If a user specifies `-e /path/to/dir` and `/path/to/dir` in a `uv pip
install` command, we want the editable to "win" (rather than erroring
due to conflicting URLs). Unfortunately, this behavior meant that when
you requested a package as editable and non-editable in conflicting
groups, the editable version was _always_ used. This PR modifies the
requisite types to use `Option<bool>` rather than `bool` for the
`editable` field, so we can determine whether a requirement was
explicitly requested as editable, explicitly requested as non-editable,
or not specified (as in the case of `/path/to/dir` in a
`requirements.txt` file). In the latter case, we allow editables to
override the "unspecified" requirement.
If a project includes a path dependency twice, once with `editable =
true` and once without any `editable` annotation, those are now
considered conflicting URLs, and lead to an error, so I've marked this
change as breaking.
Closes https://github.com/astral-sh/uv/issues/14139.
If `--workspace` is provided, we add all paths as workspace members.
If `--no-workspace` is provided, we add all paths as direct path
dependencies.
If neither is provided, then we add any paths that are under the
workspace root as workspace members, and the rest as direct path
dependencies.
Closes#14524.
While reviewing https://github.com/astral-sh/uv/pull/14107, @oconnor663
pointed out a bug where we allow `uv python pin --rm` to delete the
global pin without the `--global` flag. I think that shouldn't be
allowed? I'm not 100% certain though.
Adds environment variables for
https://github.com/astral-sh/uv/pull/14612 and
https://github.com/astral-sh/uv/pull/14614
We can't use the Clap `BoolishValueParser` here, and the reasoning is a
little hard to explain. If we used `UV_PYTHON_INSTALL_NO_BIN`, as is our
typical pattern, it'd work, but here we allow opt-in to hard errors with
`UV_PYTHON_INSTALL_BIN=1` and I don't think we should have both
`UV_PYTHON_INSTALL_BIN` and `UV_PYTHON_INSTALL_NO_BIN`.
Consequently, this pull request introduces a new `EnvironmentOptions`
abstraction which allows us to express semantics that Clap cannot —
which we probably want anyway because we have an increasing number of
environment variables we're parsing downstream, e.g., #14544 and #14369.
## Summary
When refactoring the addition PR I accidentally introduced a bug where
the debug message would not be output if the default value is used.
cc @zanieb
## Summary
When installing packages on _very_ slow/overloaded systems it'spossible
to trigger bytecode compilation timeouts, which tends to happen in
environments such as Qemu (especially without KVM/virtio), but also on
systems that are simply overloaded. I've seen this in my Nix builds if I
for example am compiling a Linux kernel at the same time as a few other
concurrent builds.
By making the bytecode compilation timeout adjustable you can work
around such issues. I plan to set `UV_COMPILE_BYTECODE_TIMEOUT=0` in the
[pyproject.nix
builders](https://pyproject-nix.github.io/pyproject.nix/build.html) to
make them more reliable.
- Related issues
* https://github.com/astral-sh/uv/issues/6105
## Test Plan
Only manual testing was applied in this instance. There is no existing
automated tests for bytecode compilation timeout afaict.
Closes#14262
## Description
Adds `UV_LIBC` environment variable and implements check within
`Libc::from_env` as recommended here:
https://github.com/astral-sh/uv/issues/14262#issuecomment-3014600313
Gave this a few passes to make sure I follow dev practices within uv as
best I am able. Feel free to call out anything that could be improved.
## Test Plan
Planned to simply run existing test suite. Open to adding more tests
once implementation is validated due to my limited Rust experience.
## Summary
We validate the `uv.toml` when it's discovered automatically, but not
when provided via `--config-file`. The same limitations exist, though --
I think the lack of enforcement is just an oversight.
Closes https://github.com/astral-sh/uv/issues/14650.
Part of #14296
This is the same as `uv tool update-shell` but handles the case where
the Python bin directory is configured to a different path.
```
❯ UV_PYTHON_BIN_DIR=/tmp/foo cargo run -q -- python install --preview 3.13.3
Installed Python 3.13.3 in 1.75s
+ cpython-3.13.3-macos-aarch64-none
warning: `/tmp/foo` is not on your PATH. To use installed Python executables, run `export PATH="/tmp/foo:$PATH"` or `uv python update-shell`.
❯ UV_PYTHON_BIN_DIR=/tmp/foo cargo run -q -- python update-shell
Created configuration file: /Users/zb/.zshenv
Restart your shell to apply changes
❯ cat /Users/zb/.zshenv
# uv
export PATH="/tmp/foo:$PATH"
❯ UV_TOOL_BIN_DIR=/tmp/bar cargo run -q -- tool update-shell
Updated configuration file: /Users/zb/.zshenv
Restart your shell to apply changes
❯ cat /Users/zb/.zshenv
# uv
export PATH="/tmp/foo:$PATH"
# uv
export PATH="/tmp/bar:$PATH"
```
Previously, if installation of executables into the bin directory failed
we'd with a non-zero code. However, if we make this behavior the default
we don't want it to be fatal. There's a `--bin` opt-in to _require_
successful executable installation and a `--no-bin` opt-out to silence
the warning / opt-out of installation entirely.
Part of https://github.com/astral-sh/uv/issues/14296 — we need this
before we can stabilize the behavior.
In #14614 we do the same for writing entries to the Windows registry.
## Summary
You can now override the cache control headers for the Simple API, file
downloads, or both:
```toml
[[tool.uv.index]]
name = "example"
url = "https://example.com/simple"
cache-control = { api = "max-age=600", files = "max-age=365000000, immutable" }
```
Closes https://github.com/astral-sh/uv/issues/10444.
## Summary
There's some inconsistent behaviour in handling symlinks when
`cache-key` is a glob or a file path. This PR attempts to address that.
- When cache-key is a path,
[`Path::metadata()`](https://doc.rust-lang.org/std/path/struct.Path.html#method.metadata)
is used to check if it's a file or not. According to the docs:
> This function will traverse symbolic links to query information about
the destination file.
So, if the target file is a symlink, it will be resolved and the
metadata will be queried for the underlying file.
- When cache-key is a glob, `globwalk` is used, specifically allowing
for symlinks:
```rust
.file_type(globwalk::FileType::FILE | globwalk::FileType::SYMLINK)
```
- However, without enabling link following, `DirEntry::metadata()` will
return an equivalent of `Path::symlink_metadata()` (and not
`Path::metadata()`), which will have a file type that looks like
```rust
FileType {
is_file: false,
is_dir: false,
is_symlink: true,
..
}
```
- Then, there's a check for `metadata.is_file()` which fails and
complains that the target entry "is a directory when file was expected".
- TLDR: glob cache-keys don't work with symlinks.
## Solutions
Option 1 (current PR): follow symlinks.
Option 2 (also doable): don't follow symlinks, but resolve the resulting
target entry manually in case its file type is a symlink. However, this
would be a little weird and unobvious in that we resolve files but not
directories for some reason. Also, symlinking directories is pretty
useful if you want to symlink directories of local dependencies that are
not under the project's path.
## Test Plan
This has been tested manually:
```rust
fn main() {
for follow_links in [false, true] {
let walker = globwalk::GlobWalkerBuilder::from_patterns(".", &["a/*"])
.file_type(globwalk::FileType::FILE | globwalk::FileType::SYMLINK)
.follow_links(follow_links)
.build()
.unwrap();
let entry = walker.into_iter().next().unwrap().unwrap();
dbg!(&entry);
dbg!(entry.file_type());
dbg!(entry.path_is_symlink());
dbg!(entry.path());
let meta = entry.metadata().unwrap();
dbg!(meta.is_file());
}
let path = std::path::PathBuf::from("./a/b");
dbg!(path.metadata().unwrap().file_type());
dbg!(path.symlink_metadata().unwrap().file_type());
}
```
Current behaviour (glob cache-key, don't follow links):
```
[src/main.rs:9:9] &entry = DirEntry("./a/b")
[src/main.rs:10:9] entry.file_type() = FileType {
is_file: false,
is_dir: false,
is_symlink: true,
..
}
[src/main.rs:11:9] entry.path_is_symlink() = true
[src/main.rs:12:9] entry.path() = "./a/b"
[src/main.rs:14:9] meta.is_file() = false
```
Glob cache-key, follow links:
```
[src/main.rs:9:9] &entry = DirEntry("./a/b")
[src/main.rs:10:9] entry.file_type() = FileType {
is_file: true,
is_dir: false,
is_symlink: false,
..
}
[src/main.rs:11:9] entry.path_is_symlink() = true
[src/main.rs:12:9] entry.path() = "./a/b"
[src/main.rs:14:9] meta.is_file() = true
```
Using `path.metadata()` for a non-glob cache key:
```
[src/main.rs:18:5] path.metadata().unwrap().file_type() = FileType {
is_file: true,
is_dir: false,
is_symlink: false,
..
}
[src/main.rs:19:5] path.symlink_metadata().unwrap().file_type() = FileType {
is_file: false,
is_dir: false,
is_symlink: true,
..
}
```
This is a continuation of the work in
* #12405
I have:
* moved to an architecture where the human output is derived from the
json structs to centralize more of the printing state/logic
* cleaned up some of the names/types
* added tests
* removed the restriction that this output is --dry-run only
I have not yet added package info, which was TBD in their design.
---------
Co-authored-by: x0rw <mahdi.svt5@gmail.com>
Co-authored-by: Zanie Blue <contact@zanie.dev>
Co-authored-by: John Mumm <jtfmumm@gmail.com>
We've seen a few cases of uv.exe exiting with an exception code as its
exit status and no user-visible output (#14563 in the field, and #13812
in CI). It seems that recent versions of Windows no longer show dialog
boxes on access violations (what UNIX calls segfaults) or similar
errors. Something is probably sent to Windows Error Reporting, and we
can maybe sign up to get the crashes from Microsoft, but the user
experience of seeing uv exit with no output is poor, both for end users
and during development. While it's possible to opt out of this behavior
or set up a debugger, this isn't the default configuration. (See
https://superuser.com/q/1246626 for some pointers.)
In order to get some output on a crash, we need to install our own
default handler for unhandled exceptions (or call all our code inside a
Structured Exception Handling __try/__catch block, which is complicated
on Rust). This is the moral equivalent of a segfault handler on Windows;
the kernel creates a new stack frame and passes arguments to it with
some processor state.
This commit adds a relatively simple exception handler that leans on
Rust's own backtrace implementation and also displays some minimal
information from the exception itself. This should be enough info to
communicate that something went wrong and let us collect enough
information to attempt to debug. There are also a handful of (non-Rust)
open-source libraries for this like Breakpad and Crashpad (both from
Google) and crashrpt.
The approach here, of using SetUnhandledExceptionFilter, seems to be the
standard one taken by other such libraries. Crashpad also seems to try
to use a newer mechanism for an out-of-tree DLL to report the crash:
https://issues.chromium.org/issues/42310037
If we have serious problems with memory corruption, it might be worth
adopting some third-party library that has already implemented this
approach. (In general, the docs of other crash reporting libraries are
worth skimming to understand how these things ought to work.)
Co-authored-by: samypr100 <3933065+samypr100@users.noreply.github.com>
## Summary
(Related PR: #13438 - would be nice to have it merged as well since it
touches on the same globwalker code)
There's a few issues with `cache-key` globs, which this PR attempts to
address:
- As of the current state, parent or absolute paths are not allowed,
which is not obvious and is not documented. E.g., cache-key paths of the
form `{file = "../dep/**"}` will be essentially ignored.
- Absolute glob patterns also don't work (funnily enough, there's logic
in `globwalk` itself that attempts to address it in
[`globwalk::glob_builder()`](8973fa2bc5/src/lib.rs (L415)),
which serves as inspiration to some parts of this PR).
- The reason for parent paths being ignored is the way globwalker is
currently being triggered in `uv-cache-info`: the base directory is
being walked over completely and each entry is then being matched to one
of the provided match patterns.
- This may also end up being very inefficient if you have a huge root
folder with thousands of files: if your match patterns are `a/b/*.rs`
and `a/c/*.py` then instead of walking over the root directory, you can
just walk over `a/b` and `a/c` and match the relevant patterns there.
- Why supporting parent paths may be important to the point of being a
blocker: in large codebases with python projects depending on other
local non-python projects (e.g. rust crates), cache-keys can be very
useful to track dependency on the source code of the latter (e.g.
`cache-keys = [{ file = "../../crates/some-dep/**" }]`.
- TLDR: parent/absolute cache-key globs don't work, glob walk can be
slow.
## Solution
- In this PR, user-provided glob patterns are first clustered
(LCP-style) into pattern groups with longest common path prefix; each of
these groups can then be walked over separately.
- Pattern groups do not overlap, so we would never walk over the same
directory twice (unless there's symlinks pointing to same folders).
- Paths are not canonicalized nor virtually normalized (which is
impossible on Unix without FS access), so the method is symlink-safe
(i.e. we don't treat `a/b/..` as `a`) and should work fine with #13438.
- Because of LCP logic, the minimal amount of directory space will be
traversed to cover all patterns.
- Absolute glob patterns will now work.
- Parent-relative glob patterns will now work.
- Glob walking will be more efficient in some cases.
## Possible improvements
- Efficiency can be further greatly improved if we limit max depth for
globwalk. Currently, a simple ".toml" will deep-traverse the whole
folder. Essentially, max depth can be always set to either N or
infinity. If a pattern at a pivot node contains `**`, we collect all
children nodes from the subtree into the same group and don't limit max
depth; otherwise, we set max depth to the length of the glob pattern.
This wouldn't change correctness though and can we done separately if
needed.
- If this is considered important enough, docs can be updated to
indicate that parent and absolute globs are supported (and symlinks are
resolved, if the relevant PR is algo merged in).
## Test Plan
- Glob splitting and clustering tests are included in the PR.
- Relative and absolute glob cache-keys were tested in an actual
codebase.
<!--
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
This is a small quality of life feature that adds a shorthand (`-w`) to
the `--with` flag for minimizing keystrokes.
Pretty minor, but I didn't see any conflicts with `-w` and thought this
could be a nice place for it.
```bash
# proposed addition (short)
uvx -w numpy ipython
# original (long)
uvx --with numpy ipython
```
## Test Plan
Added testing already in the P.R. - just copied over tests from the
`--with` flag
<!-- How was it tested? -->
<!--
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
The `version_extras` test added in
85c0fc963b needs to connect to PyPI. This
PR conditionalizes it on the `pypi` extra so that people running the
tests offline don’t have to skip that test explicitly.
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
<!-- How was it tested? -->
I already ran `cargo test` in the git checkout to confirm I didn’t
somehow introduce a syntax error. I am also applying this PR as a patch
to [the `uv` package in Fedora](https://src.fedoraproject.org/rpms/uv),
which runs tests offline with the `pypi` feature disabled.
Follow-up to https://github.com/astral-sh/uv/pull/14509 to provide the
_reason_ downloads are disabled and surface it as a hint rather than a
debug log.
e.g.,
```
❯ cargo run -q -- run --no-managed-python -p 3.13.4 python
error: No interpreter found for Python 3.13.4 in virtual environments or search path
hint: A managed Python download is available for Python 3.13.4, but the Python preference is set to 'only system'
```
This adds `alpha`, `beta`, `rc`, `stable`, `post`, and `dev` modes to
`uv version --bump`.
The components that `--bump` accepts are ordered as follows:
major > minor > patch > stable > alpha > beta > rc > post > dev
Bumping a component "clears" all lesser component (`alpha`, `beta`, and
`rc` all overwrite each other):
* `--bump minor` on `1.2.3a4.post5.dev6` => `1.3.0`
* `--bump alpha` on `1.2.3a4.post5.dev6` => `1.2.3a5`
* `--bump dev ` on `1.2.3a4.post5.dev6` => `1.2.3a4.post5.dev7`
In addition, `--bump` can now be repeated. The primary motivation of
this is "bump stable version and also enter a prerelease", but it
technically lets you express other things if you want them:
* `--bump patch --bump alpha` on `1.2.3` => `1.2.4a1` ("bump patch
version and go to alpha 1")
* `--bump minor --bump patch` on `1.2.3` => `1.3.1` ("bump minor version
and got to patch 1")
* `--bump minor --bump minor` on `1.2.3` => `1.4.0` ("bump minor version
twice")
The `--bump` flags are sorted by their priority, so that you don't need
to remember the priority yourself. This ordering is the only "useful"
one that preserves every `--bump` you passed, so there's no concern
about loss of expressiveness. For instance `--bump minor --bump major`
would just be `--bump major` if we didn't sort, as the major bump clears
the minor version. The ordering of `beta` after `alpha` means `--bump
alpha --bump beta` will just result in beta 1; this is the one case
where a bump request will effectively get overwritten.
The `stable` mode "bumps to the next stable release", clearing the pre
(`alpha`, `beta`, `rc`), `dev`, and `post` components from a version
(`1.2.3a4.post5.dev6` => `1.2.3`). The choice to clear `post` here is a
bit odd, in that `1.2.3.post4` => `1.2.3` is actually a version
decrease, but I think this gives a more intuitive model (as preserving
`post5` in the previous example is definitely wrong), and also
post-releases are extremely obscure so probably no one will notice. In
the cases where this behaviour isn't useful, you probably wanted to pass
`--bump patch` or something anyway which *should* definitely clear the
`post5` (putting it another way: the only cases where `--bump stable`
has dubious behaviour is when you wanted it to do a noop, which, is a
command you could have just not written at all).
In all cases we preserve the "epoch" and "local" components of a
version, so the `7!` and `+local` in `7!1.2.3+local` will never be
modified by `--bump` (you can use the raw version set mode if you want
to touch those). The preservation of `local` is another slightly odd
choice, but it's a really obscure feature (so again it mostly won't come
up) and when it's used it seems to mostly be used for referring to
variant releases, in which case preserving it tends to be correct.
Fixes#13223
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
<!--
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
remove redundant words in comment
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
<!-- How was it tested? -->
Signed-off-by: jingchanglu <jingchanglu@outlook.com>
Support multiple root modules in namespace packages by enumerating them:
```toml
[tool.uv.build-backend]
module-name = ["foo", "bar"]
```
This allows applications with multiple root packages without migrating
to workspaces. Since those are regular module names (we iterate over
them an process each one like a single module names), it allows
combining dotted (namespace) names and regular names. It also
technically allows combining regular and stub modules, even though this
is even less recommends.
We don't recommend this structure (please use a workspace instead, or
structure everything in one root module), but it reduces the number of
cases that need `namespace = true`.
Fixes#14435Fixes#14438
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
## Summary
This PR intends to enable `--torch-backend=auto` to detect Intel GPUs
automatically:
- On Linux, detection is performed using the `lspci` command via
`Display controller` id.
- On Windows, ~~detection is done via a `powershell` query to
`Win32_VideoController`~~. Skip support for now—revisit once a better
solution is available.
Currently, Intel GPUs (XPU) do not rely on specific driver or toolkit
versions to distribute different PyTorch wheels.
## Test Plan
<!-- How was it tested? -->
On Linux:
~~On Windows:
~~
---------
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
Reverts:
- #14349
- #14346
- #14245
Retains the test cases. Includes a `find-links` test case.
Supersedes
- https://github.com/astral-sh/uv/pull/14387
- https://github.com/astral-sh/uv/pull/14503
We originally got a report at
https://github.com/astral-sh/uv/issues/13707 that inclusion of a
trailing slash on an index URL was causing lockfile churn despite having
no semantic meaning and resolved the issue by adding normalization that
stripped trailing slashes at parse time.
We then discovered that, while there are not semantic differences for
trailing slashes on Simple API index URLs, there are differences for
some flat (or find links) indexes. As reported in
https://github.com/astral-sh/uv/issues/14367, the change in
https://github.com/astral-sh/uv/pull/14245 caused a regression for at
least one user.
We attempted to fix the regression via a few approaches.
https://github.com/astral-sh/uv/pull/14387 attempted to differentiate
between Simple API and flat index URL parsing, but failed to account for
the `Deserialize` implementation, which always assumed Simple API-style
index URLs and incorrectly trimmed trailing slashes in various cases
where we deserialized the `IndexUrl` type from a file. I attempted to
resolve this by performing a larger refactor, but it ended up being
quite painful. In particular, the `Index` type was a blocker — we don't
know the `IndexUrl` variant until we've parsed the `IndexFormat` and
having a multi-stage deserializer is not appealing but adding a new
intermediate type (i.e., `RawIndex`) is painful due to the pervasiveness
of `Index`. Given that we've regressed behavior here and there's not a
straight-forward fix, we're reverting the normalization entirely.
https://github.com/astral-sh/uv/pull/14503 attempted to perform
normalization at compare-time, but that means we'd fail to invalidate
the lockfile when the a trailing slash was added or removed and given
that a trailing slash has semantic meaning for a find-links URL... we'd
have another correctness problem.
After this revert, we'll retain all index URLs verbatim. The downside to
this approach is that we'll be adding a bunch of trailing slashes back
to lockfiles that we previously normalized out, and we'll be reverting
our fix for users with inconsistent trailing slashes on their index
URLs. Users affected by the original motivating issue should use
consistent trailing slashes on their URLs, as they do frequently have
semantic meaning. We may want to revisit normalization and type aware
index URL parsing as part of a larger change.
Closes https://github.com/astral-sh/uv/issues/14367
<!--
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
We are using UV as a library and `installer()` returned `"pip\n"`. The
packages got installed by the pip package manager and not by UV. pip
seems to add a new line to the `INSTALLER` file and UV does not.
<!-- What's the purpose of the change? What does it do, and why? -->
## Test Plan
<!-- How was it tested? -->
When [updating](https://github.com/astral-sh/uv/pull/14475) to the
latest `reqwest` version, our fragment propagation test broke. That test
was partially testing the `reqwest` behavior, so this PR moves the
fragment test to directly test our logic for constructing redirect
requests.
<!--
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? -->
In pixi we overlay the PyPI packages over the conda packages and we
sometimes need to figure out what PyPI packages are involved in the
no-solution error. We could parse the error message, but this is pretty
error-prone, so it would be good to get access to more information. A
lot of information in this module is private and should probably stay
this way, but package names are easy enough to expose. This would help
us a lot!
I collect into a HashSet to remove duplication, and did not want to
expose a rustc_hash datastructure directly, thats's why I've chosen to
expose as an iterator :)
Let me know if any changes need to be done, and thanks!
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
Hey, are you okay with exposing the `ErrorTree` for library consumers?
We have a use case that needs more information on conflicts. We need the
tree-structure of the conflict and be able to traverse it in particular.
Signed-off-by: Simon Sure <ssure@palantir.com>
The uv build backend has gone through some feedback cycles, we expect no
more major configuration changes, and we're ready to take the next step:
The uv build backend in stable.
This PR stabilizes:
* Using `uv_build` as build backend
* The documentation of the uv build backend
* The direct build fast path, where uv doesn't use PEP 517 if you're
using `uv_build` in a compatible version.
* `uv build --list`, which is limited to `uv_build`.
It does not:
* Make `uv_build` the default on `uv init`
* Make `--package` the default on `uv init`
## Summary
If we fail to acquire a lock on an environment, uv shouldn't fail; we
should just warn. In some cases, users run uv with read-only permissions
for their projects, etc.
For now, I kept any locks acquired _in the cache_ as hard failures,
since we always need write-access to the cache.
Closes https://github.com/astral-sh/uv/issues/14411.
## Summary
The idea here is that if a user runs `uv pip compile --universal`, we
should ignore the patch version on the current interpreter. I think this
makes sense... `--universal` tries to resolve for all future versions,
so it seems a bit odd that we'd start at the _current_ patch version.
Closes https://github.com/astral-sh/uv/issues/14397.
Clap does not perform global validation, so flag that are declared as
overriding can be set at the same time:
https://github.com/clap-rs/clap/issues/6049. This would previously cause
a panic. We work around this by choosing the yes-value always and
writing a warning.
An alternative would be erroring when both are set, but it's unclear to
me if this may break things we want to support. (`UV_OFFLINE=1 cargo run
-q pip --no-offline install tqdm --no-cache` is already banned).
Fixes https://github.com/astral-sh/uv/pull/14299
**Test Plan**
```
$ cargo run -q pip --offline install --no-offline tqdm --no-cache
warning: Boolean flags on different levels are not correctly supported (https://github.com/clap-rs/clap/issues/6049)
× No solution found when resolving dependencies:
╰─▶ Because tqdm was not found in the cache and you require tqdm, we can conclude that your requirements are unsatisfiable.
hint: Packages were unavailable because the network was disabled. When the network is disabled, registry packages may only be read from the cache.
```
## Summary
The basic idea here is that we can (should) reuse a build environment
across resolution (`prepare_metadata_for_build_wheel`) and installation.
This also happens to solve the build-PyTorch-from-source problem, since
we use a consistent build environment between the invocations.
Since `SourceDistributionBuilder` is stateless, we instead store the
builds on `BuildContext`, and we key them by various properties: the
underlying interpreter, the configuration settings, etc. This just
ensures that if we build the same package twice within a process, we
don't accidentally reuse an incompatible build (virtual) environment.
(Note that still drop build environments at the end of the command, and
don't attempt to reuse them across processes.)
Closes#14269.
If/when we see https://github.com/astral-sh/uv/issues/14171 again, this
should clarify whether our retry logic was skipped (i.e. a transient
error wasn't correctly identified as transient), or whether we exhausted
our retries. Previously, if you ran a local example fileserver as in
https://github.com/astral-sh/uv/issues/14171#issuecomment-3014580701 and
then you tried to install Python from it, you'd get:
```
$ export UV_TEST_NO_CLI_PROGRESS=1
$ uv python install 3.8.20 --mirror http://localhost:8000 2>&1 | cat
error: Failed to install cpython-3.8.20-linux-x86_64-gnu
Caused by: Failed to extract archive: cpython-3.8.20-20241002-x86_64-unknown-linux-gnu-install_only_stripped.tar.gz
Caused by: failed to unpack `/home/jacko/.local/share/uv/python/.temp/.tmpS4sHHZ/python/lib/libpython3.8.so.1.0`
Caused by: failed to unpack `python/lib/libpython3.8.so.1.0` into `/home/jacko/.local/share/uv/python/.temp/.tmpS4sHHZ/python/lib/libpython3.8.so.1.0`
Caused by: error decoding response body
Caused by: request or response body error
Caused by: error reading a body from connection
Caused by: Connection reset by peer (os error 104)
```
With this change you get:
```
error: Failed to install cpython-3.8.20-linux-x86_64-gnu
Caused by: Request failed after 3 retries
Caused by: Failed to extract archive: cpython-3.8.20-20241002-x86_64-unknown-linux-gnu-install_only_stripped.tar.gz
Caused by: failed to unpack `/home/jacko/.local/share/uv/python/.temp/.tmp4Ia24w/python/lib/libpython3.8.so.1.0`
Caused by: failed to unpack `python/lib/libpython3.8.so.1.0` into `/home/jacko/.local/share/uv/python/.temp/.tmp4Ia24w/python/lib/libpython3.8.so.1.0`
Caused by: error decoding response body
Caused by: request or response body error
Caused by: error reading a body from connection
Caused by: Connection reset by peer (os error 104)
```
At the same time, I'm updating the way we handle the retry count to
avoid nested retry loops exceeding the intended number of attempts, as I
mentioned at
https://github.com/astral-sh/uv/issues/14069#issuecomment-3020634281.
It's not clear to me whether we actually want this part of the change,
and I need feedback here.
## Summary
This PR ensures that we avoid cleaning up build directories until the
end of a resolve-and-install cycle. It's not bulletproof (since we could
still run into issues with `uv lock` followed by `uv sync` whereby a
build directory gets cleaned up that's still referenced in the `build`
artifacts), but it at least gets PyTorch building without error with `uv
pip install .`, which is a case that's been reported several times.
Closes https://github.com/astral-sh/uv/issues/14269.
The marker display code assumes that all versions are normalized, in
that all trailing zeroes are stripped. This is not the case for
tilde-equals and equals-star versions, where the trailing zeroes (before
the `.*`) are semantically relevant. This would cause path
dependent-behavior where we would get a different marker string
depending on whether a version with or without a trailing zero was added
to the cache first.
To handle both equals-star and tilde-equals when converting
`python_version` to `python_full_version` markers, we have to merge the
version normalization (i.e. trimming the trailing zeroes) and the
conversion both to `python_full_version` and to `Ranges`, while special
casing equals-star and tilde-equals.
To avoid churn in lockfiles, we only trim in the conversion to `Ranges`
for markers, but keep using untrimmed versions for requires-python.
(Note that this behavior is technically also path dependent, as versions
with and without trailing zeroes have the same Hash and Eq. E.q.,
`requires-python == ">= 3.10.0"` and `requires-python == ">= 3.10"` in
the same workspace could lead to either value in `uv.lock`, and which
one it is could change if we make unrelated (performance) changes.
Always trimming however definitely changes lockfiles, a churn I wouldn't
do outside another breaking or lockfile-changing change.) Nevertheless,
there is a change for users who have `requires-python = "~= 3.12.0"` in
their `pyproject.toml`, as this now hits the correct normalization path.
Fixes#14231Fixes#14270
## Summary
There's a good example of the downside of using verbatim URLs here:
https://github.com/astral-sh/uv/pull/14197#discussion_r2163599625 (we
show two relative paths that point to the same directory, but it's not
clear from the error message).
The diff:
```
2 2 │ ----- stdout -----
3 3 │
4 4 │ ----- stderr -----
5 5 │ error: Requirements contain conflicting URLs for package `library` in all marker environments:
6 │-- ../../library
7 │-- ./library
6 │+- file://[TEMP_DIR]/library
7 │+- file://[TEMP_DIR]/library (editable)
```
## Summary
<!-- What's the purpose of the change? What does it do, and why? -->
As explained in the [`codspeed-rust` v3 release
notes](https://github.com/CodSpeedHQ/codspeed-rust/releases/tag/v3.0.0),
the `v3` of the compatibility layers is now required to work with the
latest version(`v3`) of `cargo-codspeed`.
and prefer emulated x64 windows in its stead.
This is preparatory work for shipping support for uv downloading and
installing aarch64 (arm64) windows Pythons. We've [had builds for this
platform ready for a
while](https://github.com/astral-sh/python-build-standalone/pull/387),
but have held back on shipping them due to a fundamental problem:
**The Python packaging ecosystem does not have strong support for
aarch64 windows**, e.g., not many projects build aarch64 wheels yet. The
net effect of this is that, if we handed you an aarch64 python
interpreter on windows, you would have to build a lot more sdists, and
there's a high chance you will simply fail to build that sdist and be
sad.
Yes unfortunately, in this case a non-native Python interpreter simply
*works better* than the native one... in terms of working at all, today.
Of course, if the native interpreter works for your project, it should
presumably have better performance and platform compatibility.
We do not want to stand in the way of progress, as ideally this
situation is a temporary state of affairs as the ecosystem grows to
support aarch64 windows. To enable progress, on aarch64 Windows builds
of uv:
* We will still use a native python interpreter, e.g., if it's at the
front of your `PATH` or the only installed version.
* If we are choosing between equally good interpreters that differ in
architecture, x64 will be preferred.
* If the aarch64 version is newer, we will prefer the aarch64 one.
* We will emit a diagnostic on installation, and show the python request
to pass to uv to force aarch64 windows to be used.
* Will be shipping [aarch64 Windows Python
downloads](https://github.com/astral-sh/python-build-standalone/pull/387)
* Will probably add some kind of global override setting/env-var to
disable this behaviour.
* Will be shipping this behaviour in
[astral-sh/setup-uv](https://github.com/astral-sh/setup-uv)
We're coordinating with Microsoft, GitHub (for the `setup-python`
action), and the CPython team (for the `python.org` installers), to
ensure we're aligned on this default and the timing of toggling to
prefer native distributions in the future.
See discussion in
- https://github.com/astral-sh/uv/issues/12906
---
This is an alternative to
* #13719
which uses sorting rather than filtering, as discussed in
* #13721
This fixes an obscure cache collision in Python interpreter queries,
which we believe to be the root cause of CI flakes we've been seeing
where a project environment is invalidated and recreated.
This work follows from the logs in [this CI
run](https://github.com/astral-sh/uv/actions/runs/15934322410/job/44950599993?pr=14326)
which captured one of the flakes with tracing enabled. There, we can see
that the project environment is invalidated because the Python
interpreter in the environment has a different version than expected:
```
DEBUG Checking for Python environment at `.venv`
TRACE Cached interpreter info for Python 3.12.9, skipping probing: .venv/bin/python3
DEBUG The interpreter in the project environment has different version (3.12.9) than it was created with (3.9.21)
```
(this message is updated to reflect #14329)
The flow is roughly:
- We create an environment with 3.12.9
- We query the environment, and cache the interpreter version for
`.venv/bin/python`
- We create an environment for 3.9.12, replacing the existing one
- We query the environment, and read the cached information
The Python cache entries are keyed by the absolute path to the
interpreter, and rely on the modification time (ctime, nsec resolution)
of the canonicalized path to determine if the cache entry should be
invalidated. The key is a hex representation of a u64 sea hasher output
— which is very unlikely to collide.
After an audit of the Python query caching logic, we determined that the
most likely cause of a collision in cache entries is that the
modification times of underlying interpreters are identical. This seems
pretty feasible, especially if the file system does not support
nanosecond precision — though it appears that the GitHub runners do
support it.
The fix here is to include the canonicalized path in the cache key,
which ensures we're looking at the modification time of the _same_
underlying interpreter.
This will "invalidate" all existing interpreter cache entries but that's
not a big deal.
This should also have the effect of reducing cache churn for
interpreters in virtual environments. Now, when you change Python
versions, we won't invalidate the previous cache entry so if you change
_back_ to the old version we can re-use our cached information.
It's a bit speculative, since we don't have a deterministic reproduction
in CI, but this is the strongest candidate given the logs and should
increase correctness regardless.
Closes https://github.com/astral-sh/uv/issues/14160
Closes https://github.com/astral-sh/uv/issues/13744
Closes https://github.com/astral-sh/uv/issues/13745
Once it's confirmed the flakes are resolved, we should revert
- https://github.com/astral-sh/uv/pull/14275
- #13817
## Summary
In #14245, we started normalizing index URLs by dropping the trailing
slash in the lockfile. We added tests to ensure that this didn't cause
existing lockfiles to be invalidated, but we missed one of the
constructors (specifically, the path that's used with
`tool.uv.sources`).
Motivated by some code duplication highlighted in
https://github.com/astral-sh/uv/pull/14201, I noticed we weren't taking
advantage of the existing implementation for casting to a str here.
Unfortunately, we do need a special case for CPython still.
Close#7426
## Summary
Picking up on #8284, I noticed that the `requires_python` object already
has its specifiers canonicalized in the `intersection` method, meaning
`~=3.12` is converted to `>=3.12, <4`. To fix this, we check and warn in
`intersection`.
## Test Plan
Used the same tests from #8284.
## Summary
When the user provides a requirement like `==2.4.*`, we desugar that to
`>=2.4.dev0,<2.5.dev0`. These bounds then appear in error messages, and
worse, they also trick the error message reporter into thinking that the
user asked for a pre-release.
This PR adds logic to convert to the more-concise `==2.4.*`
representation when possible. We could probably do a similar thing for
the compatible release operator (`~=`).
Closes https://github.com/astral-sh/uv/issues/14177.
Co-authored-by: Zanie Blue <contact@zanie.dev>
Python `bin` installations installed with `uv python install --default
--preview` (no version specified) were not being installed as
upgradeable. Instead each link was pointed at the highest patch version
for a minor version. This change ensures that these preview default
installations are also treated as upgradeable.
The PR includes some updates to the related tests. First, it checks the
default install without specified version case. Second, since it's
adding more read link checks, it creates a new `read_link` helper method
to consolidate repeated logic and replace instances of
`#[cfg(unix/windows)` with `if cfg!(unix/windows)`.
Fixes#14247
This PR updates `IndexUrl` parsing to normalize non-file URLs by
removing trailing slashes. It also normalizes registry source URLs when
using them to validate the lockfile.
Prior to this change, when writing an index URL to the lockfile, uv
would use a trailing slash if present in the provided URL and no
trailing slash otherwise. This can cause surprising behavior. For
example, `uv lock --locked` will fail when a package is added with an
`--index` value without a trailing slash and then `uv lock --locked` is
run with a `pyproject.toml` version of the index URL that contains a
trailing slash. This PR fixes this and adds a test for the scenario.
It might be safe to normalize file URLs in the same way, but since
slashes have a well-defined meaning in the context of files and
directories, I chose not to normalize them here.
Closes#13707.
uv currently ignores URL-encoded credentials in a redirect location.
This PR adds a check for these credentials to the redirect handling
logic. If found, they are moved to the Authorization header in the
redirect request.
Closes#11097
Previously we were using the XDG data dir to avoid symlinks, but there's no
particular guarantee that that's not going to be a symlink too. Using the
canonicalized temp dir by default is also slightly nicer for a couple reasons:
It's sometimes faster (an in-memory tempfs on e.g. Arch), and it makes
overriding `$TMPDIR` or `%TMP%` sufficient to control where tests put temp
files, without needing to override `UV_INTERNAL__TEST_DIR` too.
There was a regression introduced in #13954 on Windows where creating a
venv behaved as if there was a minor version link even if none existed.
This PR adds a check to fix this.
Closes#14249.
We do not currently support passing index names to `--index` for
installing packages. However, we do accept relative paths that can look
like index names. This PR adds the requirement that `--index` values
must be disambiguated with a prefix (`./` or `../` on Unix and Windows
or `.\\` or `..\\` on Windows). For now, if an ambiguous value is
provided, uv will warn that this will not be supported in the future.
Currently, if you provide an index name like `--index test` when there
is no `test` directory, uv will error with a `Directory not found...`
error. That's not very informative if you thought index names were
supported. The new warning makes the context clearer.
Closes#13921
In workspaces with multiple packages, you usually don't want to include
shared files such as the license repeatedly. Instead, we reading from
symlinked files. This would be supported if we had used std's `is_file`
and read methods, but walkdir's `is_file` does not consider symlinked
files as files.
See https://github.com/astral-sh/uv/issues/3957#issuecomment-2994675003
<!--
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
Update [schemars
0.9.0](https://github.com/GREsau/schemars/releases/tag/v0.9.0)
There are differences in the generated JSON Schema and I will [contact
the author](https://github.com/GREsau/schemars/issues/407).
## Test Plan
---------
Co-authored-by: konstin <konstin@mailbox.org>
## Summary
This flakes often and we don't really need it to be monitored
continuously. We can always revive it from Git later.
Closes https://github.com/astral-sh/uv/issues/13952.
Don't log that we resolved a reference through the GitHub fast path if
we didn't use GitHub at all but used the cached revision. This avoids
stating that the fast path works when it's blocked due to unrelated
reasons (e.g. rate limits).
@oconnor663 discovered that executing `3.10.8` on Arch Linux ran into an
error loading `libcrypt.so.1`. This caused uv to install the latest
patch version on `uv venv` operations during upgrade tests, which
undermined their purpose (since they are checking that if you first
install `3.10.8` and then upgrade, virtual environments are
transparently upgraded). This PR updates the test to use `3.10.17`
instead to avoid this issue.
#13954 introduced an unnecessary slow-down to Python uninstall by
calling `installations.find_all()` to discover remaining installations
after an uninstall. Instead, we can filter all initial installations
against those in `uninstalled`.
As part of this change, I've updated `uninstalled` from a `Vec` to an
`IndexSet` in order to do efficient lookups in the filter. This required
a change I call out below to how we were retrieving them for messaging.
We were checking whether a path was an executable in a virtual
environment or the base directory of a virtual environment in multiple
places in the codebase. This PR consolidates this logic into one place.
Closes#13947.
This PR contains the following updates:
| Package | Type | Update | Change |
|---|---|---|---|
| [mimalloc](https://redirect.github.com/purpleprotocol/mimalloc_rust) |
dependencies | patch | `0.1.46` -> `0.1.47` |
---
> [!WARNING]
> Some dependencies could not be looked up. Check the Dependency
Dashboard for more information.
---
### Release Notes
<details>
<summary>purpleprotocol/mimalloc_rust (mimalloc)</summary>
###
[`v0.1.47`](https://redirect.github.com/purpleprotocol/mimalloc_rust/releases/tag/v0.1.47):
Version 0.1.47
[Compare
Source](https://redirect.github.com/purpleprotocol/mimalloc_rust/compare/v0.1.46...v0.1.47)
##### Changes
- Mimalloc `v2.2.4`
</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:eyJjcmVhdGVkSW5WZXIiOiI0MC42Mi4xIiwidXBkYXRlZEluVmVyIjoiNDAuNjIuMSIsInRhcmdldEJyYW5jaCI6Im1haW4iLCJsYWJlbHMiOlsiaW50ZXJuYWwiXX0=-->
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>
## Summary
Allows `--torch-backend=auto` to detect AMD GPUs. The approach is fairly
well-documented inline, but I opted for `rocm_agent_enumerator` over
(e.g.) `rocminfo` since it seems to be the recommended approach for
scripting:
https://rocm.docs.amd.com/projects/rocminfo/en/latest/how-to/use-rocm-agent-enumerator.html.
Closes https://github.com/astral-sh/uv/issues/14086.
## Test Plan
```
root@rocm-jupyter-gpu-mi300x1-192gb-devcloud-atl1:~# ./uv-linux-libc-11fb582c5c046bae09766ceddd276dcc5bb41218/uv pip install torch --torch-backend=auto
Resolved 11 packages in 251ms
Prepared 2 packages in 6ms
Installed 11 packages in 257ms
+ filelock==3.18.0
+ fsspec==2025.5.1
+ jinja2==3.1.6
+ markupsafe==3.0.2
+ mpmath==1.3.0
+ networkx==3.5
+ pytorch-triton-rocm==3.3.1
+ setuptools==80.9.0
+ sympy==1.14.0
+ torch==2.7.1+rocm6.3
+ typing-extensions==4.14.0
```
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
I was looking into `uv tool` not supporting version files, and noticed
this implementation was confusing and skipped handling like a tracing
log if `--no-config` excludes selection a file. I've refactored it in
preparation for the next change.
> NOTE: The PRs that were merged into this feature branch have all been
independently reviewed. But it's also useful to see all of the changes
in their final form. I've added comments to significant changes
throughout the PR to aid discussion.
This PR introduces transparent Python version upgrades to uv, allowing
for a smoother experience when upgrading to new patch versions.
Previously, upgrading Python patch versions required manual updates to
each virtual environment. Now, virtual environments can transparently
upgrade to newer patch versions.
Due to significant changes in how uv installs and executes managed
Python executables, this functionality is initially available behind a
`--preview` flag. Once an installation has been made upgradeable through
`--preview`, subsequent operations (like `uv venv -p 3.10` or patch
upgrades) will work without requiring the flag again. This is
accomplished by checking for the existence of a minor version symlink
directory (or junction on Windows).
### Features
* New `uv python upgrade` command to upgrade installed Python versions
to the latest available patch release:
```
# Upgrade specific minor version
uv python upgrade 3.12 --preview
# Upgrade all installed minor versions
uv python upgrade --preview
```
* Transparent upgrades also occur when installing newer patch versions:
```
uv python install 3.10.8 --preview
# Automatically upgrades existing 3.10 environments
uv python install 3.10.18
```
* Support for transparently upgradeable Python `bin` installations via
`--preview` flag
```
uv python install 3.13 --preview
# Automatically upgrades the `bin` installation if there is a newer patch version available
uv python upgrade 3.13 --preview
```
* Virtual environments can still be tied to a patch version if desired
(ignoring patch upgrades):
```
uv venv -p 3.10.8
```
### Implementation
Transparent upgrades are implemented using:
* Minor version symlink directories (Unix) or junctions (Windows)
* On Windows, trampolines simulate paths with junctions
* Symlink directory naming follows Python build standalone format: e.g.,
`cpython-3.10-macos-aarch64-none`
* Upgrades are scoped to the minor version key (as represented in the
naming format: implementation-minor version+variant-os-arch-libc)
* If the context does not provide a patch version request and the
interpreter is from a managed CPython installation, the `Interpreter`
used by `uv python run` will use the full symlink directory executable
path when available, enabling transparently upgradeable environments
created with the `venv` module (`uv run python -m venv`)
New types:
* `PythonMinorVersionLink`: in a sense, the core type for this PR, this
is a representation of a minor version symlink directory (or junction on
Windows) that points to the highest installed managed CPython patch
version for a minor version key.
* `PythonInstallationMinorVersionKey`: provides a view into a
`PythonInstallationKey` that excludes the patch and prerelease. This is
used for grouping installations by minor version key (e.g., to find the
highest available patch installation for that minor version key) and for
minor version directory naming.
### Compatibility
* Supports virtual environments created with:
* `uv venv`
* `uv run python -m venv` (using managed Python that was installed or
upgraded with `--preview`)
* Virtual environments created within these environments
* Existing virtual environments from before these changes continue to
work but aren't transparently upgradeable without being recreated
* Supports both standard Python (`python3.10`) and freethreaded Python
(`python3.10t`)
* Support for transparently upgrades is currently only available for
managed CPython installations
Closes#7287Closes#7325Closes#7892Closes#9031Closes#12977
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
This PR is a combination of #12920 and #13754. Prior to these changes,
following a redirect when searching indexes would bypass our
authentication middleware. This PR updates uv to support propagating
credentials through our middleware on same-origin redirects and to
support netrc credentials for both same- and cross-origin redirects. It
does not handle the case described in #11097 where the redirect location
itself includes credentials (e.g.,
`https://user:pass@redirect-location.com`). That will be addressed in
follow-up work.
This includes unit tests for the new redirect logic and integration
tests for credential propagation. The automated external registries test
is also passing for AWS CodeArtifact, Azure Artifacts, GCP Artifact
Registry, JFrog Artifactory, GitLab, Cloudsmith, and Gemfury.
Zanie Blue
Charlie Marsh
adisbladis
Geoffrey Thomas
konsti
github-actions[bot]
Nathan Cain
Alex Prengère
Ivan Smirnov
Aria Desires
Noam Teyssier
Ben Beasley
荆长逯
Kevin Nakamura
Yu, Guangye
Nils Koch
John Mumm
Tim de Jager
Simon Sure
Jack O'Connor
Jack O'Connor
Adrien Cacciaguerra
Aaron Ang
ya7010
Christopher Tee
renovate[bot]