## Summary
Right now, if a workspace member is first created by way of being a dev
dependency on another member, we end up duplicating it in the graph.
Instead, we should create all the roots upfront; all subsequent node
creations are robust to existing nodes.
Closes
https://github.com/astral-sh/uv/issues/13673#issuecomment-2912196406.
## Summary
Resolves https://github.com/astral-sh/uv/issues/13580.
`uv self update --offline` should fail and exit early because
self-updating requires network connection.
## Test Plan
A snapshot test is added.
---------
Co-authored-by: Aria Desires <aria.desires@gmail.com>
Co-authored-by: konsti <konstin@mailbox.org>
## Summary
This should reduce the number of filesystem operations fairly
dramatically:
- Only query actual symlinks.
- Don't recurse into package bodies (huge).
- Only traverse once (rather than twice).
Closes https://github.com/astral-sh/uv/issues/13667.
Prior to this PR, there were numerous places where uv would leak
credentials in logs. We had a way to mask credentials by calling methods
or a recently-added `redact_url` function, but this was not secure by
default. There were a number of other types (like `GitUrl`) that would
leak credentials on display.
This PR adds a `DisplaySafeUrl` newtype to prevent leaking credentials
when logging by default. It takes a maximalist approach, replacing the
use of `Url` almost everywhere. This includes when first parsing config
files, when storing URLs in types like `GitUrl`, and also when storing
URLs in types that in practice will never contain credentials (like
`DirectorySourceUrl`). The idea is to make it easy for developers to do
the right thing and for the compiler to support this (and to minimize
ever having to manually convert back and forth). Displaying credentials
now requires an active step. Note that despite this maximalist approach,
the use of the newtype should be zero cost.
One conspicuous place this PR does not use `DisplaySafeUrl` is in the
`uv-auth` crate. That would require new clones since there are calls to
`request.url()` that return a `&Url`. One option would have been to make
`DisplaySafeUrl` wrap a `Cow`, but this would lead to lifetime
annotations all over the codebase. I've created a separate PR based on
this one (#13576) that updates `uv-auth` to use `DisplaySafeUrl` with
one new clone. We can discuss the tradeoffs there.
Most of this PR just replaces `Url` with `DisplaySafeUrl`. The core is
`uv_redacted/lib.rs`, where the newtype is implemented. To make it
easier to review the rest, here are some points of note:
* `DisplaySafeUrl` has a `Display` implementation that masks
credentials. Currently, it will still display the username when there is
both a username and password. If we think is the wrong choice, it can
now be changed in one place.
* `DisplaySafeUrl` has a `remove_credentials()` method and also a
`.to_string_with_credentials()` method. This allows us to use it in a
variety of scenarios.
* `IndexUrl::redacted()` was renamed to
`IndexUrl::removed_credentials()` to make it clearer that we are not
masking.
* We convert from a `DisplaySafeUrl` to a `Url` when calling `reqwest`
methods like `.get()` and `.head()`.
* We convert from a `DisplaySafeUrl` to a `Url` when creating a
`uv_auth::Index`. That is because, as mentioned above, I will be
updating the `uv_auth` crate to use this newtype in a separate PR.
* A number of tests (e.g., in `pip_install.rs`) that formerly used
filters to mask tokens in the test output no longer need those filters
since tokens in URLs are now masked automatically.
* The one place we are still knowingly writing credentials to
`pyproject.toml` is when a URL with credentials is passed to `uv add`
with `--raw`. Since displaying credentials is no longer automatic, I
have added a `to_string_with_credentials()` method to the `Pep508Url`
trait. This is used when `--raw` is passed. Adding it to that trait is a
bit weird, but it's the simplest way to achieve the goal. I'm open to
suggestions on how to improve this, but note that because of the way
we're using generic bounds, it's not as simple as just creating a
separate trait for that method.
PackageMetadata, for whatever reason, does not have a mirrored Wire type
so it was easy to not realize that it contains markers that need to be
complexified.
Fixes#13614
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
We were previously rendering messages for the info level, carrying
overhead in pubgrub which using `log::info!`. We avoid this by only
configuring `LevelFilter::INFO` if the durations layer exists.
I've confirmed that the default `Subscriber::max_level_hint` goes from
`INFO` to `OFF` and the profile skips `Incompatibility::display`.
## Summary
Closes#13612
We check if the git error message includes `not a git repository` to
figure out if the path isn't a Git repo or if Git's broken. This PR sets
`LC_ALL=C` when invoking `git rev-parse --is-inside-work-tree` so that
the error message doesn’t change based on the user’s locale settings.
There is a runtime issue with some of these builds
Here is the testing I've seen (❌ has bug, ✅ works):
* uv version (pbs version)
* ❌ uv 0.7.7 (pbs 20250521)
* ❌ uv 0.7.6 (pbs 20250517)
* ✅ uv 0.7.5 (pbs 20250409)
* os
* ❌ linux
* ✅ windows
* arch
* ❌ x86_64
* ✅ aarch64
* python version
* ❌ 3.12
* ❌ 3.13
* ❌ 3.14
Fixes#13610
Stubs packages are different in that their name ends with `-stubs`,
their module is `<module name>-stubs` (with a dash, not the generally
legal underscore) and their modules contain a `__init__.pyi` instead of
an `__init__.py`
(https://typing.python.org/en/latest/spec/distributing.html#stub-only-packages).
We add support in the uv build backend by detecting the `-stubs` suffix.
Fixes#13546
---------
Co-authored-by: Alex Waygood <Alex.Waygood@Gmail.com>
This PR contains the following updates:
| Package | Type | Update | Change |
|---|---|---|---|
| [markdown](https://redirect.github.com/wooorm/markdown-rs) |
dependencies | major | `0.3.0` -> `1.0.0` |
---
### Release Notes
<details>
<summary>wooorm/markdown-rs (markdown)</summary>
###
[`v1.0.0`](https://redirect.github.com/wooorm/markdown-rs/releases/tag/1.0.0)
💯
Nothing changed since the last alpha.
It’s just that: this crate’s now being used a bunch and working well, so
it’s time to be stable!
</details>
---
### Configuration
📅 **Schedule**: Branch creation - "before 4am on Monday" (UTC),
Automerge - At any time (no schedule defined).
🚦 **Automerge**: Disabled by config. Please merge this manually once you
are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the
rebase/retry checkbox.
🔕 **Ignore**: Close this PR and you won't be reminded about this update
again.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check
this box
---
This PR was generated by [Mend Renovate](https://mend.io/renovate/).
View the [repository job
log](https://developer.mend.io/github/astral-sh/uv).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiIzOS4yNTcuMyIsInVwZGF0ZWRJblZlciI6IjM5LjI1Ny4zIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6WyJpbnRlcm5hbCJdfQ==-->
---------
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
Co-authored-by: konstin <konstin@mailbox.org>
## Summary
Fix the two version json tests to account for the possibility that uv
was built outside a git checkout (e.g. from an unpacked git archive) and
therefore does not have the commit info available. This approach uses
separate snapshots for the two cases, as suggested in discussion of pull
request #13251.
Fixes#13212
## Test Plan
1. `cargo test` in a git clone.
2. `cargo clean`, moved `.git` away, `cargo test` again.
---------
Co-authored-by: Aria Desires <aria.desires@gmail.com>
This adopts the logic from `uv remove` for locking and syncing, as the
scope of the changes made are ultimately similar. Unlike `uv remove`
there is no support for modifying PEP723 scripts, as these are not
versioned.
In doing this the `version` command gains a truckload of args for
configuring lock/sync behaviour. Presumably most of these are passed via
settings or env files, and not of particular concern.
The most interesting additions are:
* `--frozen`: makes `uv version` work ~exactly as it did before this PR
* `--locked`: errors if the lockfile is out of date
* `--no-sync`: updates the lockfile, but doesn't run the equivalent of
`uv sync`
* `--package name`: a convenience for referring to a package in the
workspace
Note that the existing `--dry-run` flag effectively implies `--frozen` for sets and bumps.
Fixes#13254Fixes#13548
In platform discovery we're parsing the output of the ELF interpreter,
e.g., `/lib64/ld-linux-x86-64.so.2`. This file is ld, not ldd, which was
incorrectly named in the code.
An alternative is naming everything ELF interpreter instead of ld.so.
## Summary
This test started failing, and it fails at least back to v0.6, so I
don't think it's on our end. I'm wondering if all the wheels here were
yanked? They're visible in the lockfile, but not on PyPI:
https://pypi.org/project/av/9.2.0/#files. So to get this passing, let's
just unpin it.
Edit: Ahh, ok. It looks like the project ran out of space, so they
removed wheels for all the older versions:
https://github.com/PyAV-Org/PyAV/issues/1879.
## Summary
Without the `git` feature, it fails with:
```
error: Failed to initialize Git repository at `/home/mgorny/.local/share/uv/tests/.tmp01wGGK/temp/preserve_executable_bit`
stdout:
stderr: error: `git` operations are not allowed — are you missing a cfg for the `git` feature?
```
## Test Plan
cargo test --features python --profile=fast-build --no-default-features
## Summary
Using "all extras" in `uv remove` will cause errors for projects with
conflicting extras. Now that we have a concept of "default extras", it
seems better to respect those defaults like we do for dependency groups.
Closes https://github.com/astral-sh/uv/issues/12770.
## Summary
<!-- What's the purpose of the change? What does it do, and why? -->
I've compared all the activator scripts here with the original ones in
https://github.com/pypa/virtualenv/tree/main/src/virtualenv/activation
and only the bash/POSIX script here was yielding a VIRTUAL_ENV_PROMPT
value with parenthesis and a trailing space, which should be part of the
shell prompt (PS1 for bash/POSIX) but not of the VIRTUAL_ENV_PROMPT
value itself. This fixes that small inconsistency. Fixes#13456
This reverts commit 0ec2d4e434
## Test Plan
<!-- How was it tested? -->
I didn't test this locally.
Rustfmt introduces a lot of formatting changes in the 2024 edition. To
not break everything all at once, we split out the set of formatting
changes compatible with both the 2021 and 2024 edition by first
formatting with the 2024 style, and then again with the currently used
2021 style.
Notable changes are the formatting of derive macro attributes and lines
with overly long strings and adding trailing semicolons after statements
consistently.
We may run on case-sensitive file systems (Linux, generally) or on
case-insensitive file systems (Windows, generally), while modules in
Python may be lower or upper case. For robustness over filesystem
casing, we require an explicit module name for modules with upper cases.
Fixes#13419
Love this tooling! Small adjustment to help on error messaging 🙏
<!--
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 previous message was missing the word **“like,”** which made it read
a tad awkwardly.
This PR inserts the missing word so the error reads naturally:
**After:**
```
$ uvx ch1.py
error: It looks like you tried to run a Python script at `ch1.py`, which is not supported by `uvx`
hint: Use `uv run ch1.py` instead
```
**Before:**
```
$ uvx ch1.py
error: It looks you tried to run a Python script at `ch1.py`, which is not supported by `uvx`
hint: Use `uv run ch1.py` instead
```
## Test Plan
<!-- How was it tested? -->
- `cargo run -- uvx examples/ch1.py` shows the updated message (see
“After” above).
- `cargo test` passes.
Unlike OS APIs, glob inclusion checks don't work when there are relative
path elements such as `./`. We normalize the path before using it for
the glob.
Fixes#13407
We were not correctly falling back to cached realm credentials when an
index URL was provided with only a username. This came up in a [later
comment](https://github.com/astral-sh/uv/issues/13443#issuecomment-2881115301)
on #13443 where credentials in a pip extra index in `uv.toml` were being
ignored when the same URL (but with only a username) was used at the
command line for `--extra-index-url`. I've added a test to catch this
case.
Closes#13443
## Summary
Related to https://github.com/astral-sh/uv/issues/12492
This change makes all progress bars vertically aligned. This is still a
WIP and so is not complete, in the current design I store `max_len` in
`BarState` and update it on every `on_request_start`, however this is
problematic since order matters, and if the largest name is not sent
first, the alignment is not complete. To mitigate this we'd probably
have to update all previous bars by "iterating" through the `bars` field
in `BarState` and update all request bars.
Below is an image of what happens when the largest name
(`nvidia-cusparselt-cu12`) is not the first (in this case, it was the
second to last).
## Test Plan
There are currently no tests, and I'm not sure how to design them since
from what I gather the `uv_snapshot` facilities record the final output,
not the intermediate stages.
---------
Co-authored-by: konstin <konstin@mailbox.org>
This PR redacts credentials in displayed URLs.
It mostly relies on a `redacted_url` function (and where possible
`IndexUrl::redacted`). This is a quick way to prevent leaked credentials
but it's prone to programmer error when adding new trace statements. A
better follow-on would use a `RedactedUrl` type with the appropriate
`Display` implementation. This would allow us to still extract
credentials from the URL while displaying it securely. On the plus side,
the sites where the `redacted_url` function are used serve as easy
signposts for where to use the new type in a future PR.
Closes#1714.
## Summary
We mapped both `.tgz` and `.tar.gz` to the same enum variant; later,
though, we made the assumption that a file marked with that variant
ended with exactly `.tar.gz`. Instead, we need to preserve the
originating suffix.
Closes https://github.com/astral-sh/uv/issues/13372.
## Summary
If a set of wheel tags includes a dot, this code is treating the part
_after_ the dot as an extension, and thereby failing to detect that the
entry is a symlink to an archive (and thereby removing the archive).
This is all an optimization, so this code just makes it a little
targeted: we skip specific known extensions, rather than anything with
any extension.
Closes https://github.com/astral-sh/uv/issues/13270.
This follows on from #13334 to fix another case.
<!--
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
If a dependency group contained any `{ include-group = "..." }` entries,
the sort detection would bail out. The root cause of the problem was
gating the sort detection behind `deps.iter().all(Value::is_str)`.
A public code search reveals that keeping include-groups at the top is
by far the most common, but keeping them at the bottom isn't uncommon.
In both of these cases, uv will now preserve the convention that is in
use.
Unless I've missed it, I don't think uv supports `uv add`ing an
include-group, and so that wasn't tested here.
## Test Plan
cargo test
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
This also omits bounds on constraints, and is useful for that. This
retains `--raw-sources` as an alias. I've had this on my mind for a
while, but https://github.com/astral-sh/uv/pull/12946 reminded me of it
again.
## Summary
Closes#13253
## Test Plan
```sh
❯ cat pyproject.toml | rg required
required-version = ">=0.7.3, <0.8"
❯ cargo run -q --features self-update --manifest-path ~/uv/Cargo.toml add black
error: Required uv version `>=0.7.3, <0.8` does not match the running version `0.7.2`.
hint: Update `uv` by running `uv self update`.
❯ cat pyproject.toml | rg required
required-version = ">=0.7.3"
❯ cargo run -q --features self-update --manifest-path ~/uv/Cargo.toml add black
error: Required uv version `>=0.7.3` does not match the running version `0.7.2`.
hint: Update `uv` by running `uv self update`.
❯ cat pyproject.toml | rg required
required-version = "<0.7"
❯ cargo run -q --features self-update --manifest-path ~/uv/Cargo.toml add black
error: Required uv version `<0.7` does not match the running version `0.7.2`.
❯ cat pyproject.toml | rg required
required-version = ">=0.4,<0.7"
❯ cargo run -q --features self-update --manifest-path ~/uv/Cargo.toml add black
error: Required uv version `>=0.4, <0.7` does not match the running version `0.7.2`.
```
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
The prior implementation only looks for dependencies which are sorted by
name then specifier.
I knew uv was meant to preserve sorted dependencies, but it never seemed
to work for me.
I've always used the "sort lines" feature of PyCharm/Sublime to sort
these lists, and I guess I'm not the only one. In such a case,
`flask-wtf>=1.2.1` is sorted before `flask>=3.0.2`.
After digging into the code I realised what was happening, hence this
merge request.
Maybe there's a tool I'm not aware of that people are using to sort
dependencies "properly", or are doing it by hand, but I think this is
worth supporting.
Relevant issues: https://github.com/astral-sh/uv/issues/9076,
https://github.com/astral-sh/uv/issues/10738
## Test Plan
`cargo test`
e.g., these are misleading cruft in the error message at
https://github.com/astral-sh/uv/pull/12168#discussion_r2078204601
```
❯ uv python find /foo/bar
error: No interpreter found for path `/foo/bar` in virtual environments, managed installations, or search path
❯ cargo run -q -- python find /foo/bar
error: No interpreter found at path `/foo/bar`
```
When removing a Python interpreter underneath an existing venv, uv
currently shows a not found error:
```
error: Failed to inspect Python interpreter from active virtual environment at `.venv/bin/python3`
Caused by: Python interpreter not found at `/home/konsti/projects/uv/.venv/bin/python3`
```
This is unintuitive, as the file for the Python interpreter does exist,
it is a broken symlink that needs to be replaced with `uv venv`.
I've been encountering those occasionally, and I expect users that
switch between versions a lot will, too, especially when they also use
pyenv or a similar Python manager.
The new error hints at this solution:
```
error: Failed to inspect Python interpreter from active virtual environment at `.venv/bin/python3`
Caused by: Broken symlink at `.venv/bin/python3`, was the underlying Python interpreter removed?
hint: To recreate the virtual environment, run `uv venv`
```
PEP 639 does not allow any characters that aren't in either their
limited glob syntax or the alphanumeric Unicode characters. This means
there's no way to express a glob such as `**/@test` for the excludes.
We extend the glob syntax from PEP 639 by introducing backslash escapes,
which can escape all characters but path separators (forward and
backwards slashes) to be parsed verbatim.
This means we have two glob parsers: The strict PEP 639 parser for
`project.license-files`, and our extended parser for `tool.uv`, with a
slight difference if you need to use special characters, to both adhere
to PEP 639 and to support cases such as #13280.
Fixes#13280
We have test coverage for this elsewhere, but managed Python versions
are a distinct case because we know the _full_ version before querying
the interpreter (whereas, when we find them on the `PATH`, we usually
only know `X.y` from the file name).
This pre-filter logic now matches our subsequent logic at
060be9cef1/crates/uv-python/src/discovery.rs (L2146-L2149)060be9cef1
shows the snapshot change.
## Summary
Add a `--show-extras` argument to the `uv tool list` cli, to show which
extra dependencies were installed with the tool.
i.e.
```bash
$ uv tool install fastapi --with requests --with typer==0.14
```
```bash
$ uv tool list --show-extras
fastapi v0.115.12 [extras: requests, typer==0.14]
- fastapi
```
## Test Plan
Added a new test function based on the others in the same file, with the
other arguments tested with the new argument as well.
The goal of this PR is to support reproducible builds and best-effort
platform-independent builds. Previously, while the build backend would
build the same source dist and wheel on the same machine, they would
look different across different operating systems. This PR fixes the
platform-dependent walk dir order by sorting and removes
platform-specific permissions from the source dist that had caused those
differences.
The reproducibility goal does not extend to platform-dependent
filesystem features, such as permissions and links, especially in
interaction with Git. Since most users share code across platforms
through Git, we're focusing on cross-platform behavior under Git. One of
those caveats is intentional: If a file, such as a bash script, has an
executable bit, we preserve it. This means that E.g. builds of Git
checkout of a repository with an executable shell script in the sources
will have different archives on Unix and Windows. Another relevant case
are symlinks: By default, Git on Windows replaces symlinks with a file
that contains the path to the target file
(https://stackoverflow.com/q/5917249/3549270). (This example comes from
Cargo, where it means that the package archive is different on Windows
when symlinking license from the repository root to a workspace package)
Best reviewed commit-by-commit
## Summary
This adds GraalPy download metadata so that `uv python install graalpy`
works. See https://github.com/astral-sh/uv/issues/13114
## Test Plan
The existing integration test was changed to test this functionality.
In #13302, there was an IO error without context. This error seems to be
caused by a symlink error. Switching as symlinking to `fs_err` ensures
these errors will carry context in the future.
## Summary
We now show a user-visible warning if we're using a "stale" virtual
environment due to `--no-sync`. I'd also be fine erroring here.
Closes https://github.com/astral-sh/uv/issues/13235.
Add configuration documentation for the build backend and make it the
preview default.
The build backend should generally work with default configuration
unless you want specific features such as flat layout or module
renaming, there is only a dedicated configuration, but no concept or
guide page for the build backend. Once the build backend is stable, we
can update the guide documentation to explain that uv defaults to its
own build backend, but other build backends are also supported.
The uv build backend becomes the default in preview, giving it more
exposure from users and preparing it to make it the default proper. The
current documentation retains warnings that the build backend is in
preview.
To see current uses of `uv_build` on GitHub:
https://github.com/search?q=path%3A**%2Fpyproject.toml+uv_build%3E%3D0&type=code
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
## Summary
If a script has some requirements, and you provide `--with`, we
currently ignore any constraints from those requirements. We might want
to treat them as hard constraints in the future. For now, though, we
just treat them as preferences -- so we _prefer_ those versions, but
don't require them to match and still run the `--with` resolution in
isolation.
Closes https://github.com/astral-sh/uv/issues/13173.
See commentary at
https://github.com/astral-sh/uv/issues/9828#issuecomment-2537542100
regarding the limitations and future upstream changes needed.
```
❯ cargo build --features self-update
Compiling uv v0.5.8 (/Users/zb/workspace/uv/crates/uv)
Finished `dev` profile [unoptimized + debuginfo] target(s) in 7.28s
❯ cp ./target/debug/uv ~/.cargo/bin
❯ uv self update --dry-run
info: Checking for updates...
Nothing to do. You're on the latest version of uv (v0.5.8)
❯ uv self update --dry-run 0.5.7
info: Checking for updates...
Would update uv from v0.5.8 to v0.5.7
❯ vi ~/.config/uv/uv-receipt.json # Edit the receipt to think its on an older version
❯ uv self update --dry-run
info: Checking for updates...
Would update uv from v0.5.8 to the latest version
```
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
Part of https://github.com/astral-sh/uv/issues/13173, but doesn't close
the issue. This just respects preferences if your script uses a
lockfile, since we already support that for locked _projects_.
#5577 fixed a bug on macos due to dynamically linking lzma/xz through
static linking. In #7686, this feature was moved to the performance
category.
This PR moves the `xz2/static` back to the general default features,
and, inspired by https://github.com/Homebrew/homebrew-core/pull/222211,
it structures and documents the feature flags cleaner.
We need to take care that this feature does not accidentally disable
features we want.
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
## Summary
The only thing that changed for #12175 relevant to the existing
downloads is the order of nesting, so we're checking all nested IO
errors instead of only the first one.
See #13238
## Test Plan
This is an educated guess based on what happens if I turn off the
network during a download.
```
Downloading cpython-3.13.3-linux-x86_64-gnu (download) (20.3MiB)
TRACE Considering retry of error: ExtractError("cpython-3.13.3-20250409-x86_64-unknown-linux-gnu-install_only_stripped.tar.gz", Io(Custom { kind: Other, error: TarError { desc: "failed to unpack `/home/konsti/.local/share/uv/python/.temp/.tmpe3AIvt/python/lib/libpython3.13.so.1.0`", io: Custom { kind: Other, error: TarError { desc: "failed to unpack `python/lib/libpython3.13.so.1.0` into `/home/konsti/.local/share/uv/python/.temp/.tmpe3AIvt/python/lib/libpython3.13.so.1.0`", io: Custom { kind: Other, error: reqwest::Error { kind: Decode, source: reqwest::Error { kind: Body, source: TimedOut } } } } } } }))
TRACE Cannot retry IO error: not one of `ConnectionReset` or `UnexpectedEof`
TRACE Cannot retry IO error: not one of `ConnectionReset` or `UnexpectedEof`
TRACE Cannot retry error: not an IO error
error: Failed to install cpython-3.13.3-linux-x86_64-gnu
Caused by: Failed to extract archive: cpython-3.13.3-20250409-x86_64-unknown-linux-gnu-install_only_stripped.tar.gz
Caused by: failed to unpack `/home/konsti/.local/share/uv/python/.temp/.tmpe3AIvt/python/lib/libpython3.13.so.1.0`
Caused by: failed to unpack `python/lib/libpython3.13.so.1.0` into `/home/konsti/.local/share/uv/python/.temp/.tmpe3AIvt/python/lib/libpython3.13.so.1.0`
Caused by: error decoding response body
Caused by: request or response body error
Caused by: operation timed out
```
<!--
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? -->
Fixes#11970.
## Test Plan
<!-- How was it tested? -->
Ran `cargo nextest`
There is a new idna version of testpypi. While we don't select that
version due to the exclude-newer cutoff, the version is still available
to pubgrub as an incompatible choice, changing the error message on
conflicts.
## Summary
Part of #12838. Allow users to configure `python-downloads-json-url` in
`uv.toml` and not just from env.
I followed similar PR #8695, so same as there it's also available in the
CLI (I think maybe it's better not to be configurable from the CLI, but
since the mirror parameters are, I think it's better to do the same)
## Test Plan
<!-- How was it tested? -->
## Summary
In #10939 I added the generated
`crates/uv-python/src/download-metadata-minified.json` file which is a
minified version of `crates/uv-python/download-metadata.json`.
The main reason for this PR is to avoid bloating the git objects as this
is a single-line file.
As a bonus, I also filtered the embed json to include only the versions
for the compiled target. Which should improve the binary size and
performance by a bit.
## Test Plan
<!-- How was it tested? -->
Previously, we were using the wrong `Workspace` discovery and would
report the version of the workspace root, which would iterate up from
the `--project` directory and return the workspace root (with or without
a project in the root). Instead, we need `ProjectWorkspace` discovery
that returns the closest project.
This fixes `uv version --project <path>` where `<path>` belongs to a
workspace member.
Fixes#13213
I think this regressed in https://github.com/astral-sh/uv/pull/13027 — I
misunderstood what versions could be represented in the `pyvenv.cfg` (I
assumed they _never_ included pre-release components).
Closes#13233
uv’s default index strategy was designed with dependency confusion
attacks in mind. [According to the
docs](https://docs.astral.sh/uv/configuration/indexes/#searching-across-multiple-indexes),
“if a package exists on an internal index, it should always be installed
from the internal index, and never from PyPI”. Unfortunately, this is
not true in the case where authentication fails on that internal index.
In that case, uv will simply try the next index (even on the
`first-index` strategy). This means that uv is not secure by default in
this common scenario.
This PR causes uv to stop searching for a package if it encounters an
authentication failure at an index. It is possible to opt out of this
behavior for an index with a new `pyproject.toml` option
`ignore-error-codes`. For example:
```
[[tool.uv.index]]
name = "my-index"
url = "<index-url>"
ignore-error-codes = [401, 403]
```
This will also enable users to handle idiosyncratic registries in a more
fine-grained way. For example, PyTorch registries return a 403 when a
package is not found. In this PR, we special-case PyTorch registries to
ignore 403s, but users can use `ignore-error-codes` to handle similar
behaviors if they encounter them on internal registries.
Depends on #12651Closes#9429Closes#12362
If you pass a TOML file to `uv pip install` that isn't recognized, we
should just reject it instead of assuming `requirements.txt`. I just
don't see a real case where it's better to let the command proceed.
This is a reimplementation of #7248 with a new CLI interface.
The old `uv version` is now `uv self version` (also it has gained a
`--short` flag for parity).
The new `uv version` is now an interface for getting/setting the project
version.
To give a modicum of support for migration, if `uv version` is run and
we fail to find/read a `pyproject.toml` we will fallback to `uv self
version`. `uv version --project .` prevents this fallback from being
allowed.
The new API of `uv version` is as follows:
* pass nothing to read the project version
* pass a version to set the project version
* `--bump major|minor|patch` to semver-bump the project version
* `--dry-run` to show the result but not apply it
* `--short` to have the final printout contain only the final version
* `--output-format json` to get the final printout as json
```
$ uv version
myfast 0.1.0
$ uv version --bump major --dry-run
myfast 0.1.0 => 1.0.0
$ uv version 1.2.3 --dry-run
myfast 0.1.0 => 1.2.3
$ uv version 1.2.3
myfast 0.1.0 => 1.2.3
$ uv version --short
1.2.3
$ uv version --output-format json
{
"package_name": "myfast",
"version": "1.2.3",
"commit_info": null
}
```
Fixes#6298
Some registries (like Azure Artifact) can require you to authenticate
separately for every package URL if you do not authenticate for the
/simple endpoint. These changes make the auth middleware aware of index
URL endpoints and attempts to fetch keyring credentials for such an
index URL when making a request to any URL it's a prefix of.
The current uv behavior is to cache credentials either at the request
URL or realm level. But with these changes, we also need to cache
credentials at the index level. Note that when uv does not detect an
index URL for a request URL, it will continue to apply the old behavior.
Addresses part of #4056Closes#4583Closes#11236Closes#11391Closes#11507
Closes https://github.com/astral-sh/uv/issues/7804
Includes a few small minor changes to the messaging, but the primary
change is that in, e.g., `uvx foo`, if the `foo` package does not
provide the `foo` executable we will no longer execute an arbitrary
`foo` executable if present on the `PATH`. This prevents confusing and
surprising behavior, such as the user reported where they did `uv tool
install foobar` (which provides `foo`) then `uvx foo` (which does not
provide `foo`) later falls back to the executable provided by `foobar`
since it's on the `PATH`. We don't enforce this for `--from`, so things
like `uvx --from foo bash -c "..."` are still totally valid. We also
still allow `uvx foo` where the `foo` executable is provided by a
_dependency_ of `foo` instead of `foo` itself.
Most of the diff here is consolidating the logic of the
`hint_on_not_found` and `warn_executable_not_provided_by_package `
utilities.
## Summary
If you use `--torch-backend=auto`, we want to avoid selecting (e.g.) a
`+cu124` build of `torch` alongside a `+cu126` build of `torchvision`.
<!--
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? -->
This adds `poetry-core` as a build backend choice.
## Test Plan
<!-- How was it tested? -->
---------
Co-authored-by: konstin <konstin@mailbox.org>
## Summary
Part of #8607. This is a pure refactor aimed at paving the way for
supporting the `default-extras` configuration in the `pyproject.toml`
file.
The `ExtraSpecification` struct has been refactored to align more
closely with the
[`DependencyGroups`](256b100a9e/crates/uv-configuration/src/dependency_groups.rs (L9))
struct.
## Test Plan
Existing tests.
## Summary
In https://github.com/astral-sh/uv/pull/12968, we added support for
upload time to `uv.lock`, but stylized as `upload_time`. The other keys
in `uv.lock` use kebab casing, as in common in Python formats, so this
really should've been `upload-time`. I want to change it ASAP to
minimize churn for users. Any users that already upgraded will of course
experience churn in their files a second time. But if we don't change it
now, we'll only increase the surface area of affected users.
So, this PR uses `upload-time` instead, but continues reading
`upload_time` to make it non-breaking.
Part of #11834
Currently, all Python installation are a streaming download-and-extract.
With this PR, we add the `UV_PYTHON_CACHE_DIR` variable. When set, the
installation is split into downloading the interpreter into
`UV_PYTHON_CACHE_DIR` and extracting it there from a second step. If the
archive is already present in `UV_PYTHON_CACHE_DIR`, we skip the
download.
The feature can be used to speed up tests and CI. Locally for me, `cargo
test -p uv -- python_install` goes from 43s to 7s (1,7s in release mode)
when setting `UV_PYTHON_CACHE_DIR`. It can also be used for offline
installation of Python interpreter, by copying the archives to a
directory in the offline machine, while the path rewriting is still
performed on the target machine on installation.
This PR restores #13041 and integrates two PRs from @zanieb:
* #13038
* #13040
It also adds tests for relative URI and fragment handling.
Closes#13037.
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
Just a small PR to add mentions to `pylock.toml` in the CLI manual where
appropriate.
I tried to say "PEP-751 compatible lock files" when appropriate to also
include the case `r"^pylock\.([^.]+)\.toml$"`. Feel free to change that
if you think it's cluttery.
I also tried to include the "single-use" wording when it made sense.
I also have almost never used the `uv pip` interface, so maybe there are
some other minor things to add here and there about the usage of
`pylock.toml` that I missed.
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
When working on #13025 I noticed this message was lacking versions,
which seems frustrating if you're debugging things.
I refactored the general `matches_interpreter` utilities that were added
in https://github.com/astral-sh/uv/pull/12884 into a more purpose-fit
function that returns an `Option` with the versions if there's a
mismatch.
This PR restores the `bogus_redirect` test that was
non-deterministically hanging (reverting #13076).
Mismatched package and distribution names were causing uv to hang prior
to #12917 (which added the `bogus_redirect` test). But with that fix, uv
was only checking for mismatched package names on the main thread (and
not the resolver thread). This allowed for a race condition which would
prevent uv from ever doing the check, triggering the original hang
condition. This PR adds the check to the resolver thread to prevent this
race condition.
When an index performs a bogus redirect or otherwise returns a different
distribution name than expected, uv currently hangs.
In the example case, requesting the simple index page for any package
returns the page for anyio. This mean querying the sniffio version map
returns only anyio entries, and the version maps resolves to an anyio
version. When the resolver makes a query for sniffio and waits for it to
resolve, the main thread finds an anyio and resolves only that in the
wait map, causing the hang.
We fix this by checking the name of the returned distribution against
the name of the requested distribution. For good measure, we add the
same check in `Request::Dist` and `Request::Installed`. For performance
and complexity reasons, we don't perform this check in the version map
itself, but only after a candidate distribution has been selected.
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
## Summary
Fixes several occurrences of the minor typo “This options” for “This
option.”
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
<!-- How was it tested? -->
Since this is just a typo fix in documentation and comment strings, no
particular testing was conducted.
## Notes
The typo fixes in `crates/uv-cli/src/lib.rs` would affect
`docs/reference/cli.md`. I assumed you might want to just re-generate
the reference documention, but fixing it up manually would look like:
```diff
diff --git a/docs/reference/cli.md b/docs/reference/cli.md
index 338fa0ff9..8851ca2c0 100644
--- a/docs/reference/cli.md
+++ b/docs/reference/cli.md
@@ -355,7 +355,7 @@ uv run [OPTIONS] [COMMAND]
</dd><dt id="uv-run--no-group"><a href="#uv-run--no-group"><code>--no-group</code></a> <i>no-group</i></dt><dd><p>Disable the specified dependency group.</p>
-<p>This options always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
+<p>This option always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
<p>May be provided multiple times.</p>
@@ -1757,7 +1757,7 @@ uv sync [OPTIONS]
</dd><dt id="uv-sync--no-group"><a href="#uv-sync--no-group"><code>--no-group</code></a> <i>no-group</i></dt><dd><p>Disable the specified dependency group.</p>
-<p>This options always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
+<p>This option always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
<p>May be provided multiple times.</p>
@@ -2492,7 +2492,7 @@ uv export [OPTIONS]
</dd><dt id="uv-export--no-group"><a href="#uv-export--no-group"><code>--no-group</code></a> <i>no-group</i></dt><dd><p>Disable the specified dependency group.</p>
-<p>This options always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
+<p>This option always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
<p>May be provided multiple times.</p>
@@ -2855,7 +2855,7 @@ uv tree [OPTIONS]
</dd><dt id="uv-tree--no-group"><a href="#uv-tree--no-group"><code>--no-group</code></a> <i>no-group</i></dt><dd><p>Disable the specified dependency group.</p>
-<p>This options always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
+<p>This option always takes precedence over default groups, <code>--all-groups</code>, and <code>--group</code>.</p>
<p>May be provided multiple times.</p>
```
<!--
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
Provide an in-code switch to permit using the `uv build-backend` command
rather than the default `uv-build` in the Python PEP517 backend. This
option is intended to be used by downstream packagers to provide an
option of reusing `uv` that was built already instead of having to build
a second `uv-build` executable that largely overlaps with `uv`.
Fixes#12389
## Test Plan
The option is intended for downstream consumption only, and it is tested
downstream (via attempting to build a package using the `uv_build`
backend). The backend itself is covered by tests already.
---------
Co-authored-by: konstin <konstin@mailbox.org>
## Summary
We accept `pylock.toml` as a requirements file (e.g., `uv sync
pylock.toml` or `uv pip install -r pylock.toml`). When you provide a
`pylock.toml` file, we don't allow you to provide other requirements, or
constraints, etc. And you can only provide one `pylock.toml` file, not
multiple.
We might want to remove this from `uv pip install` for now, since `pip`
may end up with a different interface (whereas `uv pip sync` is already
specific to uv), and most of the arguments aren't applicable (like
`--resolution`, etc.). Regardless, it's behind `--preview` for both
commands.
## Summary
This PR adds `uv export` support for [PEP
751](https://peps.python.org/pep-0751). We don't yet expose a way to
consume the generated lockfile, but it's a first step.
The logic to go from `uv.lock` to "flat set of packages to include, with
markers telling us when to include them" is all shared with the
`requirements.txt` export (and extracted in
https://github.com/astral-sh/uv/pull/12956). So most of the code is just
converting from our internal types to the PEP 751 schema.
As I suspected quite some time ago
(https://github.com/astral-sh/uv/pull/6738#issuecomment-2315466033),
it's problematic that we don't handle _every_ signal here. This PR adds
handling for all of the Unix signals except `SIGCHLD`, `SIGIO`, and
`SIGPOLL` which seem incorrect to forward. Also notable, we _cannot_
handle `SIGKILL` so if someone sends that to the PID instead of the
PGID, they will leave dangling subprocesses.
Instead, we could use `exec` and avoid this handling. However, we'd lose
the ability to add nice error message on failure (e.g., as someone is
trying to add in https://github.com/astral-sh/uv/pull/12201) and, more
critically, we'd need to figure out how to clean up resources properly
(i.e., temporary directories) which currently happens on `Drop`. In the
long-term, we'll probably want an option to use `exec` — but we'll need
to figure out when to clean up resources or accept that they will
dangle. This was last discussed in
https://github.com/astral-sh/uv/issues/3095 — discussion on that
approach should continue there.
A note on the implementation: I spent time time trying to write the
handler using a tokio stream, so we could dynamically iterate over a
list of signals instead of copy/pasting the implementation — I couldn't
get it to work though and it didn't seem critical.
Closes https://github.com/astral-sh/uv/issues/12830
## Summary
Before:
```console
$ uv python list py --managed-python
error: Interpreter discovery for `executable name `py`` requires `search path` but only only managed is allowed
```
After:
```console
$ uv python list py --managed-python
error: Interpreter discovery for `executable name `py`` requires `search path` but only `only managed` is allowed
```
Closes: #4567
## Summary
When adding a package with Git reference options (`--rev`, `--tag`,
`--branch`) that already has a Git source defined, use the existing Git
URL with the new reference instead of reporting an error.
This allows commands like `uv add requests --branch main` to work when
requests is already defined with a Git source in the project
configuration.
Previously, you would need to provide the whole Git url again for this
to work:
```bash
uv add git+https://github.com/psf/requests --branch main
```
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
- [x] Add unit tests for project
- [x] Add unit tests for script
- [x] Tested locally for project and script environments like below
### Testing Project
In a directory using the `uv` executable from this PR (via replacing
every `uv` with `cargo run --`) initialize a project and virtual
environment
```bash
uv init
uv venv
```
move into the environment
```bash
# on mac
source .venv/bin/activate
```
and add a dependency with a git url
```bash
uv add git+https://github.com/Textualize/rich --branch master
```
Then change the branch of the project to see that the branch can be
changed without need of the whole git url:
```bash
uv add rich --branch py310
```
### Testing Script
Create the following file, e.g. `script.py`:
```python
import time
from rich.progress import track
print("Starting")
for i in track(range(20), description="For example:"):
time.sleep(0.05)
print("Done")
```
Now using `uv` (referencing the executable of this PR) add the
dependency
```bash
uv add --script script.py 'git+https://github.com/Textualize/rich' --branch master
```
and check we can execute the script:
```bash
uv run script.py
```
To test the change update the branch
```bash
uv add --script script.py rich --branch py310
```
and check that the dependency is updated and the script is executed:
```bash
uv run script.py
```
<!-- How was it tested? -->
----
This is my first time contributing to `uv` (yay, 🤗) so let me know if
there is something obvious i am missing.
Unit tests will follow soon.
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
This is included in PEP 751, so we lose it when converting from
`uv.lock`. I think it's a good piece of information to include in the
`uv.lock` anyway.
I noticed in the trace output that we weren't obfuscating the
`Credentials` password in a trace message. This PR creates a `Password`
newtype with a custom `Debug` implementation.
uv was failing to authenticate on 302 redirects when credentials were
available. This was because it was relying on `reqwest_middleware`'s
default redirect behavior which bypasses the middleware pipeline when
trying the redirect request (and hence bypasses our authentication
middleware). This PR updates uv to retrigger the middleware pipeline
when handling a 302 redirect, correctly using credentials from the URL,
the keyring, or `.netrc`.
Closes#5595Closes#11097
Fixes#12914.
When `PythonDownloadRequest` does not have the `implementation` set, do
not set it to CPython when calling `fill`, otherwise only CPython
interpreters are shown when listing interpreters available for download,
with `uv python list`.
## Summary
This PR simplifies the version formatting by replacing `.white()` with
`.cyan()` styling for consistency.
Resolves#12940
## Test Plan
I manually recreated the code and tested it with this patch:
```diff
diff --git i/crates/uv/src/lib.rs w/crates/uv/src/lib.rs
index b9c01b002..cf051351f 100644
--- i/crates/uv/src/lib.rs
+++ w/crates/uv/src/lib.rs
@@ -1019,6 +1019,20 @@ async fn run(mut cli: Cli) -> Result<ExitStatus> {
}) => commands::self_update(target_version, token, printer).await,
#[cfg(not(feature = "self-update"))]
Commands::Self_(_) => {
+ eprintln!("{}: {}", "error".cyan().bold(), "fake error message");
+
+ let version_information = format!(
+ "from {} to {}",
+ "v0.1.1".bold().cyan(),
+ "v0.1.2".bold().cyan(),
+ );
+ eprintln!(
+ "{}{} Upgraded uv {}! {}",
+ "success".green().bold(),
+ ":".bold(),
+ version_information,
+ format!("https://github.com/astral-sh/uv/releases/tag/{}", "v0.1.2").cyan()
+ );
anyhow::bail!(
"uv was installed through an external package manager, and self-update \
is not available. Please use your package manager to update uv."
```
In a light terminal, this is what it looks like:
<img width="750" alt="image"
src="https://github.com/user-attachments/assets/dc0d283c-e845-41fb-9821-80b0a3f1c4fe"
/>
Closes#12929
## Summary
Untag the `config-settings` value to support JSON schema according to
the
[docs](https://docs.astral.sh/uv/reference/settings/#config-settings).
```toml
[tool.uv]
config-settings = { editable_mode = "compat" }
```
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
Verified using the "Even Better TOML" extension with paths to old and
new `uv.schema.json`.
## Notes
I could not reproduce the issue with either the `taplo` (on which Even
Better Toml is built, afaik) and `check-jsonschema` CLI tools; with both
old and new versions of the `uv.schema.json` validated the
`pyproject.toml`.
Maybe for these there is some additional regularization going on and
that's also how a breaking case ended up in the docs?
I'm unsure on how to test for this.
After about an hour, the Even better TOML VSCode extension was the only
way to reproduce failing validation.
Let me know if I can do something else.
<!-- How was it tested? -->
Currently, `uv init` works without a `git` executable, and with a
working `git` executable, but not with a broken `git`, be it from GitHub
Action's Windows CI or from the shim we insert.
`uv init` calls git twice: Once `git rev-parse` to check whether a git
repo already exists, and then `git init` (if there is no git repository
yet and no `--vcs none`).
By separately handling the cases where git failed during `git rev-parse`
doesn't work vs. where the is no repository when checking for an
existing repo work tree, we can avoid calling `git init` for broken git
and erroring. We have to hardcode the expected git command outputs to be
able to check.
This is a rebased and updated version of #11925 based on my review (I
didn't have permission to push to their branch).
For posterity I've preserved their commits but my final commit
essentially rewrites the whole thing anyway.
Fixes#11637
---------
Co-authored-by: Chris Lieb <clieb@bitsighttech.com>
"Only show Python downloads, exclude installed distributions." might be
misunderstood as excluding installed distributions from `uv python list
--only-downloads`, implying that versions already installed won’t be
shown.
See #12769 for the motivation. We set the 4MB not only for the main
thread, but also for all tokio and rayon threads to fix a stack overflow
while unpacking wheels in production on Windows.
There are two variables for setting the stack size: A new
`UV_STACK_SIZE` that takes precedent, and the existing `RUST_MIN_STACK`.
When setting the stack size, `UV_STACK_SIZE` should be preferred, since
`RUST_MIN_STACK` affects all Rust applications, including build backends
we call (e.g., maturin). The minimum stack size is set to 1MB, the
lowest stack size we observed on a platform (Windows main thread).
Fixes#12769
## Test Plan
Tested manually with the example from #12769
## Summary
Closes#12855
This PR also fixed an issue, where `python_request` was matched against
`PythonVersion::Default`. Previously, if `python_request` was `3.13t`,
it would match the last branch, triggering a download of the Python
version if it wasn't already installed.
6b7f60c1ea/crates/uv/src/commands/project/init.rs (L421-L448)
```console
❯ uv init -v --managed-python --python 3.13t foo
DEBUG uv 0.6.14 (a4cec56dc 2025-04-09)
DEBUG Searching for Python 3.13t in managed installations
DEBUG Searching for managed installations at `/Users/Jo/.local/share/uv/python`
DEBUG Found managed installation `cpython-3.13.1-macos-aarch64-none`
DEBUG Found `cpython-3.13.1-macos-aarch64-none` at `/Users/Jo/.local/share/uv/python/cpython-3.13.1-macos-aarch64-none/bin/python3.13` (managed installations)
DEBUG Skipping interpreter at `/Users/Jo/.local/share/uv/python/cpython-3.13.1-macos-aarch64-none/bin/python3.13` from managed installations: does not satisfy request `3.13t`
DEBUG Skipping incompatible managed installation `cpython-3.12.8-macos-aarch64-none`
DEBUG Skipping incompatible managed installation `pypy-3.11.11-macos-aarch64-none`
DEBUG Requested Python not found, checking for available download...
DEBUG Acquired lock for `/Users/Jo/.local/share/uv/python`
DEBUG Using request timeout of 30s
INFO Fetching requested Python...
Downloading cpython-3.13.3+freethreaded-macos-aarch64-none (49.9MiB)
DEBUG Downloading https://github.com/astral-sh/python-build-standalone/releases/download/20250409/cpython-3.13.3%2B20250409-aarch64-apple-darwin-freethreaded%2Bpgo%2Blto-full.tar.zst to temporary location: /Users/Jo/.local/share/uv/python/.temp/.tmpfoOLkE
DEBUG Extracting cpython-3.13.3%2B20250409-aarch64-apple-darwin-freethreaded%2Bpgo%2Blto-full.tar.zst
Downloaded cpython-3.13.3+freethreaded-macos-aarch64-none
DEBUG Moving /Users/Jo/.local/share/uv/python/.temp/.tmpfoOLkE/python/install to /Users/Jo/.local/share/uv/python/cpython-3.13.3+freethreaded-macos-aarch64-none
DEBUG Released lock at `/Users/Jo/.local/share/uv/python/.lock`
DEBUG Writing Python versions to `/private/tmp/foo/.python-version`
Initialized project `foo` at `/private/tmp/foo`
❯ cat foo/.python-version
3.13
```
After this PR, uv will not try to download it:
```console
❯ uv python uninstall 3.13t
❯ cargo run -- init -v --managed-python --python 3.13t bar
DEBUG uv 0.6.14+15 (6b7f60c1e 2025-04-12)
DEBUG Writing Python versions to `/private/tmp/bar/.python-version`
Initialized project `bar` at `/private/tmp/bar`
❯ cat bar/.python_version
3.13t
```
It was possible that a virtual environment became out of sync with the
interpreter it pointed to (for example, if a symlink was changed to an
updated Python version). In such a case, `pyvenv.cfg` and
`activate_this.py` would no longer be correct. This PR detects when the
`version` (`venv` module) or `version_info` (uv and `virtualenv`) field
in `pyvenv.cfg` is out of sync with the interpreter. In such a case, uv
recreates the virtual environment.
Closes#12461
We have been claiming in our releases that we provide
archives/installers for uv-build, but we only upload it as a wheel to
pypi. This is because cargo-dist tries to be helpful and find all your
apps, but this scales poorly to large workspaces like ours, as stuff
like this slips in. So invert the default and make uv the only package
dist will see until we say otherwise.
See e.g. https://github.com/astral-sh/uv/releases/tag/0.6.14Fixes#12883
By default, unlike on CI, a Windows machine does not allow creating
symlinks, so we have to unix-gate tests that assume symlinks.
We can't install the transformers ecosystem test on Windows due to
missing torch, so it is also unix-gated.
Windows translates error messages, so we have to filter the "File not
found" message, since it can also be a "Datei nicht gefunden".
## Summary
Closes#12687.
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
<!-- How was it tested? -->
Added the corresponding integration tests for:
- `uv sync --dry-run --locked`
- [x] Preview lock changes
- [x] Errors if lockfile is out-of-date
- `uv sync --dry-run --frozen`
- [x] Preview lock changes
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
Collapse whitespace into a single space in python_list tests, in order
to make them agnostic of padding, and therefore pass both with Python
3.12.9 and Python 3.12.10.
Fixes#12799
## Test Plan
cargo test --features python --profile=fast-build --no-default-features
This PR contains the following updates:
| Package | Type | Update | Change |
|---|---|---|---|
| [mimalloc](https://redirect.github.com/purpleprotocol/mimalloc_rust) |
dependencies | patch | `0.1.45` -> `0.1.46` |
---
### Release Notes
<details>
<summary>purpleprotocol/mimalloc_rust (mimalloc)</summary>
###
[`v0.1.46`](https://redirect.github.com/purpleprotocol/mimalloc_rust/releases/tag/v0.1.46):
Version 0.1.46
[Compare
Source](https://redirect.github.com/purpleprotocol/mimalloc_rust/compare/v0.1.45...v0.1.46)
##### Changes
- Fixed musl builds.
</details>
---
### Configuration
📅 **Schedule**: Branch creation - "before 4am on Monday" (UTC),
Automerge - At any time (no schedule defined).
🚦 **Automerge**: Disabled by config. Please merge this manually once you
are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the
rebase/retry checkbox.
🔕 **Ignore**: Close this PR and you won't be reminded about this update
again.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check
this box
---
This PR was generated by [Mend Renovate](https://mend.io/renovate/).
View the [repository job
log](https://developer.mend.io/github/astral-sh/uv).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiIzOS4yMzguMCIsInVwZGF0ZWRJblZlciI6IjM5LjIzOC4wIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6WyJpbnRlcm5hbCJdfQ==-->
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
From PEP 440:
> The exclusive ordered comparison <V MUST NOT allow a pre-release of
the specified version unless the specified version is itself a
pre-release. Allowing pre-releases that are earlier than, but not equal
to a specific pre-release may be accomplished by using <V.rc1 or
similar.
We had an additional check that would block this even if the specifier
did have a pre-release.
This likely didn't show up earlier because `Ranges` uses different code
in the resolver.
I checked these changes against `packaging` to verify their behavior:
```python
print(SpecifierSet("<1").contains("1a1", prereleases=True)) # False
print(SpecifierSet("<1a2").contains("1a1", prereleases=True)) # True
print(SpecifierSet("<1").contains("1dev1", prereleases=True)) # False
print(SpecifierSet("<1dev2").contains("1dev1", prereleases=True)) # True
print(SpecifierSet("<1a2").contains("1dev1", prereleases=True)) # True
```
Closes#12834
## Summary
This PR errors out when an Unknown Dependency Object Specifier is used
in dependency groups.
Fixes#12638
## Test Plan
The current behaviour is as follows:
```bash
➜ example git:(12638/dependency-object-specifier) ✗ cargo run -- sync
Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.21s
Running `/home/luna/Documents/uv/target/debug/uv sync`
error: Failed to generate package metadata for `example==0.1.0 @ virtual+.`
Caused by: Group `bar` contains a Dependency Object Specifier, which is not supported by uv
```
And the pyproject.toml to produce this is:
```toml
[project]
name = "example"
version = "0.1.0"
description = "Add your description here"
readme = "README.md"
requires-python = ">=3.13.2"
dependencies = []
[dependency-groups]
foo = ["pyparsing"]
bar = [{set-phasers-to = "stun"}]
```
## Summary
Closes#12806
Split `UV_INDEX` by any whitespace rather than only ASCII 32, which does
not align with the behavior of `PIP_EXTRA_INDEX_URL` and can possibly
lead to difficulties when migrating from pip to uv.
Clap unfortunately does not support passing multiple delimiters, writing
a custom parsing function involved parsing index into a Vec<Vec<Index>>
and flattening it afterwards in order to avoid breaking the --index
command line option.
There might be a prettier solution I overlooked, let me know if there is
anything I should change!
<!--
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? -->
Adds the env arg `UV_NO_EDITABLE`.
Closes#12735
## Test Plan
<!-- How was it tested? -->
I could not find a place where to add tests, any help would be
appreciated
---------
Co-authored-by: Aria Desires <aria.desires@gmail.com>
Check that the source and module directory exist when build a source
distribution, instead of delaying the check to building the wheel. This
prevents building source distributions that can never be built into
wheels.
I removed the `set_cksum` as the value of it is replaced inside of
`append_data`.
## Summary
This should fix#12762 but I don't know how to test it.
---------
Co-authored-by: konstin <konstin@mailbox.org>
## Summary
I think the lack of enforcement here is an oversight. We _do_ already
enforce this for user-level configuration files (contrary to the issue
-- at least, in my testing and from reading the code).
Closes https://github.com/astral-sh/uv/issues/12753.
## Summary
See the test cases. Previously, you could end up with something like:
```toml
[tool.uv.workspace]
members = [
"foo",
"bar",
"baz", "bop",
]
```
<!--
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
Fixes#12606.
Two options considered, thanks to @zanieb's guidance are:
1. Special-casing on parse error and encountering the `environment.yml`
filename, possibly at `RequirementsTxt::parse`
2. Adding a new `RequirementsSource::EnvironmentYml` variant and
erroring on `RequirementSpecification::from_source`
I went with the latter for the following reasons:
- This edge case is explicitly modelled within the type system. However,
it changes the semantics of `RequirementsSource` to also model
_unsupported_ sources.
- (**Separation of concerns**) The special-casing would occur in the
`uv-requirements-txt` crate, which seems to be relatively deep in the
guts of the codebase. In my opinion, maintainers working in
`uv-requirements-txt` would reasonably assume the input file to be a
`requirements.txt` file, instead of having to be concerned with it being
another file format (`environment.yml`, `pyproject.toml`, etc.)
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
Manually tested as follows:
```sh
>>> cargo run -- pip install -r environment.yml
error: Conda environment file `environment.yml` is not supported
>>> cargo run -- add -r environment.yml
error: Conda environment file `environment.yml` is not supported
```
If you can point me to the appropriate test module, I can write up tests
for these to use `insta`.
<!-- How was it tested? -->
## Summary
Add an option to overwrite the list of available Python downloads from a
local JSON file by using the environment variable
`UV_PYTHON_DOWNLOADS_JSON_URL`
as an experimental support for providing custom sources for Python
distribution binaries #8015
related #10203
I probably should make the JSON to be fetched from a remote URL instead
of a local file.
please let me know what you think and I will modify the code
accordingly.
## Test Plan
### normal run
```
root@75c66494ba8b:/# /code/target/release/uv python list
cpython-3.14.0a4+freethreaded-linux-x86_64-gnu <download available>
cpython-3.14.0a4-linux-x86_64-gnu <download available>
cpython-3.13.1+freethreaded-linux-x86_64-gnu <download available>
cpython-3.13.1-linux-x86_64-gnu <download available>
cpython-3.12.8-linux-x86_64-gnu <download available>
cpython-3.11.11-linux-x86_64-gnu <download available>
cpython-3.10.16-linux-x86_64-gnu <download available>
cpython-3.9.21-linux-x86_64-gnu <download available>
cpython-3.8.20-linux-x86_64-gnu <download available>
cpython-3.7.9-linux-x86_64-gnu <download available>
pypy-3.10.14-linux-x86_64-gnu <download available>
pypy-3.9.19-linux-x86_64-gnu <download available>
pypy-3.8.16-linux-x86_64-gnu <download available>
pypy-3.7.13-linux-x86_64-gnu <download available>
```
### empty JSON file
```sh
root@75c66494ba8b:/# export UV_PYTHON_DOWNLOADS_JSON_URL=/code/crates/uv-python/my-download-metadata.json
root@75c66494ba8b:/# cat $UV_PYTHON_DOWNLOADS_JSON_URL
{}
root@75c66494ba8b:/# /code/target/release/uv python list
root@75c66494ba8b:/#
```
### JSON file with valid version
```sh
root@75c66494ba8b:/# export UV_PYTHON_DOWNLOADS_JSON_URL=/code/crates/uv-python/my-download-metadata.json
root@75c66494ba8b:/# cat $UV_PYTHON_DOWNLOADS_JSON_URL
{
"cpython-3.11.9-linux-x86_64-gnu": {
"name": "cpython",
"arch": {
"family": "x86_64",
"variant": null
},
"os": "linux",
"libc": "gnu",
"major": 3,
"minor": 11,
"patch": 9,
"prerelease": "",
"url": "https://github.com/astral-sh/python-build-standalone/releases/download/20240814/cpython-3.11.9%2B20240814-x86_64-unknown-linux-gnu-install_only_stripped.tar.gz",
"sha256": "daa487c7e73005c4426ac393273117cf0e2dc4ab9b2eeda366e04cd00eea00c9",
"variant": null
}
}
root@75c66494ba8b:/# /code/target/release/uv python list
cpython-3.11.9-linux-x86_64-gnu <download available>
root@75c66494ba8b:/#
```
### Remote Path
```sh
root@75c66494ba8b:/# export UV_PYTHON_DOWNLOADS_JSON_URL=http://a.com/file.json
root@75c66494ba8b:/# /code/target/release/uv python list
error: Remote python downloads JSON is not yet supported, please use a local path (without `file://` prefix)
```
---------
Co-authored-by: Aria Desires <aria.desires@gmail.com>
## Summary
closes#12234
[fetch_with_cli](e0f81f0d4a/crates/uv-git/src/git.rs (L573))
doesn't respect the registry client's [connectivity
setting](e0f81f0d4a/crates/uv-client/src/registry_client.rs (L1009))
- this pr updates `fetch_with_cli` to set `GIT_ALLOW_PROTOCOL=file` when
the client's connectivity setting is `Connectivity::Offline`
## Test Plan
E2E
```sh
cargo run add "pycurl @ git+https://github.com/pycurl/pycurl.git" --directory ~/src/offline-test/ --offline
```
```sh
Compiling uv-cli v0.0.1 (/Users/justinchapman/src/uv/crates/uv-cli)
Compiling uv v0.6.11 (/Users/justinchapman/src/uv/crates/uv)
Finished `dev` profile [unoptimized + debuginfo] target(s) in 4.47s
Running `target/debug/uv add 'pycurl @ git+https://github.com/pycurl/pycurl.git' --directory /Users/justinchapman/src/offline-test/ --offline`
Updating https://github.com/pycurl/pycurl.git (HEAD) × Failed to download and build `pycurl @ git+https://github.com/pycurl/pycurl.git`
├─▶ Git operation failed
├─▶ failed to fetch into: /Users/justinchapman/.cache/uv/git-v0/db/9a596e5213c3162d
╰─▶ process didn't exit successfully: `/usr/bin/git fetch --force --update-head-ok 'https://github.com/pycurl/pycurl.git' '+HEAD:refs/remotes/origin/HEAD'` (exit status: 128)
--- stderr
fatal: transport 'https' not allowed
help: If you want to add the package regardless of the failed resolution, provide the `--frozen` flag to skip locking and syncing.
```
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
@jtfmumm mentioned a desire for this. I'm not sure how we should do
this. I kind of want to change this to something like...
```
$ uv python find
CPython 3.13 @ <path>
$ uv python find --only-path
<path>
$ uv python find --short
<path>
$ uv python find --only-version
3.13
```
The change in defaults would be breaking though.
uv has a global `--no-config` option, which indeed has an effect.
## Test Plan
```console
❯ cat pyproject.toml
[project]
name = "hello"
version = "0.1.0"
[[tool.uv.index]]
url = "http://non-exist-host.com/simple"
default = true
❯ echo requests | uv pip compile -
⠦ Resolving dependencies... error: Failed to fetch: `http://non-exist-host.com/simple/requests/`
Caused by: Could not connect, are you offline?
Caused by: Request failed after 3 retries
Caused by: error sending request for url (http://non-exist-host.com/simple/requests/)
Caused by: client error (Connect)
Caused by: dns error: failed to lookup address information: nodename nor servname provided, or not known
Caused by: failed to lookup address information: nodename nor servname provided, or not known
# Before
❯ echo requests | uv pip compile --no-config -
warning: pip-compile's `--no-config` has no effect (uv does not use a configuration file)
Resolved 5 packages in 13ms
# This file was autogenerated by uv via the following command:
# uv pip compile --no-config -
certifi==2025.1.31
# via requests
charset-normalizer==3.4.1
# via requests
idna==3.10
# via requests
requests==2.32.3
urllib3==2.3.0
# via requests
# After
❯ echo requests | uv pip compile --no-config -
Resolved 5 packages in 13ms
# This file was autogenerated by uv via the following command:
# uv pip compile --no-config -
certifi==2025.1.31
# via requests
charset-normalizer==3.4.1
# via requests
idna==3.10
# via requests
requests==2.32.3
urllib3==2.3.0
# via requests
```
## Summary
Fix a suggestion in the docs on configs through environment variables,
which lists an option value that doesn't appear to exist.
The description implies that `unsafe-best-match` was intended here.
## Test Plan
Verified by providing `unsafe-any-match` as a parameter to `uv`. It
didn't error, but appeared to use the `first-index` strategy instead.
The value I changed it to behaves as described by the documentation.
In an attempt to avoid reporting shims as their resolved
`sys.executable` path, we report the queried executable path instead.
This seems more correct for this command, broadly? Interestingly, it
changes the reported paths for Homebrew Python
<img width="1430" alt="Screenshot 2025-04-02 at 11 05 18 AM"
src="https://github.com/user-attachments/assets/0e1600e8-fb07-40c7-a6d6-56eaeb4b9293"
/>
Closes#9979
This will in principle fix the problem reported in #12611 that
`authenticate = "always"` is ignored for an index when `explicit =
true`. This change ensures all indexes are added to the URL auth
policies list passed to our auth middleware.
Incorporates #12624Fixes#12611
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
Fixes#12618
Instead of succeeding the user now gets:
```
uvdloc pip install osqp==1.0.2 --reinstall --python-platform=linux
Resolved 7 packages in 171ms
× Failed to download `osqp==1.0.2`
├─▶ Failed to extract archive
╰─▶ a computed CRC32 value did not match the expected value
```
I am not entirely sure if we have infra for testing this kind of thing,
but it would be nice to check in a test or two. I'm also not entirely
clear if there's any cases where these checks are overzealous.
## Summary
I noticed that these only support Basic credentials, but we may want to
allow users to provide Bearer tokens? This PR just generalizes the type.
serde needs to be told where to put underscores. someone clearly noticed
this when adding attributes for schemars, but they need to be present
for serde too and then schemars gets them for free.
Strictly speaking this would be a breaking change for anyone who noticed
the parsing was messed up and worked around it. So we add aliases for
backcompat, at least for a few releases.
Fixes#12590
## Summary
<!-- What's the purpose of the change? What does it do, and why? -->
Support the `UV_PROJECT` environment variable to set the project
directory.
#11946
## Test Plan
<!-- How was it tested? -->
`cargo nextest run` passed except the cache_prune.
```
export UV_PROJECT=/path/to/project
uv sync
```
works.
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
The overall strategy here is to make this code look more like
`requirements_txt.rs`: we seed the root members, then perform a DFS.
Previously, we created all nodes upfront, which caused problems when
using `--only-group`, since we'd omit "production" dependencies of
development dependencies.
Closes https://github.com/astral-sh/uv/issues/12526.
Adding dependency trace/parent comments ("via ...") to the export
command output.
This is a similar behavior to the pip compile output.
#### Note to the eager reviewer:
First of all - thanks!
Secondly, this is still a very rough draft. These are the first lines of
code I've ever written in Rust. This is still mostly an educational/fun
exercise for myself. If opening a Draft PR is creating too much noise -
I apologize and I will close it until it is ready.
## Summary
Resolves#7777
## Test Plan
- [X] manual command execution
- [x] update expected output in tests
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
uv doesn't separate the metadata block from other blocks when adding the
`script` block to a script, which results in the next block being
considered part of the script block and causes errors when running.
See #12499 for more details.
Closes#12499
## Test Plan
I manually tested the most common scenario, but there's a few edge cases
that would be good to have tests for.
I would have written the tests also, but I was running into errors like
this:
```bash
$ cargo test --package uv-scripts
Compiling uv-configuration v0.0.1 (/home/merlin/Projects/uv/crates/uv-configuration)
error: cannot find attribute `value` in this scope
--> crates/uv-configuration/src/project_build_backend.rs:8:38
|
8 | #[cfg_attr(feature = "schemars", value(hide = true))]
| ^^^^^
error: could not compile `uv-configuration` (lib) due to 1 previous error
```
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
We tend not to run tests for individual crates, which can lead to weird
situations like this, where crates are missing optional features that
are otherwise installed globally.
## Test Plan
Run `cargo test --profile fast-build -p uv-scripts`, which otherwise
fails to compile.
The idea here is that we introduce a new stdout_important method for
things that want to care about the difference between "quiet" and
"silent".
This PR is WIP because it has no actual uses of stdout_important, and we
should have at least one before landing this. Perhaps someone has a
suggestion for commands that would really benefit from this distinction?
Fixes#10431
## Summary
This PR extends `[[tool.uv.index]]` to support `--find-links`-style
"flat" indexes, so that users can point to such indexes without using
`--find-links` _and_ get access to the full functionality of
`[[tool.uv.index]]` (e.g., they can now pin packages to
`--find-links`-style indexes).
Note that, at present, `--find-links` indexes actually have some quirky
behavior, in that we combine them into a single entity and then merge
the discovered distributions into each Simple API-style index. The
motivation here, IIRC, was to match pip's behavior quite closely. I'm
interested in _removing_ that behavior, but it'd be breaking (and may
also be inconvenient for some use-cases). So, the behavior for indexes
passed in via `--find-links` remains completely unchanged. However,
`[[tool.uv.index]]` entries with `format = "flat"` are now treated
identically to those defined with `format = "simple"` (the default), in
that we stop after we find the first-matching index, etc.
Closes https://github.com/astral-sh/uv/issues/11634.
## Summary
I don't know if I actually want to commit this, but I did it on the
plane last time and just polished it off (got it to compile) while
waiting to board.
We were not applying the `authenticate = "always"` behavior to `uv pip`
commands (related to #12362). This PR addresses that, applying
authentication policies wherever we set up a registry client.
## Summary
This fixes https://github.com/astral-sh/uv/issues/12426 which helps use
a more accurate arg name in the help output.
## Test Plan
I didn't test it locally, @charliermarsh gave me guidance on what to
change so I looked around that file for another example of `value_name`
and repeated what I saw. I kept it formatted to 1 line based on it not
being a long line. The other example of `value_name` had everything on
separate lines because there were a bunch of parameters passed in.
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
This PR modifies the requirement source entities to store a (new)
container struct that wraps `IndexUrl`. This will allow us to store
user-defined metadata alongside `IndexUrl`, and propagate that metadata
throughout resolution.
Specifically, I need to store the "kind" of the index (Simple API vs.
`--find-links`), but I also ran into this problem when I tried to add
support for overriding `Cache-Control` headers on a per-index basis: at
present, we have no way to passing around metadata alongside an
`IndexUrl`.
Match the module name to its module directory with potentially different
casing.
For example, a package may have the dist-info-normalized package name
`pil_util`, but the importable module is named `PIL_util`.
We get the module name either as dist-info-normalized package name, or
explicitly from the user. For dist-info-normalizing a package name, the
rules are lowercasing, replacing `.` with `_` and replace `-` with `_`.
Since `.` and `-` are not allowed in module names, we can check whether
a directory name matches our expected module name by lowercasing it.
Fixes#12187
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
## Summary
In general, we merge `--find-links` entries into each index. If a
package is pinned to an index, though, it seems surprising (and wrong)
that we'd ever select a distribution from `--find-links`. This PR
modifies the provider to ignore `--find-links` for any explicitly pinned
packages.
Allows `uv python list <request>` to filter the installed list. I often
want this and it's not hard to add.
I tested the remote download filtering locally (#12381 is needed for
snapshot tests)
```
❯ cargo run -q -- python list --all-versions 3.13
cpython-3.13.2-macos-aarch64-none <download available>
cpython-3.13.1-macos-aarch64-none /opt/homebrew/opt/python@3.13/bin/python3.13 -> ../Frameworks/Python.framework/Versions/3.13/bin/python3.13
cpython-3.13.1-macos-aarch64-none <download available>
cpython-3.13.0-macos-aarch64-none /Users/zb/.local/share/uv/python/cpython-3.13.0-macos-aarch64-none/bin/python3.13
❯ cargo run -q -- python list --all-versions 3.13 --only-installed
cpython-3.13.1-macos-aarch64-none /opt/homebrew/opt/python@3.13/bin/python3.13 -> ../Frameworks/Python.framework/Versions/3.13/bin/python3.13
cpython-3.13.0-macos-aarch64-none /Users/zb/.local/share/uv/python/cpython-3.13.0-macos-aarch64-none/bin/python3.13
```
## Summary
I want to use the flat index client from within the registry client, so
making them both depend on the same underlying primitives rather than
having the flat index client depend on the registry client.
## Summary
We respect `--exclude-newer` during resolution, but we weren't applying
it to individual _files_ when writing the lockfile. As a result, if
wheels were added to a distribution after its initial release, we'd end
up including them in the lockfile, even if they were uploaded after the
`--exclude-newer` date.
Closes https://github.com/astral-sh/uv/issues/12296.
## Summary
It's possible that the PyTorch version the user depends on isn't in the
latest index. These indexes are equally trusted, so we should override
the policy.
Closes#12357.
<!--
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? -->
This fixes a case described in #12333, where trailing comments in
dependencies can be unexpectedly shifted when a new dependency is added.
Fixes#12333.
## Test Plan
<!-- How was it tested? -->
`cargo test` (Added a snapshot test)
## Summary
Resolves#11794.
When `uv python find` is given a `--script` option, either the existing
environment for that script or the Python executable that would be used
to create it will be returned. If neither are found, the command exits
with exit code 1.
`--script` is incompatible with all other options to the same command.
## Test Plan
Unit tests.
## Summary
This crate is for standards-compliant types, but this is explicitly a
type that's custom to uv. It's also strange because we kind of want to
reference `IndexUrl` on the registry type, but that's in a crate that
_depends_ on `uv-pypi-types`, which to me is a sign that this is off.
There was a bug where `UV_MANAGED_PYTHON` and `UV_NO_MANAGED_PYTHON`
only accepted `true` or `false`. This switches to the boolish value
parser for those flags.
Closes#12336
## Summary
Fixes the failing `cache_prune::prune_unzipped` test that was causing CI
failures in my other PR (#12328) and others like PR #12327.
The error message format changed to show a specific version constraint
(`iniconfig<=2.0.0`) rather than the generic 'all versions' message.
This PR updates the test to expect the new, more specific error message.
## Test Plan
Ran `cargo test -p uv cache_prune::prune_unzipped` to verify the test
now passes.
Previously, we required a username to perform a fetch from the keyring
because the `keyring` CLI only supported fetching password for a given
service and username. Unfortunately, this is different from the keyring
Python API which supported fetching a username _and_ password for a
given service. We can't (easily) use the Python API because we don't
expect `keyring` to be installed in a specific environment during
network requests. This means that we did not have parity with `pip`.
Way back in https://github.com/jaraco/keyring/pull/678 we got a `--mode
creds` flag added to `keyring`'s CLI which supports parity with the
Python API. Since `keyring` is expensive to invoke and we cannot be
certain that users are on the latest version of keyring, we've not added
support for invoking keyring with this flag. However, now that we have a
mode that says authentication is _required_ for an index (#11896), we
might as well _try_ to invoke keyring with `--mode creds` when there is
no username. This will address use-cases where the username is
non-constant and move us closer to `pip` parity.
This addresses a small part of #12280, namely when you have
`authenticate` set to `always`, it will output a distinct error message
for the case where you have a username but are missing a password.
## Summary
This is a prototype that I'm considering shipping under `--preview`,
based on [`light-the-torch`](https://github.com/pmeier/light-the-torch).
`light-the-torch` patches pip to pull PyTorch packages from the PyTorch
indexes automatically. And, in particular, `light-the-torch` will query
the installed CUDA drivers to determine which indexes are compatible
with your system.
This PR implements equivalent behavior under `--torch-backend auto`,
though you can also set `--torch-backend cpu`, etc. for convenience.
When enabled, the registry client will fetch from the appropriate
PyTorch index when it sees a package from the PyTorch ecosystem (and
ignore any other configured indexes, _unless_ the package is explicitly
pinned to a different index).
Right now, this is only implemented in the `uv pip` CLI, since it
doesn't quite fit into the lockfile APIs given that it relies on feature
detection on the currently-running machine.
## Test Plan
On macOS, you can test this with (e.g.):
```shell
UV_TORCH_BACKEND=auto UV_CUDA_DRIVER_VERSION=450.80.2 cargo run \
pip install torch --python-platform linux --python-version 3.12
```
On a GPU-enabled EC2 machine:
```shell
ubuntu@ip-172-31-47-149:~/uv$ UV_TORCH_BACKEND=auto cargo run pip install torch -v
Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.31s
Running `target/debug/uv pip install torch -v`
DEBUG uv 0.6.6 (e95ca063b 2025-03-14)
DEBUG Searching for default Python interpreter in virtual environments
DEBUG Found `cpython-3.13.0-linux-x86_64-gnu` at `/home/ubuntu/uv/.venv/bin/python3` (virtual environment)
DEBUG Using Python 3.13.0 environment at: .venv
DEBUG Acquired lock for `.venv`
DEBUG At least one requirement is not satisfied: torch
warning: The `--torch-backend` setting is experimental and may change without warning. Pass `--preview` to disable this warning.
DEBUG Detected CUDA driver version from `/sys/module/nvidia/version`: 550.144.3
...
```
## Summary
I think this is reasonable to change. Right now, if you're on Python
3.11, the resolver returns `multiprocess-0.70.17-py311-none-any.whl`,
but `multiprocess-0.70.17-py310-none-any.whl` is in the cache, we'll
reuse `multiprocess-0.70.17-py310-none-any.whl` (since it _is_
compatible with Python 3.11).
Instead, we now _require_ the cached wheel to match the wheel returned
by the resolver.
Closes https://github.com/astral-sh/uv/issues/12273.
Currently, for users to specify at the command line whether to use
uv-managed or system Python interpreters, they use the
`--python-preference` parameter, which takes four possible values. This
is more complex than necessary since the normal case is to either say
"only managed" or "not managed". This PR hides the old
`--python-preference` parameter from help and documentation and adds two
new flags: `--managed-python` and `--no-managed-python` to capture the
"only managed" and "not managed" cases.
I have successfully tested this locally but currently cannot add
snapshot tests because of problems with distinguishing managed vs.
system interpreters in CI (and non-determinism when run on different
developers' machines). The `--python-preference` test in
`tool-install.rs` is currently ignored for this reason. See #5144 and
#7473.
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
## Summary
I suspect this only affects packages with quotes in the requires-python,
which is typically an error but one that we correct for in
`LenientVersionSpecifiers`.
Closes https://github.com/astral-sh/uv/issues/12260.
Charlie Marsh
Dennis Dang
John Mumm
Aria Desires
konsti
Jo
Geoffrey Thomas
renovate[bot]
Michał Górny
Zanie Blue
samypr100
Eric Nielsen
Andrew Pollack
Ahmed Ilyas
Eduardo Rittner Coelho
Frazer McLean
Tobias Gårdhus
Tim Felgentreff
Christopher Tee
Meitar Reihan
Bartosz Sokorski
ReinforcedKnowledge
Ben Beasley
Max Mynter
johnthagen
Alex Prengère
Dheepak Krishnamurthy
Chandra Kiran G
Arthur Bied-Charreton
Haaris Rahman
github-actions[bot]
leiserfg
justin
Tom Boshoven
Nozomi R
Zohar Meir
Merlin
Nick Janetakis
InSync
Christian Bager Bach Houmann