Improvements to try resolve the issues with login checksums.

ptlinux/command.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+from collections import deque
 from dataclasses import asdict, dataclass
 from datetime import datetime
 import json
@@ -16,6 +17,44 @@ from typing import Callable, Mapping, Sequence, TextIO
 
 
 OutputLineFilter = Callable[[str], str | None]
+MAX_CAPTURE_CHARACTERS = 1024 * 1024
+
+
+class BoundedTextCapture:
+	"""Retain the newest subprocess output without unbounded memory growth."""
+
+	def __init__(self, max_characters: int = MAX_CAPTURE_CHARACTERS) -> None:
+		if max_characters <= 0:
+			raise ValueError("capture size must be positive")
+		self._max_characters = max_characters
+		self._parts: deque[str] = deque()
+		self._characters = 0
+		self._lock = threading.Lock()
+
+	def append(self, value: str) -> None:
+		if not value:
+			return
+		with self._lock:
+			if len(value) >= self._max_characters:
+				self._parts.clear()
+				self._parts.append(value[-self._max_characters:])
+				self._characters = self._max_characters
+				return
+			self._parts.append(value)
+			self._characters += len(value)
+			while self._characters > self._max_characters:
+				overflow = self._characters - self._max_characters
+				oldest = self._parts[0]
+				if len(oldest) <= overflow:
+					self._parts.popleft()
+					self._characters -= len(oldest)
+				else:
+					self._parts[0] = oldest[overflow:]
+					self._characters -= overflow
+
+	def getvalue(self) -> str:
+		with self._lock:
+			return "".join(self._parts)
 
 
 @dataclass(frozen=True)
@@ -103,8 +142,8 @@ class CommandRunner:
 				outcome = CommandOutcome(
 					argv=arguments,
 					returncode=completed.returncode,
-					stdout=completed.stdout,
-					stderr=completed.stderr,
+					stdout=_bounded_text(completed.stdout),
+					stderr=_bounded_text(completed.stderr),
 					duration_seconds=time.monotonic() - started,
 					timed_out=False,
 					dry_run=False,
@@ -113,8 +152,8 @@ class CommandRunner:
 				outcome = CommandOutcome(
 					argv=arguments,
 					returncode=None,
-					stdout=_as_text(exc.stdout),
-					stderr=_as_text(exc.stderr),
+					stdout=_bounded_text(_as_text(exc.stdout)),
+					stderr=_bounded_text(_as_text(exc.stderr)),
 					duration_seconds=time.monotonic() - started,
 					timed_out=True,
 					dry_run=False,
@@ -144,12 +183,12 @@ class CommandRunner:
 			bufsize=1,
 			start_new_session=True,
 		)
-		stdout: list[str] = []
-		stderr: list[str] = []
+		stdout = BoundedTextCapture()
+		stderr = BoundedTextCapture()
 
 		def relay(
 			stream: TextIO,
-			capture: list[str],
+			capture: BoundedTextCapture,
 			destination: TextIO,
 		) -> None:
 			try:
@@ -196,8 +235,8 @@ class CommandRunner:
 		return CommandOutcome(
 			argv=arguments,
 			returncode=None if timed_out else returncode,
-			stdout="".join(stdout),
-			stderr="".join(stderr),
+			stdout=stdout.getvalue(),
+			stderr=stderr.getvalue(),
 			duration_seconds=time.monotonic() - started,
 			timed_out=timed_out,
 			dry_run=False,
@@ -228,6 +267,10 @@ def _as_text(value: bytes | str | None) -> str:
 	return value
 
 
+def _bounded_text(value: str) -> str:
+	return value[-MAX_CAPTURE_CHARACTERS:]
+
+
 def _terminate_process_group(
 	process: subprocess.Popen[str],
 	*,