We stopped shipping the distributions in
6fa8fe9 in October 2023. At the time we
preserved build support just in case.

I don't believe anyone has complained about the loss of x86 Linux
distributions.

I don't believe there is compelling value in PBS supporting x86 Linux
distributions.

If someone wants to run Python on x86 Linux, they can always use
CPython's official build systen, which still supports targeting x86
Linux.

This commit rips out our support building for x86 Linux.

…627)

The `xcb` Dockerfile / image existed to provide a host build of CPython,
which was required to build libxcb, xcb-proto, and tk. The existence of
this image predated our building a host CPython in order to build
CPython itself.

With PBS now unconditionally building a host CPython, we can use our
host CPython during the builds of these packages, eliminating the need
for the xcb image.

This commit makes that transition.

We also upgrade libxcb and xcb-proto to the latest version as part of
this migration. Older versions of libxcb aren't compatible with Python
3.12+ due to removal of the `imp` module.

Closes #620

I am not sure why this change is not validated on pull request :(

Follows #630

Before commit 71882f2 (May 2022), we
built LLVM from source. This employed sccache to make builds faster.

After that commit, sccache hasn't effectively been used. But there were
remnants of sccache in the code base.

This commit removes all references to sccache in the repo.

We configure ncurses with defaults for TERMINFO_DIRS that work on I can
find (it works on Debian family, Red Hat family, NixOS, and Alpine), so
we shouldn't phrase either the title or the body as if this is a problem
that affects most people or setting the environment variable is a
necessary thing in general. (Also I think even in cases where we don't
have terminfo for your terminal, backspace usually still works, as
evidenced by `TERM=potato python`).

musl distributions are now dynamic binaries by default and dynamic
loading works.

The libcrypt transition is behind us.

The python3-config file (since its first version, I think) knows how to
correct the prefix in output. While it does contain hard-coded paths
internally, it uses that for the purpose of fixing its output, so I
don't think that fact is helpful to call out.

python-build-standalone users who get it through uv are (hopefully) not
affected by absolute paths in sysconfig; call this out explicitly for
the benefit of people evaluting whether uv will work for them. Also
mention sysconfigpatcher for everyone else.

In general, tighten up the content.

#623)

We need this to get inherited to dependent libraries, i.e., to extension
modules.

We were killed in
https://github.com/astral-sh/python-build-standalone/actions/runs/15085967950/job/42408534018

astral-sh/uv#13610 reported a misbehavior that is the result of a
subclass of str incorrectly having its ->tp_as_number->nb_add slot
filled in with the value of PyUnicode_Type->tp_as_sequence->sq_concat.
There are some times when this is an appropriate thing to do iwhen
subclassing, but this is not one of them. The logic to prevent it in
this case relies on two helper functions in the file, wrap_binaryfunc
and wrap_binaryfunc_l, having different addresses, even though they
contain identical code.

For some reason BOLT does not do this optimization in the shared library
(even though those are static functions and not exported), so we only
started seeing this in the static build.

BOLT in LLVM 20+ supports "safe" code folding, which uses heuristics about
relocations to determine whether a function's address is used in any way other
than a call. This seems to be enough to fix the issue. Add a patch to switch
to -icf=safe, submitted upstream as python/cpython#134642