[ Web Proxy ]
URL:
Viewing: https://raw.githubusercontent.com/Navegos/RustPython/main/scripts/compare_bytecode.py [Back]  [Original]

#!/usr/bin/env python3
"""Compare bytecode between CPython and RustPython.

Compiles all Python files under Lib/ with both interpreters and reports
differences in the generated bytecode instructions.

Usage:
    python scripts/compare_bytecode.py
    python scripts/compare_bytecode.py --detail
    python scripts/compare_bytecode.py --filter "asyncio/*.py"
    python scripts/compare_bytecode.py --summary-json report.json
"""

import argparse
import fnmatch
import json
import os
import random
import subprocess
import sys
import tempfile
from collections import defaultdict

SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
PROJECT_ROOT = os.path.dirname(SCRIPT_DIR)
DIS_DUMP = os.path.join(SCRIPT_DIR, "dis_dump.py")
DEFAULT_REPORT = os.path.join(PROJECT_ROOT, "compare_bytecode.report")
DUMP_TIMEOUT = 600


def find_rustpython():
    """Locate the RustPython binary, allowing release builds only."""
    if "RUSTPYTHON" in os.environ:
        path = os.environ["RUSTPYTHON"]
        normalized = os.path.normpath(path)
        debug_fragment = os.path.join("target", "debug", "rustpython")
        if normalized.endswith(debug_fragment):
            raise ValueError(
                "RUSTPYTHON must point to a release binary, not target/debug/rustpython"
            )
        return path

    path = os.path.join(PROJECT_ROOT, "target", "release", "rustpython")
    if os.path.isfile(path) and os.access(path, os.X_OK):
        return path
    return None


def collect_targets(lib_dir, pattern=None):
    """Collect Python files to compare, relative to lib_dir."""
    targets = []
    for root, dirs, files in os.walk(lib_dir):
        dirs[:] = sorted(
            d for d in dirs if d != "__pycache__" and not d.startswith(".")
        )
        for fname in sorted(files):
            if not fname.endswith(".py"):
                continue
            fpath = os.path.join(root, fname)
            relpath = os.path.relpath(fpath, lib_dir)
            if pattern and not fnmatch.fnmatch(relpath, pattern):
                continue
            targets.append((relpath, fpath))
    return targets


def _start_one(interpreter, targets, base_dir):
    """Start a single dis_dump.py subprocess."""
    env = os.environ.copy()
    if interpreter != sys.executable:
        env["RUSTPYTHONPATH"] = base_dir

    files_file = None
    output_file = None
    try:
        files_file = tempfile.NamedTemporaryFile(
            mode="w",
            encoding="utf-8",
            delete=False,
            prefix="compare-bytecode-files-",
        )
        output_file = tempfile.NamedTemporaryFile(
            mode="w",
            encoding="utf-8",
            delete=False,
            prefix="compare-bytecode-output-",
        )
        for _, path in targets:
            files_file.write(path)
            files_file.write("\n")
        files_file.close()
        output_file.close()
        cmd = [
            interpreter,
            DIS_DUMP,
            "--base-dir",
            base_dir,
            "--files-from",
            files_file.name,
            "--output",
            output_file.name,
            "--progress",
            "10",
        ]
        proc = subprocess.Popen(
            cmd,
            stdout=subprocess.PIPE,
            stderr=None,  # inherit stderr so progress dots appear on terminal
            env=env,
            cwd=PROJECT_ROOT,
        )
        return {
            "proc": proc,
            "files_file": files_file.name,
            "output_file": output_file.name,
            "targets": targets,
            "interpreter": interpreter,
            "base_dir": base_dir,
        }
    except Exception:
        for handle in (files_file, output_file):
            if handle is None:
                continue
            try:
                handle.close()
            finally:
                if os.path.exists(handle.name):
                    os.unlink(handle.name)
        raise


def _load_dump_output(output_file):
    try:
        with open(output_file, encoding="utf-8") as f:
            content = f.read().strip()
    except OSError as e:
        print("  Failed to read dump output: %s" % e, file=sys.stderr)
        return None
    if not content:
        return {}
    try:
        return json.loads(content)
    except json.JSONDecodeError as e:
        print("  JSON parse error: %s" % e, file=sys.stderr)
        return None


def _run_sync_dump(interpreter, targets, base_dir, timeout=DUMP_TIMEOUT):
    job = _start_one(interpreter, targets, base_dir)
    proc = job["proc"]
    stdout = b""
    timed_out = False
    try:
        stdout = proc.communicate(timeout=timeout)[0]
    except subprocess.TimeoutExpired:
        proc.kill()
        proc.communicate()
        print(f"  Timeout ({timeout}s)", file=sys.stderr)
        timed_out = True

    try:
        data = _load_dump_output(job["output_file"])
    finally:
        for path in (job["files_file"], job["output_file"]):
            if os.path.exists(path):
                os.unlink(path)

    if timed_out:
        return {}
    if proc.returncode != 0:
        print("  Warning: exited with code %d" % proc.returncode, file=sys.stderr)
    stray = stdout.decode(errors="replace").strip()
    if stray:
        print("  Warning: unexpected stdout from dump helper", file=sys.stderr)
    return data


def _rerun_missing_targets(interpreter, targets, base_dir):
    recovered = {}
    failed = []
    empty = []
    for target in targets:
        relpath = target[0]
        data = _run_sync_dump(interpreter, [target], base_dir)
        if data is None:
            failed.append(relpath)
            recovered[relpath] = {
                "status": "error",
                "error": "dump helper failed while rerunning target",
            }
        elif data:
            recovered.update(data)
        else:
            empty.append(relpath)
            recovered[relpath] = {
                "status": "error",
                "error": "dump helper produced no data while rerunning target",
            }
    if failed:
        print(
            "  Warning: rerun failed for %d file(s): %s"
            % (len(failed), ", ".join(failed[:5])),
            file=sys.stderr,
        )
    if empty:
        print(
            "  Warning: rerun produced no data for %d file(s): %s"
            % (len(empty), ", ".join(empty[:5])),
            file=sys.stderr,
        )
    return recovered


def _finish_one(job, timeout=DUMP_TIMEOUT):
    """Wait for a single dis_dump.py process and return parsed JSON."""
    proc = job["proc"]
    expected = {relpath for relpath, _ in job["targets"]}
    stdout = b""
    try:
        stdout = proc.communicate(timeout=timeout)[0]
    except subprocess.TimeoutExpired:
        proc.kill()
        proc.communicate()
        print(
            f"  Timeout ({timeout}s), retrying {len(job['targets'])} file(s) serially",
            file=sys.stderr,
        )
        data = None
    else:
        data = _load_dump_output(job["output_file"])
    finally:
        for path in (job["files_file"], job["output_file"]):
            if os.path.exists(path):
                os.unlink(path)

    if proc.returncode != 0:
        print("  Warning: exited with code %d" % proc.returncode, file=sys.stderr)

    stray = stdout.decode(errors="replace").strip()
    if stray:
        print("  Warning: unexpected stdout from dump helper", file=sys.stderr)

    if data is None:
        return _rerun_missing_targets(
            job["interpreter"], job["targets"], job["base_dir"]
        )

    missing = [
        target
        for target in job["targets"]
        if target[0] not in data and target[0] in expected
    ]
    if missing:
        print(
            "  Re-running %d missing file(s) serially" % len(missing),
            file=sys.stderr,
        )
        data.update(
            _rerun_missing_targets(job["interpreter"], missing, job["base_dir"])
        )
    return data


def start_dump(interpreter, targets, base_dir, num_workers=1):
    """Start dis_dump.py under the given interpreter, split across workers."""
    if num_workers  50:
                    p("  ... and %d more" % (len(rp_error_files) - 50))
                p()

            if diff_files:
                p("-" * 60)
                p("  Bytecode Differences")
                p("-" * 60)
                for fp, code_diffs in diff_files:
                    p()
                    p("  %s:" % fp)
                    for code_path, diffs in code_diffs:
                        shown = min(len(diffs), args.max_diffs)
                        p("    %s: %d difference(s)" % (code_path, len(diffs)))
                        for idx, cp_inst, rp_inst in diffs[:shown]:
                            if idx == -1:
                                p("      %s" % (cp_inst or rp_inst))
                            else:
                                p("      [%3d] CPython:     %s" % (idx, cp_inst))
                                p("            RustPython:  %s" % rp_inst)
                        if len(diffs) > shown:
                            p("      ... and %d more" % (len(diffs) - shown))
                p()
        else:
            list_limit = 0 if args.summary_json else max(args.list_limit, 0)
            if diff_summaries and list_limit:
                shown = min(list_limit, len(diff_summaries))
                total = len(diff_summaries)
                p(f"Top differing files ({shown} shown of {total}):")
                top = sorted(
                    diff_summaries,
                    key=lambda item: (
                        item["diff_instructions"],
                        item["diff_code_objects"],
                        item["path"],
                    ),
                    reverse=True,
                )[:list_limit]
                for item in top:
                    p(
                        "  %s  (%d code objects, %d instruction diffs)"
                        % (
                            item["path"],
                            item["diff_code_objects"],
                            item["diff_instructions"],
                        )
                    )
                p()
                p("Use --detail to see specific instruction differences.")
                p()

    # Summary JSON output
    if args.summary_json:
        summary = {
            "total": total,
            "sample": args.sample,
            "sample_seed": sample_seed,
            "match": match,
            "differ": differ,
            "rp_error": rp_err,
            "cp_error": cp_err,
            "both_error": both_err,
            "rp_missing": rp_miss,
            "match_pct": round(100.0 * match / total, 2) if total else 0,
            "diff_files": [fp for fp, _ in diff_files]
            if need_detailed_diffs
            else [item["path"] for item in diff_summaries],
            "top_diff_files": sorted(
                diff_summaries,
                key=lambda item: (
                    item["diff_instructions"],
                    item["diff_code_objects"],
                    item["path"],
                ),
                reverse=True,
            )[: min(20, len(diff_summaries))],
            "rp_error_files": [fp for fp, _ in rp_error_files],
        }
        with open(args.summary_json, "w") as f:
            json.dump(summary, f, indent=2)
        log("Summary JSON: %s" % args.summary_json)

    log("Done: %d match, %d differ, %d errors" % (match, differ, rp_err))
    sys.exit(0 if differ == 0 and rp_err == 0 else 1)


if __name__ == "__main__":
    main()

Web Proxy Viewer  |  New URL  |  Original Page