Fix tests, update README, unique (#1)

Signed-off-by: Christian Heimes <[email protected]>

README.md

@@ -1,3 +1,68 @@
 # ELF deps
 
-Python implementation of RPM [elfdeps](https://github.com/rpm-software-management/rpm/blob/master/tools/elfdeps.c).
+Python implementation of RPM [`elfdeps`](https://github.com/rpm-software-management/rpm/blob/master/tools/elfdeps.c). The `elfdeps` tool can extract dependencies and provides from an ELF binary.
+
+## Example
+
+```shell-session
+$ elfdeps --requires /usr/bin/python3.12
+libc.so.6(GLIBC_2.34)(64bit)
+libc.so.6(GLIBC_2.2.5)(64bit)
+libpython3.12.so.1.0()(64bit)
+libc.so.6()(64bit)
+rtld(GNU_HASH)
+
+$ elfdeps --provides /usr/lib64/libpython3.12.so
+libpython3.12.so.1.0()(64bit)
+```
+
+```shell-session
+$ elfdeps --provides /lib64/libc.so.6
+libc.so.6(GLIBC_2.2.5)(64bit)
+libc.so.6(GLIBC_2.2.6)(64bit)
+libc.so.6(GLIBC_2.3)(64bit)
+...
+libc.so.6(GLIBC_2.36)(64bit)
+libc.so.6(GLIBC_2.38)(64bit)
+libc.so.6(GLIBC_ABI_DT_RELR)(64bit)
+libc.so.6(GLIBC_PRIVATE)(64bit)
+libc.so.6()(64bit)
+```
+
+## RPM
+
+In Fedora-based distributions, RPM packages provide and require virtual packages with ELF sonames and versions. The package manager can install virtual provides.
+
+The `python3` base package depends on `libpython3.12.so.1.0()(64bit)` and `libc.so.6(GLIBC_2.34)(64bit)`:
+
+```shell-session
+$ rpm -qR python3
+libc.so.6()(64bit)
+libc.so.6(GLIBC_2.2.5)(64bit)
+libc.so.6(GLIBC_2.34)(64bit)
+libpython3.12.so.1.0()(64bit)
+...
+rtld(GNU_HASH)
+```
+
+The `python3-libs` package virtually provides `libpython3.12.so.1.0()(64bit)`:
+
+```shell-session
+$ rpm -qP python3-libs
+bundled(libb2) = 0.98.1
+libpython3.12.so.1.0()(64bit)
+libpython3.so()(64bit)
+python-libs = 3.12.3-2.fc39
+python3-libs = 3.12.3-2.fc39
+python3-libs(x86-64) = 3.12.3-2.fc39
+python3.12-libs = 3.12.3-2.fc39
+```
+
+```shell-session
+$ sudo dnf install 'libc.so.6(GLIBC_2.34)(64bit)' 'libpython3.12.so.1.0()(64bit)'
+Package glibc-2.38-18.fc39.x86_64 is already installed.
+Package python3-libs-3.12.3-2.fc39.x86_64 is already installed.
+Dependencies resolved.
+Nothing to do.
+Complete!
+```

pyproject.toml

@@ -22,7 +22,6 @@ classifiers = [
     "Programming Language :: Python",
     "Programming Language :: Python :: 3",
     "Programming Language :: Python :: 3 :: Only",
-    "Programming Language :: Python :: 3.9",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
@@ -70,7 +69,21 @@ select = [
   "TID", # flake8-tidy-imports
 ]
 ignore = [
+    "UP006",  # not compatible with 3.9 type annotations
 ]
 
 [tool.ruff.lint.isort]
 known-first-party = ["elfdeps"]
+
+[tool.mypy]
+python_version = "3.9"
+warn_return_any = true
+warn_unused_configs = true
+
+[[tool.mypy.overrides]]
+module = ["elfdeps.*"]
+disallow_untyped_defs = true
+
+[[tool.mypy.overrides]]
+module = ["elftools.*"]
+ignore_missing_imports = true

src/elfdeps/__main__.py

@@ -3,6 +3,7 @@
 import argparse
 import pathlib
 import pprint
+import typing
 
 from ._elfdeps import ELFDeps
 
@@ -46,16 +47,24 @@
     dest="require_interp",
     help="Include ELF interpreter name",
 )
+parser.add_argument(
+    "--unique",
+    "-u",
+    action="store_true",
+    dest="unique",
+    help="Remove duplicate entries",
+)
 
 
-def main(argv=None):
+def main(argv: typing.Optional[typing.List[str]] = None) -> None:
     args = parser.parse_args(argv)
     e = ELFDeps(
         args.filename,
         soname_only=args.soname_only,
         fake_soname=args.fake_soname,
         filter_soname=args.filter_soname,
         require_interp=args.require_interp,
+        unique=args.unique,
     )
     if args.provides:
         for p in e.info.provides:

src/elfdeps/_elfdeps.py

@@ -11,14 +11,15 @@
 import os
 import pathlib
 import stat
+import typing
 
 from elftools.elf.constants import VER_FLAGS
 from elftools.elf.dynamic import DynamicSection
 from elftools.elf.elffile import ELFFile
 from elftools.elf.gnuversions import GNUVerDefSection, GNUVerNeedSection
 
 
-@dataclasses.dataclass(slots=True, frozen=True)
+@dataclasses.dataclass(frozen=True, order=True)
 class SOInfo:
     """Shared object information"""
 
@@ -38,16 +39,19 @@ def __repr__(self) -> str:
 
 @dataclasses.dataclass
 class ELFInfo:
+    # requires and provides are ordered by occurence in ELF metadata and
+    # can contain duplicate entries. The order can be different than output
+    # of elfdeps.c, but that usually does not matter.
     requires: list[SOInfo]
     provides: list[SOInfo]
-    machine: str | None = None
+    machine: typing.Optional[str] = None
     is_dso: bool = False
     is_exec: bool = False
     got_debug: bool = False
     got_hash: bool = False
     got_gnuhash: bool = False
-    soname: str | None = None
-    interp: str | None = None
+    soname: typing.Optional[str] = None
+    interp: typing.Optional[str] = None
     marker: str = ""
 
 
@@ -79,6 +83,7 @@ def __init__(
         fake_soname: bool = True,
         filter_soname: bool = False,
         require_interp: bool = False,
+        unique: bool = True,
     ) -> None:
         if not isinstance(filename, pathlib.Path):
             raise TypeError(f"filename is not a pathlib.Path: {type(filename)}")
@@ -87,11 +92,13 @@ def __init__(
         self.fake_soname = fake_soname
         self.filter_soname = filter_soname
         self.require_interp = require_interp
+        self.unique = unique
 
         self.info = ELFInfo(
             requires=[],
             provides=[],
         )
+        self._seen: typing.Set[typing.Tuple[bool, SOInfo]] = set()
         with self.filename.open("rb") as f:
             self._process_file(f)
 
@@ -137,23 +144,30 @@ def _process_file(self, f: io.BufferedReader) -> None:
             # direct add
             self.info.requires.append(SOInfo(self.info.interp, version="", marker=""))
 
-    def _add_soinfo(self, provides: bool, soname: str, version: str | None) -> bool:
+    def _add_soinfo(
+        self, provides: bool, soname: str, version: typing.Optional[str]
+    ) -> None:
         if skip_soname(soname, filter_soname=self.filter_soname):
-            return False
+            return
         version = version if version else ""
         marker = self.info.marker or ""
         soinfo = SOInfo(soname, version, marker)
+
+        key = (provides, soinfo)
+        if self.unique and key in self._seen:
+            return
+        self._seen.add(key)
+
         if provides:
             self.info.provides.append(soinfo)
         else:
             self.info.requires.append(soinfo)
-        return True
 
-    def add_provides(self, soname: str, version: str | None = None) -> bool:
-        return self._add_soinfo(True, soname, version)
+    def add_provides(self, soname: str, version: typing.Optional[str] = None) -> None:
+        self._add_soinfo(True, soname, version)
 
-    def add_requires(self, soname: str, version: str | None = None) -> bool:
-        return self._add_soinfo(False, soname, version)
+    def add_requires(self, soname: str, version: typing.Optional[str] = None) -> None:
+        self._add_soinfo(False, soname, version)
 
     @property
     def gen_requires(self) -> bool:
@@ -200,7 +214,7 @@ def process_verdef(self, sec: GNUVerDefSection) -> None:
 
         processVerDef(Elf_Scn *scn, GElf_Shdr *shdr, elfInfo *ei)
         """
-        soname: str | None = None
+        soname: typing.Optional[str] = None
         for verdef, vernaux in sec.iter_versions():
             for aux in vernaux:
                 if not aux.name:
@@ -242,12 +256,13 @@ def process_dynamic(self, sec: DynamicSection) -> None:
             elif d_tag == "DT_NEEDED":
                 self.add_requires(tag.needed)
 
-    def process_prog_headers(self, elffile: ELFFile) -> str | None:
+    def process_prog_headers(self, elffile: ELFFile) -> typing.Optional[str]:
         """Get interpreter from PT_INTERP segment
 
         void processProgHeaders(elfInfo *ei, GElf_Ehdr *ehdr)
         """
         for seg in elffile.iter_segments("PT_INTERP"):
-            return seg.get_interp_name()
+            interp: str = seg.get_interp_name()
+            return interp
         else:
             return None

tox.ini

@@ -1,24 +1,31 @@
 # SPDX-License-Identifier: Apache-2.0
 
 [tox]
-envlist=py{9,10,11,12},lint
+envlist=py3{9,10,11,12},lint
 
 [testenv]
+package = wheel
+wheel_build_env = pkg
 extras = test
 commands =
     python -m pytest tests
 
 [testenv:lint]
-deps=
+basepython = python3
+deps =
     ruff
+    mypy
+    pytest-stub
 commands =
     ruff check src tests
     ruff format --check src tests
+    mypy src tests
 skip_install = true
 skip_sdist = true
 
 [testenv:fix]
-deps=
+basepython = python3
+deps =
     ruff
 commands =
     ruff format src tests