python · maurycy · Jul 2, 2026 · Jul 2, 2026 · Jul 2, 2026 · Jul 2, 2026
@@ -1,5 +1,6 @@
 """Thin Python wrapper around C binary writer for profiling data."""
 
+import sys
 import time
 
 import _remote_debugging
@@ -61,6 +62,7 @@ def __init__(self, filename, sample_interval_usec, *, skip_idle=False,
         self.filename = filename
         self.sample_interval_usec = sample_interval_usec
         self.skip_idle = skip_idle
+        self.running = True
 
         compression_type = _resolve_compression(compression)
         start_time_us = int(time.monotonic() * 1_000_000)
@@ -79,9 +81,17 @@ def collect(self, stack_frames, timestamp_us=None):
             timestamp_us: Optional timestamp in microseconds. If not provided,
                           uses time.monotonic() to generate one.
         """
+        if not self.running:
+            return
         if timestamp_us is None:
             timestamp_us = int(time.monotonic() * 1_000_000)
-        self._writer.write_sample(stack_frames, timestamp_us)
+        try:
+            self._writer.write_sample(stack_frames, timestamp_us)
+        except OverflowError as e:
 /*[clinic input] 
 _remote_debugging.BinaryWriter.__exit__ 
     exc_type: object = None 
     exc_val: object = None 
     exc_tb: object = None 
 Exit context manager, finalizing the file. 
 [clinic start generated code]*/ 
 def __exit__(self, exc_type, exc_val, exc_tb): 
     """Context manager exit - finalize unless there was an error.""" 
     if exc_type is None: 
         self._writer.finalize() 
     else: 
         self._writer.close() 
 /* Only finalize on normal exit (no exception) */ 
 except (RuntimeError, UnicodeDecodeError, MemoryError, OSError): 
 /*[clinic input] 
 _remote_debugging.BinaryWriter.__exit__ 
     exc_type: object = None 
     exc_val: object = None 
     exc_tb: object = None 
  
 Exit context manager, finalizing the file. 
 [clinic start generated code]*/ 
 def __exit__(self, exc_type, exc_val, exc_tb): 
     """Context manager exit - finalize unless there was an error.""" 
     if exc_type is None: 
         self._writer.finalize() 
     else: 
         self._writer.close() 
 /* Only finalize on normal exit (no exception) */ 
 except (RuntimeError, UnicodeDecodeError, MemoryError, OSError): 
+            self.running = False
+            print(f"Warning: {e}; stopping early and keeping the data "
+                  "collected so far.",
+                  file=sys.stderr)
 
     def collect_failed_sample(self):
         """Record a failed sample attempt (no-op for binary format)."""

@@ -9,6 +9,8 @@
 import unittest
 from collections import defaultdict
 
+from test.support import captured_stderr
+
 try:
     import _remote_debugging
     from _remote_debugging import (
@@ -143,6 +145,12 @@ def samples_to_by_thread(samples):
 class BinaryFormatTestBase(unittest.TestCase):
     """Base class with common setup/teardown for binary format tests."""
 
+    HDR_OFF_SAMPLES = 28
+    HDR_OFF_THREADS = 36
+    HDR_OFF_STR_TABLE = 40
+    HDR_OFF_FRAME_TABLE = 48
+    FILE_HEADER_PLACEHOLDER_SIZE = 64
+
     def setUp(self):
         self.temp_files = []
 
@@ -994,15 +1002,89 @@ def test_writer_total_samples_after_close_returns_zero(self):
         w.close()
         self.assertEqual(w.total_samples, 0)
 
+    def test_binary_collector_stops_gracefully_on_overflow(self):
+        """OverflowError from the writer stops collection via the running
+        protocol instead of propagating and corrupting the file.
+        See gh-151292."""
+        with tempfile.NamedTemporaryFile(suffix=".bin", delete=False) as f:
+            filename = f.name
+        self.temp_files.append(filename)
+
+        collector = BinaryCollector(filename, 1000, compression="none")
+        self.assertTrue(collector.running)
+
+        sample = [
+            make_interpreter(0, [make_thread(1, [make_frame("a.py", 1, "f")])])
+        ]
+
+        # Collect real samples first, then hit the limit.
+        for i in range(3):
+            collector.collect(sample, timestamp_us=(i + 1) * 1000)
+        self.assertTrue(collector.running)
+
+        real_writer = collector._writer
+
+        class _OverflowingWriter:
+            def write_sample(self, stack_frames, timestamp_us):
+                raise OverflowError("too many samples for binary format")
+
+        collector._writer = _OverflowingWriter()
+        with captured_stderr() as stderr:
+            collector.collect(sample, timestamp_us=4000)
+
+        self.assertFalse(collector.running)
+        self.assertIn("too many samples", stderr.getvalue())
+
+        # The real writer can still be finalized into a valid file that
+        # keeps the samples collected before the limit was hit.
+        collector._writer = real_writer
+        collector.export(None)
+
+        with open(filename, "rb") as f:
+            header = f.read(self.FILE_HEADER_PLACEHOLDER_SIZE)
+        magic, version = struct.unpack_from("=II", header, 0)
+        self.assertEqual(magic, 0x54414348)  # "TACH"
+        self.assertEqual(version, 1)
+        (sample_count,) = struct.unpack_from(
+            "=Q", header, self.HDR_OFF_SAMPLES
+        )
+        self.assertEqual(sample_count, 3)
+
+        reader_collector = RawCollector()
+        with BinaryReader(filename) as reader:
+            self.assertEqual(reader.replay_samples(reader_collector), 3)
+
+    def test_interpreter_id_overflow_rejected(self):
+        """An interpreter_id wider than u32 raises OverflowError before any
+        writer state is mutated: subsequent valid samples are still accepted
+        and finalize produces a readable file."""
+        with tempfile.NamedTemporaryFile(suffix=".bin", delete=False) as f:
+            filename = f.name
+        self.temp_files.append(filename)
+
+        good = [
+            make_interpreter(0, [make_thread(1, [make_frame("a.py", 1, "f")])])
+        ]
+        bad = [
+            make_interpreter(2**32, [make_thread(1, [make_frame("a.py", 1, "f")])])
+        ]
+
+        writer = _remote_debugging.BinaryWriter(filename, 1000, 0, compression=0)
+        writer.write_sample(good, 1000)
+        with self.assertRaises(OverflowError):
+            writer.write_sample(bad, 2000)
+        writer.write_sample(good, 3000)
+        writer.finalize()
+        self.assertEqual(writer.total_samples, 2)
+
+        reader_collector = RawCollector()
+        with BinaryReader(filename) as reader:
+            self.assertEqual(reader.replay_samples(reader_collector), 2)
+
 
 class TestBinaryFormatValidation(BinaryFormatTestBase):
     """Tests for malformed binary files."""
 
-    HDR_OFF_SAMPLES = 28
-    HDR_OFF_THREADS = 36
-    HDR_OFF_STR_TABLE = 40
-    HDR_OFF_FRAME_TABLE = 48
-    FILE_HEADER_PLACEHOLDER_SIZE = 64
     FILE_FOOTER_SIZE = 32
     FTR_OFF_STRINGS = 0
     FTR_OFF_FRAMES = 4

diff --git a/Misc/NEWS.d/next/Library/2026-07-02-15-26-51.gh-issue-151292.nmnQlp.rst b/Misc/NEWS.d/next/Library/2026-07-02-15-26-51.gh-issue-151292.nmnQlp.rst
@@ -0,0 +1,3 @@
+Fix ``profiling.sampling --binary`` leaving unreadable profile files when
+the ``_remote_debugging`` binary writer raises :exc:`OverflowError`. Patch
+by Maurycy Pawłowski-Wieroński.