When a process is running and another calls `uv cache clean` or `uv
cache prune` we currently deadlock - sometimes until the CI timeout
(https://github.com/astral-sh/setup-uv/issues/588). To avoid this, we
add a default 5 min timeout waiting for a lock. 5 min balances allowing
in-progress builds to finish, especially with larger native
dependencies, while also giving timely errors for deadlocks on (remote)
systems.
Commit 1 is a refactoring.
This branch also fixes a problem with the logging where acquired and
released resources currently mismatch:
```
DEBUG Acquired lock for `https://github.com/tqdm/tqdm`
DEBUG Using existing Git source `https://github.com/tqdm/tqdm`
DEBUG Released lock at `C:\Users\Konsti\AppData\Local\uv\cache\git-v0\locks\16bb813afef8edd2`
```
Fixes https://github.com/astral-sh/uv/issues/16447
Passing this around explicitly uncovers some behaviors where we pass
e.g. the credentials store to reading the lockfile. The changes in this
PR should preserve the existing behavior for now, they only make the
locations we read from more explicit.
Labeling this PR as "Enhancement" instead of "Internal" in case this
changes behavior when it shouldn't have.
## Summary
Small nit, but I wanted to avoid doing this access in the hot path.
(Probably not important in practice.)
---------
Co-authored-by: Zanie Blue <contact@zanie.dev>
Dependabot appears to run a proxy which intercepts all requests and adds
credentials — credentials are _not_ provided via the CLI or environment
variables and there's no way for a user to do so. This means that when
`authenticate = "always"` is used (or when the index URL is on a pyx
domain), uv will fail even though Dependabot may intercept the request
and add credentials.
See
https://github.com/dependabot/dependabot-core/#private-registry-credential-management
## Summary
This PR enables users to mark a URL as an S3 endpoint, at which point uv
will sign requests to that URL by detecting credentials from the
standard AWS environment variables, configuration files, etc.
Signing is handled by the
[reqsign](https://docs.rs/reqsign/latest/reqsign/) crate, which we can
also use in the future to sign requests for other providers.
## Summary
I spent time trying to figure out how to support this but came up empty.
It _seems_ like maybe the `DefaultFields` implementation in
`tracing-subscriber` uses debug formatting for fields...? So if you have
a string with ANSI codes, they end up printing as unformatted values? I
even reverted all our custom formatting in `logging.rs` and saw the same
thing.
Closes https://github.com/astral-sh/uv/issues/15840.
## Summary
This PR adds support for pyx to `uv auth login`, `uv auth logout`, and
`uv auth token`. These are generic uv commands that can be used to store
credentials for arbitrary indexes and other URLs, but we include a
fast-path for pyx that initiates the appropriate login or logout flow.
We're not sure what the best way to expose the native store to users is
yet and it's a bit weird that you can use this in the `uv auth` commands
but can't use any of the other keyring provider options. The simplest
path forward is to just not expose it to users as a keyring provider,
and instead frame it as a preview alternative to the plaintext uv
credentials store. We can revisit the best way to expose configuration
before stabilization.
Note this pull request retains the _internal_ keyring provider
implementation — we can refactor it out later but I wanted to avoid a
bunch of churn here.
Adds locking of the credentials store for concurrency safety. It's
important to hold the lock from read -> write so credentials are not
dropped during concurrent writes.
I opted not to attach the lock to the store itself. Instead, I return
the lock on read and require it on write to encourage safe use. Maybe
attaching the source path to the store struct and adding a `lock(&self)`
method would make sense? but then you can forget to take the lock at the
right time. The main problem with the interface here is to write a _new_
store you have to take the lock yourself, and you could make a mistake
by taking a lock for the wrong path or something. The fix for that would
be to introduce a new `CredentialStoreHandle` type or something, but
that seems overzealous rn. We also don't eagerly drop the lock on token
read, although we could.
Adds a default plain text storage mechanism to `uv auth`.
While we'd prefer to use the system store, the "native" keyring support
is experimental still and I don't want to ship an unusable interface.
@geofft also suggested that the story for secure credential storage is
much weaker on Linux than macOS and Windows and felt this approach would
be needed regardless.
We'll switch over to using the native keyring by default in the future.
On Linux, we can now fallback to a plaintext store the secret store is
not configured, which is a nice property.
Right now, we store credentials in a TOML file in the uv state
directory. I expect to also read from the uv config directory in the
future, but we don't need it immediately.
Picks up the work from
- #14559
- https://github.com/astral-sh/uv/pull/14896
There are some high-level changes from those pull requests
1. We do not stash seen credentials in the keyring automatically
2. We use `auth login` and `auth logout` (for future consistency)
3. We add a `token` command for showing the credential that will be used
As well as many smaller changes to API, messaging, testing, etc.
---------
Co-authored-by: John Mumm <jtfmumm@gmail.com>
## Summary
Make the use of `Self` consistent. Mostly done by running `cargo clippy
--fix -- -A clippy::all -W clippy::use_self`.
## Test Plan
<!-- How was it tested? -->
No need.
## Summary
If `HF_TOKEN` is set, we'll automatically wire it up to authenticate
requests when hitting private `huggingface.co` URLs in `uv run`.
## Test Plan
An unauthenticated request:
```
> cargo run -- run https://huggingface.co/datasets/cmarsh/test/resolve/main/main.py
File "/var/folders/nt/6gf2v7_s3k13zq_t3944rwz40000gn/T/mainYadr5M.py", line 1
Invalid username or password.
^^^^^^^^
SyntaxError: invalid syntax
```
An authenticated request:
```
> HF_TOKEN=hf_... cargo run run https://huggingface.co/datasets/cmarsh/test/resolve/main/main.py
Hello from main.py!
```
By default, Rust does not support safe cast from `&U` to `&T` for
`#[repr(transparent)] T(U)` even if the newtype opts in. The dtolnay
ref-cast crate fills this gap, allowing to remove `DisplaySafeUrlRef`.
This PR implements a few review follow-ups from #13560. In particular,
it
* Makes `DisplaySafeUrlRef` implement `Copy` so that it can be passed by
value.
* Updates `to_string_with_credentials` methods with
`displayable_with_credentials`, returning an `impl Display` instead of
`String` for greater flexibility.
* Updates the `DisplaySafeUrl` and `DisplaySafeUrlRef` `Debug`
implementations to match the underlying `Url` `Debug` implementation
(with the exception that credentials are masked).
* Replaces an unnecessary `DisplaySafeUrl::from(Url::from_file_path`
with `DisplaySafeUrl::from_file_path`
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.
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
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
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
## 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.
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.
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.
Adds a new optional key `auth-policy` to `[tool.uv.index]` that sets the
authentication policy for the index URL.
The default is `"auto"`, which attempts to authenticate when necessary.
`"always"` always attempts to authenticate and fails if the endpoint is
unauthenticated. `"never"` never attempts to authenticate.
These policy address two kinds of cases:
* Some indexes don’t fail on unauthenticated requests; instead they just
forward to the public PyPI. This can leave the user confused as to why
their package is missing. The "always" policy prevents this.
* "never" allows users to ensure their credentials couldn't be leaked to
an unexpected index, though it will only allow for successful requests
on an index that doesn't require credentials.
Closes#11600
## Summary
These were moved as part of a broader refactor to create a single
integration test module. That "single integration test module" did
indeed have a big impact on compile times, which is great! But we aren't
seeing any benefit from moving these tests into their own files (despite
the claim in [this blog
post](https://matklad.github.io/2021/02/27/delete-cargo-integration-tests.html),
I see the same compilation pattern regardless of where the tests are
located). Plus, we don't have many of these, and same-file tests is such
a strong Rust convention.
## Summary
Resolves#7685
## Test Plan
```console
$ echo "this is an invalid netrc" > .netrc
$ NETRC=.netrc cargo run -- pip install anyio --index-url https://pypi-proxy.fly.dev/basic-auth/simple --strict -v
DEBUG uv 0.4.24 (f4d5fba61 2024-10-19)
DEBUG Searching for default Python interpreter in system path or `py` launcher
DEBUG Found `cpython-3.11.2-windows-x86_64-none` at `D:\Projects\Rust\uv\.venv\Scripts\python.exe` (virtual environment)
DEBUG Using Python 3.11.2 environment at .venv
DEBUG Acquired lock for `.venv`
DEBUG At least one requirement is not satisfied: anyio
DEBUG Using request timeout of 30s
DEBUG Solving with installed Python version: 3.11.2
DEBUG Solving with target Python version: >=3.11.2
DEBUG Adding direct dependency: anyio*
DEBUG No cache entry for: https://pypi-proxy.fly.dev/basic-auth/simple/anyio/
WARN Error reading netrc file: parsing error: bad toplevel token 'this' (line 1) in the file '.netrc'
DEBUG Searching for a compatible version of anyio (*)
DEBUG No compatible version found for: anyio
× No solution found when resolving dependencies:
╰─▶ Because anyio was not found in the package registry and you require anyio, we can conclude that your
requirements are unsatisfiable.
hint: An index URL (https://pypi-proxy.fly.dev/basic-auth/simple) could not be queried due to a lack of valid
authentication credentials (401 Unauthorized).
DEBUG Released lock at `D:\Projects\Rust\uv\.venv\.lock`
error: process didn't exit successfully: `target\debug\uv.exe pip install anyio --index-url https://pypi-proxy.fly.dev/basic-auth/simple --strict -v` (exit code: 1)
```
Basically, if username-only authentication came from the _cache_ instead
of being present on the _request URL_ to start, we'd end up ignoring it
during password lookups which breaks keyring.
Includes some cosmetic changes to the logging and commentary in the
middleware, because I was confused when reading the code and logs.
As per
https://matklad.github.io/2021/02/27/delete-cargo-integration-tests.html
Before that, there were 91 separate integration tests binary.
(As discussed on Discord — I've done the `uv` crate, there's still a few
more commits coming before this is mergeable, and I want to see how it
performs in CI and locally).
While working on https://github.com/astral-sh/uv/pull/6389 I discovered
we never checked `cache.get_url` here, which is wrong — though I don't
think it had much effect in practice since the realm would typically
match first. The main problem is that when we call `get_url` later we
hard-code the username to `None` because we assume we checked up here
with the username if present.
Partially addresses https://github.com/astral-sh/uv/issues/4056
We were incorrectly omitting the port from requests to `keyring` when
falling back to a realm/host query, e.g. `localhost` was used instead of
`localhost:1234`. We still won't include "standard" ports like `80` for
an HTTP request.
konsti
Charlie Marsh
Zanie Blue
William Woodruff
adamnemecek
John Mumm
samypr100
Jo
Amos Wenger
Ed Morley
renovate[bot]