WIP: Bintar deps validation script

Author: cvicentiu

runtime_config/bintar_deps/deps_10.11.yaml

@@ -0,0 +1,111 @@
+import logging
+import tarfile
+import subprocess
+import shutil
+import sys
+import re
+import os
+from typing import Tuple
+
+from pathlib import Path
+
+
+def setup_logging(level: int):
+    # ANSI escape codes for colors
+    RESET = "\033[0m"
+    GREEN = "\033[32m"
+    RED = "\033[31m"
+    YELLOW = "\033[33m"
+
+
+    # Custom log formatter to include colors
+    class ColoredFormatter(logging.Formatter):
+        def format(self, record):
+            if record.levelno == logging.INFO:
+                color = GREEN
+            elif record.levelno == logging.ERROR:
+                color = RED
+            elif record.levelno == logging.WARNING:
+                color = YELLOW
+            else:
+                color = RESET
+
+            # Apply color to the message
+            record.msg = f"{color}{record.levelname}{RESET}: {record.msg}"
+            return super().format(record)
+
+    # Basic logging configuration
+    logging.basicConfig(
+        level=level,
+        format="%(message)s",  # No logger name or timestamp
+        handlers=[
+            logging.StreamHandler()
+        ]
+    )
+
+    # Apply the custom formatter
+    logging.getLogger().handlers[0].setFormatter(ColoredFormatter("%(message)s"))
+
+
+# Helper functions
+def run_command(command):
+    """Run a shell command and return the output."""
+    try:
+        result = subprocess.run(command, shell=True, check=True, text=True,
+                                stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        return result.stdout.strip()
+    except subprocess.CalledProcessError as e:
+        logging.error(f"Error running command '{command}': {e} {e.stderr.strip()}")
+        return None
+
+
+def _unpack_archive(tarball_path: Path, dst_path: Path):
+    logging.info(f"Extracting archive {tarball_path}")
+    with tarfile.open(str(tarball_path), 'r:*') as tar:
+        tar.extractall(path=str(dst_path), filter='fully_trusted')
+
+
+def _parse_archive_path(archive_path: Path) -> Tuple[str, str]:
+    archive_name = archive_path.name
+
+    # Removes the last extension (e.g., .gz)
+    base_name = Path(archive_name).stem
+    # Check and remove the .tar extension
+    if base_name.endswith(".tar"):
+        base_name = Path(base_name).stem
+
+    # Let's extract the product version from the archive:
+    match = re.search('([1-9][0-9]+\\.[0-9]+\\.[0-9]+)', base_name)
+    if not match:
+        logging.error(f'Archive name {archive_name} must contain product version')
+        sys.exit(1)
+
+    # Only interested in major and minor version numbers, not point.
+    version = match.group(0).split('.')
+    major_minor = f'{version[0]}.{version[1]}'
+
+    logging.info(f'Product version (major.minor) {major_minor}')
+
+    return base_name, major_minor
+
+
+def prepare_test_directory(archive_path: Path, tests_path: Path) -> Tuple[Path, str]:
+
+    base_name, major_minor = _parse_archive_path(archive_path)
+    # The archive contains a folder with the same name as the archive.
+    # We are interested in the contents within that folder, as thats where
+    # the files are.
+    files_path = tests_path / base_name
+
+    # Cleanup any previous run.
+    shutil.rmtree(files_path, ignore_errors=True)
+
+    # Create the test directory.
+    tests_path.mkdir(parents=True, exist_ok=True)
+
+    _unpack_archive(archive_path, tests_path)
+
+    # Sanity check that the archive has maintained its format.
+    assert files_path.is_dir()
+
+    return files_path, major_minor

runtime_config/bintar_deps/deps_10.5.yaml

@@ -0,0 +1,256 @@
+import argparse
+import os
+import re
+import sys
+from concurrent.futures import ProcessPoolExecutor
+from typing import Iterable, Tuple, Generator
+import logging
+
+import magic
+import yaml
+
+from common import run_command, setup_logging, prepare_test_directory
+from pathlib import Path
+
+
+def check_file_is_elf_binary_callback(file_path: str) -> str | None:
+    global mime
+    try:
+        file_type = mime.from_file(file_path)
+        if "elf" in file_type.lower():  # Identify ELF files
+            return file_path
+    except Exception as e:
+        logging.error(f"Error checking file {file_path}: {e}")
+    return None
+
+
+def start_worker():
+    global mime
+    mime = magic.Magic()
+
+
+def get_file_paths(path: str) -> Generator[str, None, None]:
+    # Generator to feed file paths to processes.
+    for root, _, files in os.walk(path):
+        for file in files:
+            yield os.path.join(root, file)
+
+
+def get_executables(path: str):
+    """
+    Recursively searches for ELF executable files and libraries in the given
+    path using a multiprocess approach (to speed up).
+
+    Args:
+        path (str): Root directory to search.
+
+    Returns:
+        list: List of paths to ELF executables and libraries.
+    """
+    executables = []
+
+    # Use ProcessPoolExecutor to process files in parallel
+    # This offers a 10x speed up compared to single threaded.
+    with ProcessPoolExecutor(initializer=start_worker,
+                             max_workers=os.cpu_count()) as executor:
+        results = executor.map(check_file_is_elf_binary_callback,
+                               get_file_paths(path))
+
+    # Collect non-None results
+    executables = [result for result in results if result]
+
+    return executables
+
+
+def get_file_dependencies_callback(file: str) -> Tuple[str, set[str]]:
+    result = set()
+    output = run_command(f'readelf -d {file}')
+    if output is None:
+        logging.error(f"Failed to check libraries for {file}.")
+        return file, False
+
+    pattern = "Shared library: \\[(\\S*)\\]"
+    regex_shared_library = re.compile(pattern)
+
+    for line in output.splitlines():
+        # Here is an example line we match:
+        # 0x0000000000000001 (NEEDED)      Shared library: [libsystemd.so.0]
+
+        match = regex_shared_library.search(line)
+        if not match:
+            continue
+        library = match.group(1)
+        result.add(library)
+
+    return file, result
+
+
+def get_dependencies_for_files(files: Iterable[str]) -> dict[str, list[str]]:
+    with ProcessPoolExecutor(initializer=start_worker,
+                             max_workers=os.cpu_count()) as executor:
+        results = executor.map(get_file_dependencies_callback, files)
+
+    deps = {}
+    for full_file_path, file_deps in results:
+        # TODO(cvicentiu) Perhaps this should be marked as a failure.
+        # Unable to read file dependencies, skip the file.
+        if file_deps is False:
+            continue
+
+        deps[full_file_path] = file_deps
+
+    return deps
+
+
+def remove_base_path_from_files(dependencies: dict[str, list[str]],
+                                base_path: str) -> dict[str, list[str]]:
+    """
+        For all keys in dependencies, remove the base_path prefix.
+        "./tests/mariadb-11.6.2-linux-systemd-x86_64/lib/libgalera_smm.so"
+        becomes
+        "lib/libgalera_smm.so"
+    """
+    result = {}
+    for full_file_name, deps in dependencies.items():
+        # If this assert fails, there is a bug in the testing script.
+        assert full_file_name.startswith(base_path)
+        file_name = full_file_name[len(base_path)+1:]
+        result[file_name] = deps
+    return result
+
+
+def dependencies_to_canonical_repr(
+        dependencies: dict[str, set[str]],
+        version: str,
+        base_path: Path
+) -> dict[str, dict[str, list[str]]]:
+
+    dependencies = remove_base_path_from_files(dependencies, base_path.as_posix())
+    result = {
+        'version': version,
+        'files': {},
+    }
+
+    for file, deps in dependencies.items():
+        result['files'][file] = list(sorted(deps))
+
+    return result
+
+
+def get_standard_dependencies(path: str):
+    with open(path, 'r') as spec_file:
+        return yaml.safe_load(spec_file)
+
+
+def get_executable_files_dependencies(path: str):
+    files = get_executables(path)
+    return get_dependencies_for_files(files)
+
+
+def compare_versions(archive_deps, standard_deps,
+                     allow_cross_version: bool):
+    a_version = archive_deps['version']
+    s_version = standard_deps['version']
+
+    if a_version != s_version:
+        if allow_cross_version:
+            logging.warn(f'WARNING: version mismatch {a_version} {s_version}')
+        else:
+            logging.error(f'version mismatch {a_version} {s_version}')
+            return True
+    return False
+
+
+def compare_dependencies(archive_deps, standard_deps):
+    error = False
+    files = archive_deps['files']
+    control = standard_deps['files']
+
+    files_set = set(files.keys())
+    control_set = set(control.keys())
+
+    files_extra = files_set.difference(control_set)
+    files_missing = control_set.difference(files_set)
+    common = files_set.intersection(control_set)
+
+    if files_extra:
+        logging.error(f'We have extra files! {files_extra}')
+        error = True
+
+    if files_missing:
+        logging.error(f'We have missing files from the archive! {files_missing}')
+        error = True
+
+    for file in common:
+        deps_extra = set(files[file]).difference(control[file])
+        deps_missing = set(control[file]).difference(files[file])
+
+        if deps_extra:
+            logging.error(f'We have extra deps for {file}! {deps_extra}')
+            error = True
+        if deps_missing:
+            logging.error(f'We have missing deps for {file}! {deps_missing}')
+            error = True
+
+    return error
+
+
+def main(archive_path: Path,
+         tests_path: Path,
+         deps_file: Path,
+         record: bool,
+         allow_cross_version: bool):
+    error = False  # track any errors so we can return properly.
+
+    files_path, major_minor = prepare_test_directory(archive_path, tests_path)
+
+    logging.info("Fetching archive dependencies")
+    dependencies = get_executable_files_dependencies(files_path)
+
+    canonical_deps = dependencies_to_canonical_repr(dependencies,
+                                                    version=major_minor,
+                                                    base_path=files_path)
+
+    if record:
+        logging.info(f"Recording new result to {deps_file}")
+        with open(deps_file, 'w') as f:
+            yaml.dump(canonical_deps, f)
+        return
+
+    # Validate dependencies.
+    standard = get_standard_dependencies(deps_file)
+
+    error |= compare_versions(canonical_deps, standard, allow_cross_version)
+    error |= compare_dependencies(canonical_deps, standard)
+
+    if error:
+        logging.error("Some tests failed")
+        sys.exit(1)
+
+    logging.info("All OK")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+                    prog='bintar_deps.py',
+                    description='Checks/Records bintar files and dependencies')
+    parser.add_argument('archive',
+                        help='Path to the binary tarball archive')
+    parser.add_argument('deps_file',
+                        help='Path to YAML file with a list of dependencies')
+    parser.add_argument('--record', action='store_true',
+                        help='Use the bintar archive to generate a deps file')
+    parser.add_argument('--test_directory', type=str, default='./tests/',
+                        help='Where to extract the archive and run tests.')
+    parser.add_argument('--allow_cross_version', action='store_true',
+                        help='Tests pass even if there is a '
+                             'version mismatch between the archive and '
+                             'the deps_file version')
+    args = parser.parse_args()
+
+    setup_logging(logging.INFO)
+    main(archive_path=Path(args.archive),
+         tests_path=Path(args.test_directory),
+         deps_file=Path(args.deps_file),
+         record=args.record,
+         allow_cross_version=args.allow_cross_version)