uv add confuses dependency versions when installing multiple libraries

[ThirVondukr]

I've been migrating a project from pdm to uv and faced an error when installed dev libraries.

Environment info

uv version:

$ uv --version
uv 0.4.2 (736ccb950 2024-09-01)

OS: Windows, MacOS

Steps to reproduce

uv init

uv add --dev mypy ruff radon vulture pytest coverage factory-boy sqlalchemy-pytest types-python-dateutil jinja2 types-click deptry types-openpyxl types-aiobotocore-s3

Resolved 32 packages in 1.64s
  × No solution found when resolving dependencies:
  ╰─▶ Because only ruff<=0.6.3 is available and test-uv:dev depends on ruff>=2.9.0.20240821, we can conclude
      that test-uv:dev's requirements are unsatisfiable.
      And because your project depends on test-uv:dev, we can conclude that your project's requirements are
      unsatisfiable.
  help: If this is intentional, run `uv add --frozen` to skip the lock and sync steps.

What interesting is that in lockfile correct version of ruff is used (0.6.3),

In this particular case version 2.9.0.20240821 comes from types-python-dateutil

[erob-archim]

Adding another example in case it's helpful:

>>> uv add torchgeo torch numpy geopandas result shapely pyproj opencv-python-headless rasterio rustedpy-maybe loguru adbc-driver-postgresql scipy pyarrow
Resolved 95 packages in 0.90ms
  × No solution found when resolving dependencies for split (python_full_version < '3.11' and platform_system == 'Darwin' and sys_platform == 'linux'):
  ╰─▶ Because only loguru<=0.7.2 is available and your project depends on loguru>=4.10.0.84, we can conclude that your project's requirements are unsatisfiable.
  help: If this is intentional, run `uv add --frozen` to skip the lock and sync steps.
>>> uv add torchgeo torch numpy geopandas result shapely pyproj opencv-python-headless rasterio rustedpy-maybe adbc-driver-postgresql scipy pyarrow
Resolved 94 packages in 354ms
  × No solution found when resolving dependencies for split (python_full_version < '3.11' and platform_system == 'Darwin' and sys_platform == 'linux'):
  ╰─▶ Because only numpy<=2.1.0 is available and your project depends on numpy>=4.10.0.84, we can conclude that your project's requirements are unsatisfiable.
  help: If this is intentional, run `uv add --frozen` to skip the lock and sync steps.

Interesting that it's saying the same incorrect version number for both numpy and loguru

[charliermarsh]

Thanks, I'll look into this ASAP.

[charliermarsh]

It's a bug from #6388. I'll either fix and re-release, or revert.