Apply lockfile marker normalization for fork markers by konstin · Pull Request #18116 · astral-sh/uv

konstin · 2026-02-20T13:03:04Z

While dependency markers get a roundtrip through simplify/complexify (crates/uv-resolver/src/lock/mod.rs#L4846-L4848, #15994), this treatment was missing for fork markers, causing errors with --locked --refresh on a fresh lockfile.

Fixes #16839
Closes #18024

While dependency markers get a roundtrip through simplify/complexify (https://github.com/astral-sh/uv/blob/3223b1c39f8011a4460f2b5d56ace19e5d26e16d/crates/uv-resolver/src/lock/mod.rs#L4846-L4848), this treatment was missing for fork markers, causing errors with `--locked --refresh` on a fresh lockfile. Fixes #16839 Closes #18024

konstin · 2026-02-20T13:52:49Z

@BurntSushi I tagged you for expertise on simplify/complexify; I don't have a full understanding of when we do which I one (and sometimes both in normalization pass), I've basically just copied the approach from dependency markers to fork markers here.

charliermarsh · 2026-02-24T02:38:48Z

Is it not sufficient to (1) store simplified markers and (2) ensure we always compare the simplified markers here?

…change detection Instead of normalizing fork markers via a simplify/complexify round-trip (as proposed in astral-sh#18116), store both simplified and complexified forms in a new `ForkMarkers` type. PartialEq compares the simplified forms, ensuring markers from the resolver match markers deserialized from the lockfile regardless of whether they've been through a requires-python round-trip. This follows the same pattern as `Dependency`, which stores both `simplified_marker` and `complexified_marker`. Fixes astral-sh#16839 Closes astral-sh#18024 https://claude.ai/code/session_017Wu3GxDzMrmiymfDHiXHBq

zanieb · 2026-02-24T11:18:57Z

That'd look like main...zaniebot:claude/investigate-uv-pr-H6fIZ

konstin · 2026-02-24T13:27:28Z

Is it not sufficient to (1) store simplified markers and (2) ensure we always compare the simplified markers here?

I don't follow, isn't that what this PR does?

BurntSushi

I think this LGTM in the sense that it does match what we do elsewhere.

BurntSushi · 2026-02-24T14:32:12Z

crates/uv-resolver/src/lock/mod.rs

+            .iter()
+            .map(|marker| {
+                let simplified = SimplifiedMarkerTree::new(&requires_python, marker.combined());
+                UniversalMarker::from_combined(simplified.into_marker(&requires_python))


I think what confuses me is that deserializing a lock file is already doing this step?

uv/crates/uv-resolver/src/lock/mod.rs

Lines 2279 to 2284 in 7ba5946

let fork_markers = wire

.fork_markers

.into_iter()

.map(|simplified_marker| simplified_marker.into_marker(&wire.requires_python))

.map(UniversalMarker::from_combined)

.collect();

The bug in #16839 is that we compare the resolver output to the deserialized output, and then see a mismatch, even if there was no change, as the one from the lockfile had a roundtrip through complexify/simplify, but the one from the resolver didn't.

BurntSushi · 2026-02-24T14:36:41Z

crates/uv-resolver/src/lock/mod.rs

        };
+        // Normalize fork markers with a simplify/complexify round-trip through
+        // `requires-python`. This ensures markers from the resolver (which don't include
+        // `requires-python` bounds) match markers deserialized from the lockfile (which get


My understanding/memory is that "complexified" markers are what we use inside the resolver. That is, "simplified" markers are really only a thing in the context of writing a lock file whose markers are simplified given the context of a requires-python. But those simplified markers can't really be used on their own, because they lack the constraints that would otherwise be present with requires-python.

charliermarsh · 2026-02-24T15:09:19Z

(Defer to Andrew's review.)

EliteTK

When we discussed this on Discord it seemed to make sense and it does indeed now match the other code.

This MR contains the following updates: | Package | Update | Change | |---|---|---| | [uv](https://github.com/astral-sh/uv) | patch | `0.10.4` → `0.10.6` | MR created with the help of [el-capitano/tools/renovate-bot](https://gitlab.com/el-capitano/tools/renovate-bot). **Proposed changes to behavior should be submitted there as MRs.** --- ### Release Notes <details> <summary>astral-sh/uv (uv)</summary> ### [`v0.10.6`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0106) [Compare Source](astral-sh/uv@0.10.5...0.10.6) Released on 2026-02-24. ##### Bug fixes - Apply lockfile marker normalization for fork markers ([#18116](astral-sh/uv#18116)) - Fix Python version selection for scripts with a `requires-python` conflicting with `.python-version` ([#18097](astral-sh/uv#18097)) - Preserve file permissions when using reflinks on Linux ([#18187](astral-sh/uv#18187)) ##### Documentation - Remove verbose documentation from optional dependencies help text ([#18180](astral-sh/uv#18180)) ### [`v0.10.5`](https://github.com/astral-sh/uv/blob/HEAD/CHANGELOG.md#0105) [Compare Source](astral-sh/uv@0.10.4...0.10.5) Released on 2026-02-23. ##### Enhancements - Add hint when named index is found in a parent config file ([#18087](astral-sh/uv#18087)) - Add warning for `uv lock --frozen` ([#17859](astral-sh/uv#17859)) - Attempt to use reflinks by default on Linux ([#18117](astral-sh/uv#18117)) - Fallback to hardlinks after reflink failure before copying ([#18104](astral-sh/uv#18104)) - Filter `pylock.toml` wheels by tags and `requires-python` ([#18081](astral-sh/uv#18081)) - Validate wheel filenames are normalized during `uv publish` ([#17783](astral-sh/uv#17783)) - Fix message when `exclude-newer` invalidates the lock file ([#18100](astral-sh/uv#18100)) - Change the missing files log level to debug ([#18075](astral-sh/uv#18075)) ##### Performance - Improve performance of repeated conflicts with an extra ([#18094](astral-sh/uv#18094)) ##### Bug fixes - Fix `--no-emit-workspace` with `--all-packages` on single-member workspaces ([#18098](astral-sh/uv#18098)) - Fix `UV_NO_DEFAULT_GROUPS` rejecting truthy values like `1` ([#18057](astral-sh/uv#18057)) - Fix iOS detection ([#17973](astral-sh/uv#17973)) - Propagate project-level conflicts to package extras ([#18096](astral-sh/uv#18096)) - Use a global build concurrency semaphore ([#18054](astral-sh/uv#18054)) ##### Documentation - Update documentation heading for environment variable files ([#18122](astral-sh/uv#18122)) - Fix comment about `uv export` formats ([#17900](astral-sh/uv#17900)) - Make it clear that Windows is supported in user- and system- level configuration docs ([#18106](astral-sh/uv#18106)) </details> --- ### Configuration 📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever MR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this MR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this MR, check this box --- This MR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate).

konstin requested review from BurntSushi and EliteTK February 20, 2026 13:03

konstin added the bug Something isn't working label Feb 20, 2026

zanieb approved these changes Feb 20, 2026

View reviewed changes

BurntSushi approved these changes Feb 24, 2026

View reviewed changes

EliteTK approved these changes Feb 24, 2026

View reviewed changes

konstin merged commit 55cfaf9 into main Feb 24, 2026
53 checks passed

konstin deleted the konsti/comsimplifigoration-of-fork-markers branch February 24, 2026 16:43

BrewTestBot mentioned this pull request Feb 25, 2026

uv 0.10.6 Homebrew/homebrew-core#269308

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Apply lockfile marker normalization for fork markers#18116

Apply lockfile marker normalization for fork markers#18116
konstin merged 1 commit intomainfrom
konsti/comsimplifigoration-of-fork-markers

konstin commented Feb 20, 2026

Uh oh!

konstin commented Feb 20, 2026

Uh oh!

charliermarsh commented Feb 24, 2026

Uh oh!

zanieb commented Feb 24, 2026

Uh oh!

konstin commented Feb 24, 2026

Uh oh!

BurntSushi left a comment

Uh oh!

BurntSushi Feb 24, 2026

Uh oh!

konstin Feb 24, 2026

Uh oh!

BurntSushi Feb 24, 2026

Uh oh!

charliermarsh commented Feb 24, 2026

Uh oh!

EliteTK left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

	let fork_markers = wire
	.fork_markers
	.into_iter()
	.map(\|simplified_marker\| simplified_marker.into_marker(&wire.requires_python))
	.map(UniversalMarker::from_combined)
	.collect();

Conversation

konstin commented Feb 20, 2026

Uh oh!

konstin commented Feb 20, 2026

Uh oh!

charliermarsh commented Feb 24, 2026

Uh oh!

zanieb commented Feb 24, 2026

Uh oh!

konstin commented Feb 24, 2026

Uh oh!

BurntSushi left a comment

Choose a reason for hiding this comment

Uh oh!

BurntSushi Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

konstin Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

BurntSushi Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

charliermarsh commented Feb 24, 2026

Uh oh!

EliteTK left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants