utils.py

# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at https://mozilla.org/MPL/2.0/.

import collections
import gzip
import hashlib
import http.client
import io
import json
import multiprocessing
import os
import pathlib
import platform
import random
import stat
import string
import subprocess
import sys
import tarfile
import time
import urllib.error
import urllib.request
import zipfile

import yaml
import zstandard

from .downloads import DOWNLOADS
from .logging import log


def get_targets(yaml_path: pathlib.Path):
    """Obtain the parsed targets YAML file."""
    with yaml_path.open("rb") as fh:
        return yaml.load(fh, Loader=yaml.SafeLoader)


def get_target_settings(yaml_path: pathlib.Path, target: str):
    """Obtain the settings for a named target."""
    return get_targets(yaml_path)[target]


def supported_targets(yaml_path: pathlib.Path):
    """Obtain a set of named targets that we can build."""
    targets = set()

    for target, settings in get_targets(yaml_path).items():
        for host_platform in settings["host_platforms"]:
            if sys.platform == "linux" and host_platform == "linux64":
                targets.add(target)
            elif sys.platform == "darwin" and host_platform == "macos":
                targets.add(target)

    return targets


def target_needs(yaml_path: pathlib.Path, target: str, python_version: str):
    """Obtain the dependencies needed to build the specified target."""
    settings = get_targets(yaml_path)[target]

    needs = set(settings["needs"])

    # We only ship libedit linked readline extension on 3.10+ to avoid a GPL
    # dependency.
    if not python_version.startswith("3.9"):
        needs.discard("readline")

    return needs


def release_tag_from_git():
    return (
        subprocess.check_output(
            [
                "git",
                "log",
                "-n",
                "1",
                "--date=format:%Y%m%dT%H%M",
                "--pretty=format:%ad",
            ]
        )
        .strip()
        .decode("ascii")
    )


def hash_path(p: pathlib.Path):
    h = hashlib.sha256()

    with p.open("rb") as fh:
        while True:
            chunk = fh.read(65536)
            if not chunk:
                break

            h.update(chunk)

    return h.hexdigest()


def get_target_support_file(
    search_dir, prefix, python_version, host_platform, target_triple
):
    candidates = [
        search_dir / ("%s.%s.%s" % (prefix, python_version, target_triple)),
        search_dir / ("%s.%s.%s" % (prefix, python_version, host_platform)),
    ]

    for path in candidates:
        if path.exists():
            return path

    raise Exception(
        "Could not find support file %s for (%s, %s, %s)",
        prefix,
        python_version,
        host_platform,
        target_triple,
    )


def write_if_different(p: pathlib.Path, data: bytes):
    """Write a file if it is missing or its content is different."""
    if p.exists():
        with p.open("rb") as fh:
            existing = fh.read()
        write = existing != data
    else:
        write = True

    if write:
        with p.open("wb") as fh:
            fh.write(data)


def write_triples_makefiles(
    targets, dest_dir: pathlib.Path, support_search_dir: pathlib.Path
):
    """Write out makefiles containing make variable settings derived from config."""
    dest_dir.mkdir(parents=True, exist_ok=True)

    for triple, settings in targets.items():
        for host_platform in settings["host_platforms"]:
            # IMPORTANT: if we ever vary the content of these Makefiles by
            # Python versions, the variable names will need add the Python
            # version and the Makefile references updated to point to specific
            # versions. If we don't do that, multi-version builds will fail
            # to work correctly.

            makefile_path = dest_dir / ("Makefile.%s.%s" % (host_platform, triple))

            lines = []
            for need in settings.get("needs", []):
                lines.append(
                    "NEED_%s := 1\n" % need.upper().replace("-", "_").replace(".", "_")
                )

            image_suffix = settings.get("docker_image_suffix", "")

            lines.append("DOCKER_IMAGE_BUILD := build%s\n" % image_suffix)
            lines.append("DOCKER_IMAGE_XCB := xcb%s\n" % image_suffix)

            entry = clang_toolchain(host_platform, triple)
            lines.append(
                "CLANG_FILENAME := %s-%s-%s.tar\n"
                % (entry, DOWNLOADS[entry]["version"], host_platform)
            )

            lines.append(
                "PYTHON_SUPPORT_FILES := $(PYTHON_SUPPORT_FILES) %s\n"
                % (support_search_dir / "extension-modules.yml")
            )

            write_if_different(makefile_path, "".join(lines).encode("ascii"))


def write_package_versions(dest_path: pathlib.Path):
    """Write out versions of packages to files in a directory."""
    dest_path.mkdir(parents=True, exist_ok=True)

    for k, v in DOWNLOADS.items():
        p = dest_path / ("VERSION.%s" % k)
        content = "%s_VERSION := %s\n" % (k.upper().replace("-", "_"), v["version"])
        write_if_different(p, content.encode("ascii"))


def write_cpython_version(dest_path: pathlib.Path, version: str):
    """Write a CPython version in a directory."""
    dest_path.mkdir(parents=True, exist_ok=True)

    major_minor = ".".join(version.split(".")[:2])
    k = "cpython-%s" % major_minor
    p = dest_path / ("VERSION.%s" % k)
    content = "%s_VERSION := %s\n" % (k.upper().replace("-", "_"), version)
    write_if_different(p, content.encode("ascii"))


def write_target_settings(targets, dest_path: pathlib.Path):
    dest_path.mkdir(parents=True, exist_ok=True)

    for triple, settings in targets.items():
        payload = {}

        for key in ("host_cc", "host_cxx", "target_cc", "target_cflags"):
            payload[key] = settings.get(key)

        serialized_payload = json.dumps(payload, indent=4).encode("utf-8")

        write_if_different(dest_path / triple, serialized_payload)


class IntegrityError(Exception):
    """Represents an integrity error when downloading a URL."""


def secure_download_stream(url, size, sha256):
    """Securely download a URL to a stream of chunks.

    If the integrity of the download fails, an IntegrityError is
    raised.
    """
    h = hashlib.sha256()
    length = 0

    with urllib.request.urlopen(url) as fh:
        if not url.endswith(".gz") and fh.info().get("Content-Encoding") == "gzip":
            fh = gzip.GzipFile(fileobj=fh)

        while True:
            chunk = fh.read(65536)
            if not chunk:
                break

            h.update(chunk)
            length += len(chunk)

            yield chunk

    digest = h.hexdigest()

    if length != size or digest != sha256:
        raise IntegrityError(
            "integrity mismatch on %s: wanted size=%d, sha256=%s; got size=%d, sha256=%s"
            % (url, size, sha256, length, digest)
        )


def download_to_path(url: str, path: pathlib.Path, size: int, sha256: str):
    """Download a URL to a filesystem path, possibly with verification."""

    # We download to a temporary file and rename at the end so there's
    # no chance of the final file being partially written or containing
    # bad data.
    print("downloading %s to %s" % (url, path))

    if path.exists():
        good = True

        if path.stat().st_size != size:
            print("existing file size is wrong; removing")
            good = False

        if good:
            if hash_path(path) != sha256:
                print("existing file hash is wrong; removing")
                good = False

        if good:
            print("%s exists and passes integrity checks" % path)
            return

        path.unlink()

    # Need to write to random path to avoid race conditions. If there is a
    # race, worst case we'll download the same file N>1 times. Meh.
    tmp = path.with_name(
        "%s.tmp%s"
        % (
            path.name,
            "".join(random.choices(string.ascii_uppercase + string.digits, k=8)),
        )
    )

    for attempt in range(5):
        try:
            try:
                with tmp.open("wb") as fh:
                    for chunk in secure_download_stream(url, size, sha256):
                        fh.write(chunk)

                break
            except IntegrityError:
                tmp.unlink()
                raise
        except http.client.HTTPException as e:
            print(f"HTTP exception on {url}; retrying: {e}")
            time.sleep(2**attempt)
        except urllib.error.URLError as e:
            print(f"urllib error on {url}; retrying: {e}")
            time.sleep(2**attempt)
    else:
        raise Exception("download failed after multiple retries: %s" % url)

    tmp.rename(path)
    print("successfully downloaded %s" % url)


def download_entry(key: str, dest_path: pathlib.Path, local_name=None) -> pathlib.Path:
    entry = DOWNLOADS[key]
    url = entry["url"]
    size = entry["size"]
    sha256 = entry["sha256"]

    assert isinstance(url, str)
    assert isinstance(size, int)
    assert isinstance(sha256, str)

    local_path = dest_path / (local_name or url[url.rindex("/") + 1 :])
    download_to_path(url, local_path, size, sha256)

    return local_path


def create_tar_from_directory(fh, base_path: pathlib.Path, path_prefix=None):
    with tarfile.open(name="", mode="w", fileobj=fh) as tf:
        for root, dirs, files in os.walk(base_path):
            dirs.sort()

            for f in sorted(files):
                full = base_path / root / f
                rel = full.relative_to(base_path)
                if path_prefix:
                    rel = pathlib.Path(path_prefix) / rel
                tf.add(full, rel)


def extract_tar_to_directory(source: pathlib.Path, dest: pathlib.Path):
    with tarfile.open(source, "r") as tf:
        tf.extractall(dest)


def extract_zip_to_directory(source: pathlib.Path, dest: pathlib.Path):
    with zipfile.ZipFile(source, "r") as zf:
        zf.extractall(dest)


# 2024-01-01T00:00:00Z
DEFAULT_MTIME = 1704067200


def normalize_tar_archive(data: io.BytesIO) -> io.BytesIO:
    """Normalize the contents of a tar archive.

    We want tar archives to be as deterministic as possible. This function will
    take tar archive data in a buffer and return a new buffer containing a more
    deterministic tar archive.
    """
    members = []

    with tarfile.open(fileobj=data) as tf:
        for ti in tf:
            # We don't care about directory entries. Tools can handle this fine.
            if ti.isdir():
                continue

            filedata = tf.extractfile(ti)
            if filedata is not None:
                filedata = io.BytesIO(filedata.read())

            members.append((ti, filedata))

    # Sort the archive members. We put PYTHON.json first so metadata can
    # be read without reading the entire archive.
    def sort_key(v):
        if v[0].name == "python/PYTHON.json":
            return 0, v[0].name
        else:
            return 1, v[0].name

    members.sort(key=sort_key)

    # Normalize attributes on archive members.
    for entry in members:
        ti = entry[0]

        # The pax headers attribute takes priority over the other named
        # attributes. To minimize potential for our assigns to no-op, we
        # clear out the pax headers. We can't reset all the pax headers,
        # as this would nullify symlinks.
        for a in ("mtime", "uid", "uname", "gid", "gname"):
            try:
                ti.pax_headers.__delattr__(a)
            except AttributeError:
                pass

        ti.pax_headers = {}

        ti.mtime = DEFAULT_MTIME
        ti.uid = 0
        ti.uname = "root"
        ti.gid = 0
        ti.gname = "root"

        # Give user/group read/write on all entries.
        ti.mode |= stat.S_IRUSR | stat.S_IWUSR | stat.S_IRGRP | stat.S_IWGRP

        # If user executable, give to group as well.
        if ti.mode & stat.S_IXUSR:
            ti.mode |= stat.S_IXGRP

    dest = io.BytesIO()
    with tarfile.open(fileobj=dest, mode="w") as tf:
        for ti, filedata in members:
            tf.addfile(ti, filedata)

    dest.seek(0)

    return dest


def clang_toolchain(host_platform: str, target_triple: str) -> str:
    if host_platform == "linux64":
        # musl currently has issues with LLVM 15+.
        if "musl" in target_triple:
            return "llvm-14-x86_64-linux"
        else:
            return "llvm-18-x86_64-linux"
    elif host_platform == "macos":
        if platform.mac_ver()[2] == "arm64":
            return "llvm-aarch64-macos"
        else:
            return "llvm-x86_64-macos"
    else:
        raise Exception("unhandled host platform")


def compress_python_archive(
    source_path: pathlib.Path, dist_path: pathlib.Path, basename: str
):
    dest_path = dist_path / ("%s.tar.zst" % basename)
    temp_path = dist_path / ("%s.tar.zst.tmp" % basename)

    print("compressing Python archive to %s" % dest_path)

    try:
        with source_path.open("rb") as ifh, temp_path.open("wb") as ofh:
            params = zstandard.ZstdCompressionParameters.from_level(
                22, strategy=zstandard.STRATEGY_BTULTRA2
            )
            cctx = zstandard.ZstdCompressor(compression_params=params)
            cctx.copy_stream(ifh, ofh, source_path.stat().st_size)

        temp_path.rename(dest_path)
    finally:
        temp_path.unlink(missing_ok=True)

    print("%s has SHA256 %s" % (dest_path, hash_path(dest_path)))

    return dest_path


def add_licenses_to_extension_entry(entry):
    """Add licenses keys to a ``extensions`` entry for JSON distribution info."""

    have_licenses = False
    licenses = set()
    license_paths = set()
    license_public_domain = None

    have_local_link = False

    for link in entry["links"]:
        name = link["name"]
        if "path_static" in link or "path_dynamic" in link:
            have_local_link = True

        for value in DOWNLOADS.values():
            if name not in value.get("library_names", []):
                continue

            # Don't add licenses annotations if they aren't defined. This leaves
            # things as "unknown" to consumers.
            if "licenses" not in value:
                continue

            have_licenses = True
            licenses |= set(value["licenses"])
            license_paths.add("licenses/%s" % value["license_file"])
            license_public_domain = value.get("license_public_domain", False)

    if have_local_link and not have_licenses:
        raise Exception(
            "missing license for local library for extension entry: %s" % entry
        )

    if not have_licenses:
        return

    entry["licenses"] = sorted(licenses)
    entry["license_paths"] = sorted(license_paths)
    entry["license_public_domain"] = license_public_domain


def add_env_common(env):
    """Adds extra keys to environment variables."""

    cpu_count = multiprocessing.cpu_count()
    env["NUM_CPUS"] = "%d" % cpu_count
    env["NUM_JOBS_AGGRESSIVE"] = "%d" % max(cpu_count + 2, cpu_count * 2)

    if "CI" in os.environ:
        env["CI"] = "1"

    env_path = os.path.expanduser("~/.python-build-standalone-env")
    try:
        with open(env_path, "r") as fh:
            for line in fh:
                line = line.strip()
                if line.startswith("#"):
                    continue

                key, value = line.split("=", 1)

                print("adding %s from %s" % (key, env_path))
                env[key] = value
    except FileNotFoundError:
        pass

    # Proxy sccache settings.
    for k, v in os.environ.items():
        if k.startswith("SCCACHE_"):
            env[k] = v

    # Proxy cloud provider credentials variables to enable sccache to
    # use stores in those providers.
    for k in ("AWS_ACCESS_KEY_ID", "AWS_SECRET_ACCESS_KEY"):
        if k in os.environ:
            env[k] = os.environ[k]


def exec_and_log(args, cwd, env):
    p = subprocess.Popen(
        args,
        cwd=cwd,
        env=env,
        bufsize=1,
        stdout=subprocess.PIPE,
        stderr=subprocess.STDOUT,
    )

    for line in iter(p.stdout.readline, b""):
        log(line.rstrip())

    p.wait()

    if p.returncode:
        if "PYBUILD_BREAK_ON_FAILURE" in os.environ:
            import pdb

            pdb.set_trace()

        print("process exited %d" % p.returncode)
        sys.exit(p.returncode)


def validate_python_json(info, extension_modules):
    """Validate a PYTHON.json file for problems.

    Raises an exception if an issue is detected.
    """

    if extension_modules:
        missing = set(info["build_info"]["extensions"].keys()) - set(
            extension_modules.keys()
        )
        if missing:
            raise Exception(
                "extension modules in PYTHON.json lack metadata: %s"
                % ", ".join(sorted(missing))
            )

    for name, variants in sorted(info["build_info"]["extensions"].items()):
        for ext in variants:
            local_links = set()

            for link in ext["links"]:
                if "path_static" in link:
                    local_links.add(link["path_static"])
                if "path_dynamic" in link:
                    local_links.add(link["path_dynamic"])

                if not local_links and "framework" not in link and "system" not in link:
                    raise Exception(
                        f"Invalid link entry for extension {name}: link type not defined"
                    )

            if (
                local_links
                and not ext.get("licenses")
                and not ext.get("license_public_domain")
            ):
                raise Exception(
                    "Missing license annotations for extension %s for library files %s"
                    % (name, ", ".join(sorted(local_links)))
                )


def release_download_statistics(mode="by_asset"):
    with urllib.request.urlopen(
        "https://api.github.com/repos/indygreg/python-build-standalone/releases"
    ) as fh:
        data = json.load(fh)

    by_tag = collections.Counter()
    by_build = collections.Counter()
    by_build_install_only = collections.Counter()

    for release in data:
        tag = release["tag_name"]

        for asset in release["assets"]:
            name = asset["name"]
            count = asset["download_count"]

            by_tag[tag] += count

            if name.endswith(".tar.zst"):
                # cpython-3.10.2-aarch64-apple-darwin-debug-20220220T1113.tar.zst
                build_parts = name.split("-")
                build = "-".join(build_parts[2:-1])
                by_build[build] += count
            elif name.endswith("install_only.tar.gz"):
                # cpython-3.10.13+20240224-x86_64-apple-darwin-install_only.tar.gz
                build_parts = name.split("-")
                build = "-".join(build_parts[2:-1])
                by_build_install_only[build] += count

            if mode == "by_asset":
                print("%d\t%s\t%s" % (count, tag, name))

    if mode == "by_build":
        for build, count in sorted(by_build.items()):
            print("%d\t%s" % (count, build))
    elif mode == "by_build_install_only":
        for build, count in sorted(by_build_install_only.items()):
            print("%d\t%s" % (count, build))
    elif mode == "by_tag":
        for tag, count in sorted(by_tag.items()):
            print("%d\t%s" % (count, tag))
    elif mode == "total":
        print("%d" % by_tag.total())
    else:
        raise Exception("unhandled display mode: %s" % mode)