gh-151378: Fix unbounded memory growth when recording binary and the thread is sleeping

[maurycy]

Fixes #151378

I've described the root cause in #151378.

Instead of adding some LIMIT and checking it, I'm doing two tricks here that spare us one struct and some complexity: using a fixed buffer (lazy alloc, so threads do not pay any price if there's no repeats), instead of a dynamic array, and eager encoding, instead of having PendingRLESample.

Truth be told, it's not easy to write a test. We currently have no test for any OverflowError

Also, I'm wondering if we couldn't simplify alloc-dealloc in the reader, too:

cpython/Modules/_remote_debugging/binary_io_reader.c

Lines 1098 to 1114 in ee78d43

	if (i == 0 || status != batch_status) {
	if (timestamps_list) {
	int rc = emit_batch(state, collector, thread_id, interpreter_id,
	batch_status, ts->current_stack, ts->current_stack_depth,
	reader, timestamps_list, batch_idx);
	Py_DECREF(timestamps_list);
	if (rc < 0) {
	return -1;
	}
	}
	timestamps_list = PyList_New(count - i);
	if (!timestamps_list) {
	return -1;
	}
	batch_status = status;
	batch_idx = 0;
	}

cpython/Modules/_remote_debugging/binary_io_reader.c

Line 958 in ee78d43

if (PyList_SetSlice(timestamps_list, actual_size, PyList_GET_SIZE(timestamps_list), NULL) < 0) {

but it's out of the current scope (and maybe PyList_New(run_len) would be sufficient there; just call it a day.)

(No) reproduction

The same as in #151378

2026-06-24T15:42:07.243116000+0200 maurycy@gimel /Users/maurycy/work/cpython (rle-overlow-fixed ee78d43+) % ./python.exe -c "
import time
time.sleep(3600)
" & TARGET=$!

sudo ./python.exe -m profiling.sampling attach --binary -r 1000khz -d 300 -o /tmp/rle.bin $TARGET &
sleep 2; PROF=$(pgrep -fn "profiling.sampling attach")
for i in $(seq 5); do printf "t=%2dmin  RSS=%d MB\n" $i $(($(ps -o rss= -p $PROF|tr -d ' ')/1024)); sleep 60; done
[1] 65625
[2] 65626
t= 1min  RSS=32 MB
t= 2min  RSS=33 MB
t= 3min  RSS=33 MB
t= 4min  RSS=33 MB
t= 5min  RSS=33 MB
Captured 99,316,092 samples in 300.00 seconds
Sample rate: 331,053.64 samples/sec
Error rate: 0.00
  Binary Encoding:
    Records:          12,131
      RLE repeat:     12,130 (100.0%) [99,308,159 samples]
      Full stack:     1 (0.0%)
      Suffix match:   0 (0.0%)
      Pop-push:       0 (0.0%)
  Frame Efficiency:
    Frames written:   1
    Frames saved:     99,308,159 (100.0%)
    Bytes (pre-zstd): 189.6 MB
Warning: missed 200683909 samples from the expected total of 300000001 (66.89%)
Binary profile written to /tmp/rle.bin (99316092 samples)

[2]  + done       sudo ./python.exe -m profiling.sampling attach --binary -r 1000khz -d 300 -o 
2026-06-24T15:47:52.516465000+0200 maurycy@gimel /Users/maurycy/work/cpython (rle-overlow-fixed ee78d43+) % ./python.exe -VV
Python 3.16.0a0 (heads/main-dirty:ee78d4323c1, Jun 24 2026, 12:32:26) [Clang 21.0.0 (clang-2100.1.1.101)]

• Issue: _remote_debugging: unbounded memory growth when recording binary and the thread is sleeping #151378

[pablogsal]

LGTM. Nice fix. I checked the RLE boundary case locally and it held up.

[pablogsal]

Some minor comments

[pablogsal]

Amazing! great job!

[miss-islington-app]

Thanks @maurycy for the PR, and @pablogsal for merging it 🌮🎉.. I'm working now to backport this PR to: 3.15.
🐍🍒⛏🤖

[bedevere-app]

GH-153047 is a backport of this pull request to the 3.15 branch.