We have a lot of jobs downstream of the `build-binary-linux` job, but
the job is significantly slower than the other binary builds because we
need to configure musl. Instead, we split this into two jobs (as it was
before https://github.com/astral-sh/uv/pull/2309#discussion_r1520101330)
to speed things up.
The libc job takes ~1m and its _downstream_ jobs finish before the musl
build does. The musl job takes ~5m.
## Summary
In preview mode on windows, register und un-register the managed python build standalone installations in the Windows registry following PEP 514.
We write the values defined in the PEP plus the download URL and hash. We add an entry when installing a version, remove an entry when uninstalling and removing all values when uninstalling with `--all`. We update entries only by overwriting existing values, there is no "syncing" involved.
Since they are not official builds, pbs gets a prefix. `py -V:Astral/CPython3.13.1` works, `py -3.13` doesn't.
```
$ py --list-paths
-V:3.12 * C:\Users\Konsti\AppData\Local\Programs\Python\Python312\python.exe
-V:3.11.9 C:\Users\Konsti\.pyenv\pyenv-win\versions\3.11.9\python.exe
-V:3.11 C:\Users\micro\AppData\Local\Programs\Python\Python311\python.exe
-V:3.8 C:\Users\micro\AppData\Local\Programs\Python\Python38\python.exe
-V:Astral/CPython3.13.1 C:\Users\Konsti\AppData\Roaming\uv\data\python\cpython-3.13.1-windows-x86_64-none\python.exe
```
Registry errors are reported but not fatal, except for operations on the company key since it's not bound to any specific python interpreter.
On uninstallation, we prune registry entries that have no matching Python installation (i.e. broken entries).
The code uses the official `windows_registry` crate of the `winreg` crate.
Best reviewed commit-by-commit.
## Test Plan
We're reusing an existing system check to test different (un)installation scenarios.
In the interest of expanding these tests and debugging weird behaviors,
I've moved the smoke tests out of the `cargo test` job and into
dedicated `smoke test` jobs. We explicitly build `uvx` in the `build
binary` jobs instead of relying on the implicit build for the test run.
I also added a `uvx` test case to the smoke tests: `uvx ruff --version`
See https://github.com/astral-sh/uv/issues/4204 for motivation
This doesn't really reach the user experience I'd expect — i.e., we end
up saying a virtual environment "does not exist" which is a little
silly. However, I think improving the error messaging on interpreter
queries in general should be solved separately. I did one small
"general" change in
89e11d0222
— otherwise we don't show the message at all.
---------
Co-authored-by: konsti <konstin@mailbox.org>
When using the standard Windows runners (as opposed to the _larger_
GitHub runners), an undocumented `D:` drive is available and performant.
We can save some money on by using this on a standard runner instead of
a larger runner with an ReFS drive. Switching to the `D:` drive was not
acceptable for `cargo test` >25m runtime.
Inspired by https://github.com/pypa/pip/pull/13129
See https://github.com/actions/runner-images/issues/8755
Timings (grain of salt — GitHub is super noisy):
- clippy: 2m 18s -> 2m 11s
- build binary: 2m 3s -> 2m 35s
- trampoline check (x86-64): 2m 32s -> 1m 50s (other architectures
similar)
- trampoline test (x86-64): 4m 12s -> 6m 7s
- trampoline test (i686): 6m 44s -> 5m 35s
Previously, we couldn't use a DevDrive
(https://github.com/astral-sh/uv/pull/3522#issuecomment-2111448930)
because our Windows version was not sufficient.
Recently, I upgraded our larger runners to Windows 2025 preview
(https://github.com/astral-sh/uv/pull/10298) which I presume has support
for this.
I removed ReFS in
953c3535c3
which didn't seem to do anything to performance.
I also found some notes on "trusted" DevDrives and "disabling anti-virus
filtering" which I simply have to try.
The shellcheck action we uses misses some files, so they fell out of
spec for what we support. This PR first and foremost adds them to the
scanning list, and then fixes the issues found.
Fixes#7480
I'm renaming our runners to be more explicit about their size,
architecture, and version.
Switching to Windows 2025 over 2022 in some of our jobs in the hope that
it's faster.
Enable `lzma-sys/static` through the performance feature not only in uv,
but in uv-dev and uv-bench too, to avoid the system dependency on
`liblzma-dev`.
Ref #9880
When trying to upload without a password but with the keyring, check
that the keyring has a password for the upload URL and username and warn
if it doesn't.
Fixes#8781
## Summary
In the Windows Clippy job, the workspace is transferred to
`UV_WORKSPACE`. However, `cargo clippy` continues to execute in the
`github.workspace`, and `Swatinem/rust-cache` only caches the
`UV_WORKSPACE/target`, resulting in `cargo clippy` having no cache.
This adjustment will take effect when any changes are made to
`Cargo.toml` or `Cargo.lock`, prompting `Swatinem/rust-cache` to updat
the cache.
Incorporating #8637 into #8458
- Adds `python-managed` feature selection to Windows CI for `python
install` tests
- Adds trampoline sniffing utilities to `uv-trampoline-builder`
- Uses a trampoline to install Python executables into the `PATH` on
Windows
Updates `uv python install` to link `python3.x` in the executable
directory (i.e., `~/.local/bin`) to the the managed interpreter path.
Includes
- #8569
- #8571
Remaining work
- #8663
- #8650
- Add an opt-out setting and flag
- Update documentation
Currently, our trampoline is used to convert `<command> [args]` to
`python <command> [args]` for script entrypoints installed into virtual
environments. For #8458, it'd be nice to convert a shim `python3.12
[args]` to `python [args]`. Here, we modify the trampolines to support
this use-case.
The only change we really need here is to avoid injecting `<command>`
into the child process. We change the "magic number" at the end of the
trampoline executables from `UVUV` to `UVSC` and `UVPY` which define
"script" and "python" variants to the trampoline. We then omit the
`<command>` injection in the latter case. We also omit writing the zip
script payload.
To support construction of the new variant, a new
`uv-trampoline-builder` crate is introduced — this avoids requirements
on `uv-install-wheel` in future work. I also use `uv-trampoline-builder`
to consolidate some of the test setup for `uv-trampoline`.
There should be no backwards compatibility concerns, since trampolines
are fully self-referential.
I rebased to fix the commits at the end, as this took many iterations to
get working via CI. This should roughly be reviewable by commit if you
prefer.
It seems unintentional that we basically did nothing on these
alternative platforms? It seems like an artifact from some previous
change.
I'm not sure it's worth running Clippy multiple times. We could also
just reduce the matrix here.
Zanie Blue
konsti
Divyanshu Kalra
Charlie Marsh
Aria Desires
renovate[bot]
Jo
Mathieu Kniewallner
Hamir Mahal