Este conteúdo não está disponível no idioma selecionado.
Chapter 5. Packaging Python 3 RPMs
You can install Python packages on your system by using the DNF package manager. DNF uses the RPM package format, which offers more downstream control over the software.
Packaging a Python project into an RPM package provides the following advantages compared to native Python packages:
- Dependencies on Python and non-Python packages are possible and strictly enforced by the DNF package manager.
- You can cryptographically sign the packages. With cryptographic signing, you can verify, integrate, and test contents of RPM packages with the rest of the operating system.
- You can execute tests during the build process.
The packaging format of native Python packages is defined by Python Packaging Authority (PyPA) Specifications. Historically, most Python projects used the distutils or setuptools utilities for packaging and defined package information in the setup.py file. However, possibilities of creating native Python packages have evolved over time:
-
To package Python software that uses the
setup.pyfile, follow this document. -
To package more modern packages with
pyproject.tomlfiles, see theREADMEfile in pyproject-rpm-macros. Note thatpyproject-rpm-macrosis included in the CodeReady Linux Builder (CRB) repository, which contains unsupported packages, and it can change over time to support newer Python packaging standards.
5.1. A spec file description for an example Python package Copiar o linkLink copiado para a área de transferência!
An RPM spec file for Python projects has some specifics compared to non-Python RPM spec files.
Note that it is recommended for any RPM package name of a Python library to include the python3- prefix.
See the notes about Python RPM spec files specifics in the following example of the python3-pello package.
- An example SPEC file for the pello program written in Python
- 1
- By defining the
python3_pkgversionmacro, you set which Python version this package will be built for. To build for the default Python version3.12, remove the line. - 2
- When packaging a Python project into RPM, always add the
python-prefix to the original name of the project. The project name here isPelloand, therefore, the name of the Source RPM (SRPM) ispython-pello. - 3
BuildRequiresspecifies what packages are required to build and test this package. InBuildRequires, always include items providing tools necessary for building Python packages:python3-develand the relevant projects needed by the specific software that you package, for example,python3-setuptoolsor the runtime and testing dependencies needed to run the tests in the%checksection.- 4
- When choosing a name for the binary RPM (the package that users will be able to install), add a versioned Python prefix. Use the
python3-prefix for the default Python 3.12. You can use the%{python3_pkgversion}macro, which evaluates to3for the default Python version3.12unless you set it to an explicit version, for example, when a later version of Python is available (see footnote 1). - 5
- The
%py3_buildand%py3_installmacros run thesetup.py buildandsetup.py installcommands, respectively, with additional arguments to specify installation locations, the interpreter to use, and other details.NoteUsing the
setup.py buildandsetup.py installcommands from thesetuptoolspackage is deprecated and will be removed in the future major RHEL release. You can use pyproject-rpm-macros instead. - 6
- The
%checksection runs the tests of the packaged project. The exact command depends on the project itself, but you can use the%pytestmacro to run thepytestcommand in an RPM-friendly way.
5.2. Common macros for Python 3 RPMs Copiar o linkLink copiado para a área de transferência!
In a Python RPM spec file, always use the macros for Python 3 RPMs rather than hardcoding their values.
You can redefine which Python 3 version is used in these macros by defining the python3_pkgversion macro on top of your spec file. For more information, see A spec file description for an example Python package. If you define the python3_pkgversion macro, the values of the macros described in the following table will reflect the specified Python 3 version.
| Macro | Normal Definition | Description |
|---|---|---|
| %{python3_pkgversion} | 3 | The Python version that is used by all other macros. Can be redefined to any future Python versions that will be added. |
| %{python3} | /usr/bin/python3 | The Python 3 interpreter. |
| %{python3_version} | 3.12 | The major.minor version of the Python 3 interpreter. |
| %{python3_sitelib} | /usr/lib/python3.12/site-packages | The location where pure-Python modules are installed. |
| %{python3_sitearch} | /usr/lib64/python3.12/site-packages | The location where modules containing architecture-specific extension modules are installed. |
| %py3_build |
Expands to the | |
| %py3_install |
Expands to the | |
| %{py3_shebang_flags} | sP |
The default set of flags for the Python interpreter directives macro, |
| %py3_shebang_fix |
Changes Python interpreter directives to |
5.3. Using automatically generated dependencies for Python RPMs Copiar o linkLink copiado para a área de transferência!
You can automatically generate dependencies for Python RPMs by using upstream-provided metadata.
Prerequisites
-
A
specfile for the RPM exists. For more information, see A spec file description for an example Python package.
Procedure
Include one of the following directories in the resulting RPM:
-
.dist-info .egg-infoThe RPM build process automatically generates virtual
pythonX.Ydistprovides from these directories, for example:python3.12dist(pello)
python3.12dist(pello)Copy to Clipboard Copied! Toggle word wrap Toggle overflow The Python dependency generator then reads the upstream metadata and generates runtime requirements for each RPM package using the generated
pythonX.Ydistvirtual provides. Example of a generated requirements tag:Requires: python3.12dist(requests)
Requires: python3.12dist(requests)Copy to Clipboard Copied! Toggle word wrap Toggle overflow
-
-
Inspect the generated
Requires. -
To remove some of the generated
Requires, modify the upstream-provided metadata in the%prepsection of thespecfile. -
To disable the automatic requirements generator, include the
%{?python_disable_dependency_generator}macro above the main package’s%descriptiondeclaration.