Summary

This is one ((or possibly two) minor) bug(s) I frequently encounter when working on astropy, reproduced here with a minimal example.
Both of these are small sources of friction in my workflow, but I hit them frequently enough that I decided to report them any way.

Starting from the following pyproject.toml contents:

[project]
name = "test-uv"
version = "0.1.0"
requires-python = ">=3.10"
dependencies = [
    "attrs>=25.4.0",     # perhaps weirdly comment formatted comment
]

[build-system]
requires = ["uv_build>=0.9.9,<0.10.0"]
build-backend = "uv_build"

[tool.ruff]
lint.extend-select = []
exclude = []
lint.ignore = []

And using, for instance

uv add pytest

produces the following patch

diff --git a/pyproject.toml b/pyproject.toml
index a2b9bad..82d55d1 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -3,7 +3,8 @@ name = "test-uv"
 version = "0.1.0"
 requires-python = ">=3.10"
 dependencies = [
-    "attrs>=25.4.0",     # perhaps weirdly comment formatted comment
+    "attrs>=25.4.0", # perhaps weirdly comment formatted comment
+    "pytest>=9.0.1",
 ]
 
 [build-system]
@@ -12,5 +13,5 @@ build-backend = "uv_build"
 
 [tool.ruff]
 lint.extend-select = []
-exclude = []
 lint.ignore = []
+exclude = []

The problems should be evident, but let's list them anyway:

• a comment from an unrelated (albeit adjadent) line is moved
• a completely unrelated table (here, tool.ruff) is re-formatted

At least no actual data is lost, but having larger than strictly needed footprints means that I can't just commit the result of a uv add invoke, and must resort to manual editing to only commit the bits I want.

Platform

Darwin 24.6.0 arm64

Version

uv 0.9.9 (4fac4cb 2025-11-12)

Python version

Python 3.14.0 (shouldn't be relevant though)