest_library_architecture(
 			for candidate in candidates:
 				architecture = _elf_architecture(candidate)
 				if architecture is not None:
-					return architecture
+					architectures.add(architecture)
+	if architectures:
+		return architectures
 	lowered = manifest.name.casefold()
 	if any(token in lowered for token in ("i386", "i486", "i586", "i686", "x86_32")):
-		return 32
+		return {32}
 	if any(token in lowered for token in ("x86_64", "amd64")):
-		return 64
-	return None
+		return {64}
+	return set()
 
 
 def _library_candidates(root: Path, prefix: str) -> tuple[Path, ...]:

ptlinux/ui/__init__.py

@@ -36,6 +36,7 @@ from .viewmodels.dashboard import (
 	game_client_activity_text,
 	interrupt_game_supervisor,
 	launcher_child_command,
+	launcher_child_environment,
 	launcher_restart_spec,
 	launcher_update_presentation,
 	news_card_columns,
@@ -131,6 +132,7 @@ __all__ = [
 	"interrupt_game_supervisor",
 	"launch_external_url",
 	"launcher_child_command",
+	"launcher_child_environment",
 	"launcher_restart_spec",
 	"launcher_update_presentation",
 	"news_card_columns",

ptlinux/ui/dashboard.py

@@ -123,6 +123,7 @@ from .viewmodels.dashboard import (
 	game_client_activity_text,
 	interrupt_game_supervisor,
 	launcher_child_command,
+	launcher_child_environment,
 	launcher_restart_spec,
 	launcher_update_presentation,
 	news_card_columns,
@@ -1064,7 +1065,8 @@ def run_dashboard(paths: AppPaths) -> None:
 					stdout=subprocess.PIPE,
 					stderr=subprocess.STDOUT,
 					bufsize=1,
-				)
+					env=launcher_child_environment(),
+					)
 			except OSError as process_error:
 				publish_output(f"Could not start operation: {process_error}\n")
 				if game_client_id is None:

ptlinux/ui/viewmodels/dashboard.py

@@ -14,6 +14,10 @@ from typing import TYPE_CHECKING, Callable, Mapping
 
 from ...paths import AppPaths
 from ...self_update import LauncherUpdateCheck
+from ...telemetry import (
+	TELEMETRY_OPERATION_PROCESS_ROLE,
+	TELEMETRY_PROCESS_ROLE_VARIABLE,
+)
 
 
 if TYPE_CHECKING:
@@ -465,6 +469,18 @@ def launcher_child_command(*arguments: str) -> list[str]:
 	return [sys.executable, "-B", "-m", "ptlinux", *arguments]
 
 
+def launcher_child_environment(
+	environment: Mapping[str, str] | None = None,
+) -> dict[str, str]:
+	"""Mark an internal CLI process as part of the current dashboard session."""
+
+	child_environment = dict(os.environ if environment is None else environment)
+	child_environment[TELEMETRY_PROCESS_ROLE_VARIABLE] = (
+		TELEMETRY_OPERATION_PROCESS_ROLE
+	)
+	return child_environment
+
+
 def dashboard_poll_interval(
 	*,
 	operation_active: bool,