https://fedoraproject.org/wiki/Changes/Python_Upstream_Architecture_Names
== Summary == Use CPython upstream architecture naming in Fedora's Python ecosystem (mostly in filenames) instead of the previously patched Fedora names. For example, have <code>/usr/lib64/python3.9/lib-dynload/array.cpython-39-powerpc64le-linux-gnu.so</code> instead of <code>/usr/lib64/python3.9/lib-dynload/array.cpython-39-ppc64le-linux-gnu.so</code>. This makes packaging of Python itself a tad trickier, but it moves Fedora's Python closer to upstream and solves interoperability problems with ppc64le manylinux wheels. The change has impact only on '''ppc64le''' and '''armv7hl''' (considering the architectures built by koji.fedoraproject.org). Packages assuming the filenames always contain <code>%{_arch}-linux%{_gnu}</code> will need to be adapted.
== Owner == * Name: [[User:Churchyard|Miro Hrončok]] * Name: [[User:Lbalhar|Lumír Balhar]] * Email: python-maint@redhat.com
== Detailed Description ==
=== The Saga ===
==== A Long Time Ago in the Fedora Galaxy ====
Many releases ago, when Fedora wasn't being built for power and arm yet, the Python maintainers mapped the Python "platform triplet" or "multiarch name" to <code>%{_arch}-linux%{_gnu}</code>. This worked. For example, on x86_64, this is <code>x86_64-linux-gnu</code> on Fedora and this is consistent with the "platform triplet" used in filenames in upstream.
==== The Phantom Technical Debt ====
Later around the year 2015, as more architectures were added, Python build scripts were patched to use "the Fedora's architecture names":
* <code>powerpc</code> was changed to <code>ppc</code> https://src.fedoraproject.org/rpms/python3/c/0efd3d31 * the arm triplet was patched https://src.fedoraproject.org/rpms/python3/c/2f6352e7 * even the mips one https://src.fedoraproject.org/rpms/python3/c/4bc70e0c
At the time, that was a reasonable decision: the idea of cross-Linux builds was sci-fi, and Fedora was not trying to stay close to upstream as it is now (we had around 60 patches; today we're down to around 6).
==== Rise of the Manylinux Wheels ====
In the meantime, cross-Linux builds become a thing. The [https://www.python.org/dev/peps/pep-0513/ manylinux1 standard] was created in 2016, allowing to build Python wheels with compiled extension modules on one Linux platform and ship them to many. The first manylinux version only supported x86_64 and i686 and hence it was not impacted by Fedora's patching decisions.
The manylinux standard arguably made the upstream Python packaging ecosystem a much nicer place. Installing packages with compiled extension modules was no longer such a pain. One could just run <code>pip install numpy</code> and not worry about a disturbing lack of a Fortran compiler. For that reason, manylinux wheels become widely adopted by the most popular projects.
==== A New Architecture ====
With the third manylinux version -- [https://www.python.org/dev/peps/pep-0599/ manylinux2014] (created in 2019, named after the oldest Linux it supports -- CentOS 7), support for more architectures was introduced: x86_64, i686, aarch64, armv7l, ppc64, ppc64le, s390x. The adaption of new architectures is somehow slow, because the [https://github.com/pypa/manylinux#manylinux2014 official manylinux2014 containers] only currently (August 2020) exist for x86_64, i686, aarch64, ppc64le and s390x.
==== Revenge of the Patches ====
We have discovered [https://github.com/pypa/manylinux/issues/687 a problem with the ppc64le manylinux2014 wheels]: The CentOS 7 manylinux2014 container images ship upstream Python without RHEL/CentOS/EPEL patches. When an extension module is built there, it is named with an upstream named suffix: <code>.cpython-XY(m)-powerpc64le-linux-gnu.so</code>. The wheel is installable on Fedora (with Fedora's patches), but the module won't (even be considered for) import, because Fedora's Pythons expect the extension to be <code>.cpython-XY(m)-ppc64le-linux-gnu.so</code>.
In theory, we have the same problem on armv7hl, but there are no manylinux2014 containers available for that platform, so there are no such wheels out there yet (to our knowledge).
The same problem also exists the other way around, albeit it's arguably less severe. It is possible to build manylinux wheels on (some version of) Fedora or EL (using the Python from that distribution). However extension modules from such ppc64le wheels won't import on other Linux distributions.
==== The Workaround Awakens ====
To allow importing extension modules from ppc64le manylinux wheels, we have [https://src.fedoraproject.org/rpms/python3.9/pull-request/24 patched Pythons (3.5+) in Fedora] to consider both "Fedora's" and upstream extension suffixes when importing extension modules. This workarounds works well for users installing manylinux wheels on Fedora, but does not solve the problem when building the wheels on Fedora.
=== The Change ===
With this change proposal, we plan to [https://src.fedoraproject.org/rpms/python3.9/pull-request/28 switch to use the upstream architecture names] and keep the workaround to preserve backwards compatibility. When we do that the following will happen:
# The Python standard library extension module suffixes will change to <code>.cpython-39-powerpc64le-linux-gnu.so</code> and <code>.cpython-39-arm-linux-gnueabihf.so</code>. Python will still import extension modules with the legacy suffixes <code>.cpython-39-ppc64le-linux-gnu.so</code> and <code>.cpython-39-arm-linux-gnu.so</code>. Other architectures not built by koji.fedoraproject.org will also be renamed, see [https://src.fedoraproject.org/rpms/python3.9/pull-request/28#comment-0 the pull request for a complete regex]. This will happen for Python 3.7, 3.8, 3.5, 3.6 and 3.9. # The newly built Python packages with extension modules will also change the suffixes. Packages that assume the platform triplet is always <code>%{_arch}-linux%{_gnu}</code> (e.g. in the <code>%files</code> section) will need to be adapted (see the [[#New_Macros|New Macros]] section). A mix of legacy and upstream suffixes will co-exist and work together. # When safe, we will drop the workaround to support the legacy names. For example, when we initially package Python 3.10, it will be packaged without the workaround. On the other hand, older Python versions might never be able to drop it, because users will carry their own built extension modules from previous releases.
=== New Macros ===
For packagers' convenience we will add 2 new Python macros:
'''%python3_ext_suffix'''
Defined as:
%python3_ext_suffix %(%{__python3} -Ic "import sysconfig; print(sysconfig.get_config_var('EXT_SUFFIX'))")
Values will be:
* <code>.cpython-39-x86_64-linux-gnu.so</code> * <code>.cpython-39-powerpc64le-linux-gnu.so</code> on Fedora 34+ / <code>.cpython-39-ppc64le-linux-gnu.so</code> on older Fedoras * <code>.cpython-39-arm-linux-gnueabihf.so</code> on Fedora 34+ / <code>.cpython-39-arm-linux-gnu.so</code> on older Fedoras * etc.
Beware that due to chnages in Python, this macro only works with Python 3. For Python 3.4, the value is <code>.cpython-34m.so</code> on all architectures (similarly on older Python 3 versions, but we don't have such in Fedora or EPEL).
'''%python3_platform_triplet'''
Defined as:
%python3_platform_triplet %(%{__python3} -Ic "import sysconfig; print(sysconfig.get_config_var('MULTIARCH'))")
Values will be:
* <code>x86_64-linux-gnu</code> * <code>powerpc64le-linux-gnu</code> / <code>ppc64le-linux-gnu</code> * <code>arm-linux-gnueabihf</code> / <code>arm-linux-gnu</code> * etc.
Beware that due to changes in Python, this macro only works with Python versions 3.6 and newer.
Both macros will be backported to stable Fedoras and EPEL 7+ and will have the corresponding <code>%python_</code> variant.
== Feedback == Eighth_Doctor: it's the cleverest written change proposal I've seen in over a decade :)
== Benefit to Fedora ==
Users of ppc64le and armv7hl Fedora (and future RHEL) will have a closer-to-upstream Python experience and will no longer suffer from compatibility issues when they install or build manylinux wheels. The upstream-downstream balance will be restored.
== Scope == * Proposal owners: ** Review and merge https://src.fedoraproject.org/rpms/python3.9/pull-request/28 ** Add the new macros to all Fedoras and EPEL 7+ and mention them in the guidelines ** Monitor builds of Python packages with extension modules, fix issues if they arise. ** Backport to Python 3.5-3.8 ** Drop the workaround from Python 3.10 once packaged
* Other developers: Mostly nothing, adapt the <code>%files</code> section if needed * Release engineering: a check of an impact with Release Engineering is not needed * Policies and guidelines: the new macros will be documented in the Python packaging guidelines * Trademark approval: not needed for this Change * Alignment with Objectives: no
== Upgrade/compatibility impact == No significant user visible upgrade/compatibility problem is anticipated. Filenames will be different, but the old filenames are still supported. Scripts that hardcode filename assumptions might break.
== How To Test == On ppc64le, try to install a manylinux wheel and import from it. It should work on any Python ≥ 3.5. E.g.:
pip install simple-manylinux-demo python -c 'from dummyextension import extension'
On ppc64le, try to build a manylinux wheel and import from it on another Linux. It should work on any Python ≥ 3.5. E.g.:
pip wheel . # on some project with extension module auditwheel repair ...whl wormhole send ...whl # or any other way
On another ppc64le Linux (such as Debian or openSUSE):
wormhole receive ... pip install ...whl python -c 'from ... import ...'
You can also build a regular (non-manylinux) wheel on Fedora 33/32 and install and import it on Fedora 34. It should work. The other way around will most likely also work, unless Fedora 34 has an incompatible glibc update.
== User Experience == Users of ppc64le and armv7hl Fedora (and future RHEL) will have a closer-to-upstream Python experience and will no longer suffer from compatibility issues when they install or build manylinux wheels.
== Dependencies == No known dependencies. May the force be with us.
== Contingency Plan == * Contingency mechanism: Revert the change and rebuild all affected packages. * Contingency deadline: Soft before the mass rebuild, so we could leverage it for the revert-rebuilds. Hard before the beta freeze. * Blocks release? No * Blocks product? No
== Documentation == This page is the documentation.