See https://github.com/astral-sh/uv/issues/16368
Current behavior of `uv sync`:
* Requested: `1.0.0+cpu`, Installed: `1.0.0`: Install new package
* Requested: `1.0.0+cpu`, Installed: `1.0.0+cu128`: Install new package
* Requested: `1.0.0`, Installed: `1.0.0+cpu`: Keep installed package
The new behavior is to always reinstall when the local version part is
different, for consistency and to fix torch.
This behavior happens because internally, we translate the version from
the lockfile to an `=={version}` request, and version matching says
local versions are allowed if the specifier has no local version.
This is a (minor) behavior change: When running `uv sync` in a venv with
a package after installing the same package with the same version except
the local version, uv will now remove the installed package and use the
one from the lockfile instead. This seems more correct, as the other
version was not matching the lockfile. There is still a gap as what we
actually want to track is the index URL (or even better, the package
hash), but as that information is not tracked in the venv, checking the
local version is the next bests thing we can do. The main motivation is
fixing torch, our main user of packages with both local and non-local
versions (https://github.com/astral-sh/uv/issues/16368).
<!--
Thank you for contributing to uv! To help us out with reviewing, please
consider the following:
- Does this pull request include a summary of the change? (See below.)
- Does this pull request include a descriptive title?
- Does this pull request include references to any relevant issues?
-->
## Summary
<!-- What's the purpose of the change? What does it do, and why? -->
> 4. Hyphenate compound words, e.g., use "platform-specific" not
"platform specific".
Applies to "command line arguments" -> "command-line arguments".
## Summary
Changes the `length` parameter to `PythonDownloadReporter::new` to an
`Option<u64>` to resolve the issue of setting a single-item reporter
length.
Resolves: #15404
## Test Plan
- build test
- ran test suite
## Summary
Upgrades to the latest Rust edition and pyo3 version.
Change initialized module to "Inline Declaration" format.
https://pyo3.rs/v0.27.1/module.html#inline-declaration
The output of `maturin new` is also updating to the new declarative
format
342239a95a
## Test Plan
Updated the relevant snapshot tests and to confirm ran
`cargo nextest run --all-features --no-fail-fast maturin`
Also used the updated template to generate a library in a different
project with
```
cargo run -- init --lib --build-backend maturin --name try-template ../_OTHER_PROJECT_/backend/try-template
```
and the generated workspace member functioned as expected.
## Summary
<!-- What's the purpose of the change? What does it do, and why? -->
This pull request updates the GitHub Actions workflow documentation to
use the latest version of the `actions/setup-python` action. This
ensures compatibility with recent improvements and bug fixes in the
action.
Workflow version updates:
* Updated the `uses: actions/setup-python` step from version `v5` to
`v6` in two separate workflow job examples in
`docs/guides/integration/github.md`.
[[1]](diffhunk://#diff-e864b910728c865e8e16ddb7892761fc2ef4838f2bf256eb1e20c35b24edd9fbL96-R96)
[[2]](diffhunk://#diff-e864b910728c865e8e16ddb7892761fc2ef4838f2bf256eb1e20c35b24edd9fbL119-R119)
Made to address this comment:
https://github.com/astral-sh/uv/pull/16103#discussion_r2437498249
This PR sorts the distributions collected by
`FlatIndexClient::read_from_directory` (used for `--find-links`) so
results are ordered deterministically by filename and index.
<!--
Thank you for contributing to uv! To help us out with reviewing, please
consider the following:
- Does this pull request include a summary of the change? (See below.)
- Does this pull request include a descriptive title?
- Does this pull request include references to any relevant issues?
-->
## Summary
<!-- What's the purpose of the change? What does it do, and why? -->
I am a new uv user and I was reading the docs to better understand the
project scope & best practices. The section on [signal
forwarding](https://docs.astral.sh/uv/concepts/projects/run/#signal-handling)
with `uv run` caught my eye because I've used tools that use SIGHUP to
trigger config reloads or SIGUSR1/2 to enable debugging/profiling/etc so
I was a little concerned about using a runner that might block those
signals. After some searching in issues/PRs, I found that this behavior
was actually [changed earlier this
year](https://github.com/astral-sh/uv/pull/13017) to forward additional
signals (awesome!) and thought I would update the docs and save the next
person/llm from thinking their tool won't work as expected if it uses
custom signal handling.
Thanks for all your great work!
P.S. If you think it makes more sense to explicitly list all forwarded
signals as opposed to just the exclusions, I'm happy to edit.
This should avoid hitting the GitHub API to determine the latest
release, and more generally we should not automatically fetch the latest
version of Maturin in our release pipeline as it opens us to
supply-chain attacks.
## Summary
The latest version should fix our rate-limiting issues.
## Test Plan
See what happens in CI.
Signed-off-by: William Woodruff <william@astral.sh>
Closes https://github.com/astral-sh/uv/issues/16218
This occurs when using a mounted file system in Docker.
We're almost always removing a virtual environment to replace it, and
removing the parent directory isn't necessary for that.
Provide a good error message when the discovered workspace members
mismatch with the locked workspace members in `uv export --frozen`,
instead of panicking.
Fixes#16406
I had some qualms with https://github.com/astral-sh/uv/pull/16318
1. The snapshot was specific to uv's managed interpreter due to
inclusion of the externally-managed output. This will break downstream
distros. We either need to filter the message, or, as done here, install
a managed interpreter.
2. It had a custom filter for the interpreter path, which we shouldn't
need. If needed, we should fix that in the test context.
3. We already had an implicit hint to create a virtual environment. The
change in styling of the new hint hint following it is confusing. It's
also confusing to hint creating a virtual environment when `--system`
was used.
4. There were unresolved requested changes to the language for the
message / it didn't fit stylistic with our existing ones.
5. The message was also very confusing for a uv-managed interpreter,
which is both a "system" Python interpreter (in that it's global) and
the opposite of a "system" interpreter per UV_PYTHON_PREFERENCE.
Also problematic, but not addressed here, is that there are other
commands that display an externally-managed message, e.g., `uv pip
sync`, but #16318 was just limited to `pip install`. We should not just
implement this in one place — I'll open a tracking issue to consolidate
and reuse the logic.
Hello,
# Summary
This PR fixes the confusing error message when running `uv lock --check`
with an outdated lockfile.
Previously, the error message incorrectly stated that `--locked` was
provided, even when the user used `--check`.
Now, the error message correctly indicates which flag was used: either
`--check` or `--locked`.
This closes#14105.
# Test plan
- I updated the existing integration test (`check_outdated_lock` in
`lock.rs`) to verify the new error message includes the correct flag.
- I ran existing tests to ensure I have no introduced regressions for
other commands.
## Summary
I tried to fix minor typos in the project
---------
Co-authored-by: Samuel Rigaud <rigaud@gmail.com>
Co-authored-by: konstin <konstin@mailbox.org>
## Summary
Fixes the logic for escaping a double quoted string in PowerShell by not
escaping a backslash but escaping the Unicode double quote variants that
PowerShell treats the same as the ASCII double quotes.
<img width="685" height="99" alt="image"
src="https://github.com/user-attachments/assets/ac1368c2-d915-4e49-b67f-ac71ee0f7d46"
/>
## Test Plan
There does not seem to be any tests for this. I'm fairly new to rust but
happy to add some if someone can point me in the right direction.
---------
Co-authored-by: Aria Desires <aria.desires@gmail.com>
Hello,
# Summary
This PR makes the error message clearer when you try to install packages
into an externally managed Python environment with the `--system` flag.
Now, instead of just failing, the error explains that you're hitting
this because you explicitly used `--system`.
This closes#15639.
# Test plan
- I added a new integration test (`install_with_system_interpreter` in
`pip_install.rs`) that checks the error message includes the hint.
- I tried to run `uv pip install --system -r requirements.txt` to see
the actual error message in action, but had permission issues.
Fixes#16340
## Summary
Some package registries (PyTorch, corporate PyPI mirrors) return `403
Forbidden` when a package is not found, instead of `404 Not Found`. The
previous error message incorrectly suggested this was always an
authentication issue, causing confusion for users.
This PR updates the error hint to clarify that a 403 error could
indicate either missing authentication credentials OR that the package
doesn't exist on the index.
## Test Plan
- Updated existing snapshot test in `crates/uv/tests/it/edit.rs` to
reflect the new error message
- The change is purely a message improvement with no behavioral changes
## Changes
### Before
hint: An index URL (https://example.com/simple) could not be queried due
to a lack of valid authentication credentials (403 Forbidden).
### After
hint: An index URL (https://example.com/simple) returned a 403 Forbidden
error. This could indicate missing authentication credentials, or the
package may not exist on this index.
## Files Changed
- `crates/uv-resolver/src/pubgrub/report.rs` - Updated error message
- `crates/uv/tests/it/edit.rs` - Updated snapshot test expectation
---------
Co-authored-by: eun2ce <eun2ce@eun2ceui-MacBookPro.local>
Co-authored-by: konstin <konstin@mailbox.org>
## Summary
fixes issue #15938
- show platform wheel hint with a concrete
`tool.uv.required-environments` example so users know how to configure
compatibility
- add `WheelTagHint::suggest_environment_marker` to pick a sensible
environment marker based on the available wheel tags
- update the `sync_required_environment_hint` integration snapshot to
expect the new multi-line hint
## Test Plan
cargo test --package uv --test it --
sync::sync_required_environment_hint
An opinionated write-up on why Python packaging needs metadata
consistency, and that we need to extend metadata to accommodate ML and
scientific users.
I didn't add a paragraph related to CUDA or accelerators in general and
wheel variants, as this is currently support neither by wheel tags nor
by PEP 508 markers, so it's not a strict metadata consistency concern,
plus this would get outdated quickly as wheel variants progress.
<!--
Thank you for contributing to uv! To help us out with reviewing, please
consider the following:
- Does this pull request include a summary of the change? (See below.)
- Does this pull request include a descriptive title?
- Does this pull request include references to any relevant issues?
-->
## Summary
<!-- What's the purpose of the change? What does it do, and why? -->
The warning shown when `egg-info` lacks `top_level.txt` incorrectly
warns about missing `top-level.txt`
ee2649feaa/crates/uv-install-wheel/src/uninstall.rs (L179-L188)ee2649feaa/crates/uv/src/commands/pip/operations.rs (L687-L693)
The non fatal warning with incorrect filename was introduced in
https://github.com/astral-sh/uv/pull/6881 which changed previous fatal
error https://github.com/astral-sh/uv/issues/6872 with correct
`top_level.txt` output.
## Test Plan
<!-- How was it tested? -->
Updated unit test to reflect change in warning
## Summary
Last time I ran the `update_schemastore.py` script in ty, due to what I
assume was a `npm` version mismatch, the `package-lock.json` file was
updated while running `npm install` in the `schemastore`. Due to the use
of `git commit -a`, it was accidentally included in the commit for the
semi-automated schemastore PR. The solution here is to only add the
actual file that we want to commit.
Same as https://github.com/astral-sh/ty/pull/1391
## Test Plan
I did a dry-run of this script (by commenting out the final `push`) and
verified that the commit did include the schema, but not the updated
`package-lock.json` file.
This PR enables `--torch-backend=auto` to automatically detect Intel
GPUs. It follows up on
[#14386](https://github.com/astral-sh/uv/pull/14386).
On Windows, detection is implemented by querying the
`Win32_VideoController` class via the [WMI
crate](https://github.com/ohadravid/wmi-rs/tree/v0.16.0).
Currently, Intel GPUs (XPU) do not depend on specific driver or toolkit
versions to determine which PyTorch wheel to use.
<!--
Thank you for contributing to uv! To help us out with reviewing, please
consider the following:
- Does this pull request include a summary of the change? (See below.)
- Does this pull request include a descriptive title?
- Does this pull request include references to any relevant issues?
-->
## Summary
HTTP1.1 [RFC 9112 -
HTTP/1.1](https://www.rfc-editor.org/rfc/rfc9112.html#name-status-line)
section 4 defines the response status code to optionally include a text
description (human readable) of the reason for the status code.
[RFC9113 - HTTP/2](https://www.rfc-editor.org/rfc/rfc9113) is the HTTP2
protocol standard and the response status only considers the [status
code](https://www.rfc-editor.org/rfc/rfc9113#name-response-pseudo-header-fiel)
and not the reason phrase, and as such important information can be lost
in helping the client determine a route cause of a failure.
As per discussion on this
[PR](https://github.com/astral-sh/uv/pull/15979) the current feeling is
that implementing the RFC9457 standard might be the preferred route.
This PR makes those changes to aid the discussion which has also been
moved to the [PEP
board](https://discuss.python.org/t/block-download-of-components-when-violating-policy/104021/1)
## Test Plan
Pulling components that violate our policy over HTTP2 and without any
RFC9457 implementation the following message is presented to the user:
<img width="1482" height="104" alt="image"
src="https://github.com/user-attachments/assets/0afcd0d8-ca67-4f94-a6c2-131e3b6d8dcc"
/>
With the RFC9457 standard implemented, below you can see the advantage
in the extra context as to why the component has been blocked:
<img width="2171" height="127" alt="image"
src="https://github.com/user-attachments/assets/25bb5465-955d-4a76-9f30-5477fc2c866f"
/>
---------
Co-authored-by: konstin <konstin@mailbox.org>
konstin
Hugo van Kemenade
elle
Matthijs Kok
renovate[bot]
Basil Sh.
Zanie Blue
SADIK KUZU
liam
Brian Dentino
William Woodruff
github-actions[bot]
Charles-Meldhine Madi Mnemoi
Samuel Rigaud
Jordan Borean
Andrei Berenda
eun2ce
Parham MohammadAlizadeh
Björn Dahlgren
David Peter
Yu, Guangye
Mark Dodgson