[jdx]

Summary

• resolve Aqua install metadata with package- and override-prefixed version candidates when no exact release tag is available
• preserve exact release-tag selection when a GitHub lock URL provides one
• add regression coverage for Cabal-style locked HTTP packages whose archive format comes from a prefix-scoped override

Root cause

Locked HTTP URLs such as Cabal's downloads.haskell.org artifacts do not contain a GitHub release tag, and mise skips tag discovery when a lockfile URL already exists. After prefix-aware Aqua override matching was introduced, install resolution still supplied only the bare version and its v variant. Cabal's cabal-install-v override therefore did not match, its format: tar.xz metadata was lost, and the archive was copied as the executable.

Fixes the regression reported in #11494.

Impact

Locked Aqua HTTP installs retain prefix-scoped format and file metadata, so archives are extracted correctly. Packages with an exact release tag continue to use only that tag, preserving disambiguation between multiple prefix families.

Validation

• cargo test backend::aqua::tests::test_package_version_candidates -- --nocapture
• cargo test backend::aqua::lock_candidate_tests:: -- --nocapture
• mise run lint-fix

AI-assisted — Tool: Codex; model: unavailable/unavailable; version: unavailable.

---

Note

Medium Risk
Changes install-time Aqua package/override resolution for locked installs; wrong candidate logic could mis-apply formats or overrides, but scope is narrow and covered by new regression tests.

Overview
Fixes Aqua installs where a lockfile URL exists but carries no GitHub release tag (e.g. Cabal on downloads.haskell.org). Install resolution no longer passes only the bare version and optional v variant into override matching; it uses install_package_version_candidates, which delegates to package_version_candidates so registry version_prefix values (like cabal-install-v) still match and inherit metadata such as format: tar.xz.

When the lock URL does encode a GitHub release tag, resolution still uses only that tag, avoiding ambiguity between prefix families. The redundant async package_with_options helper is removed in favor of loading the package once and calling package_with_options_for_pkg directly. A Cabal-style HTTP regression test asserts prefix and format are applied.

^{Reviewed by Cursor Bugbot for commit 4708fe7. Bugbot is set up for automated code reviews on this repo. Configure here.}

Summary by CodeRabbit

• Bug Fixes
  • Improved installation handling for packages that use version prefixes and format overrides.
  • Preserved inherited version metadata for HTTP packages during installs, including when a release tag is not available.
  • Corrected install-time version resolution to use the appropriate registry-derived candidate set for overrides.
• Tests
  • Updated coverage to validate inherited version-prefix behavior with install-time options.

[coderabbitai]

📝 Walkthrough

Walkthrough

Aqua installation now fetches the registry package directly, derives install-time version candidates through a dedicated helper, and applies options with package_with_options_for_pkg. A regression test covers inherited version_prefix and overridden format metadata for HTTP packages.

Changes

Aqua install resolution

Layer / File(s)	Summary
Install candidate derivation src/backend/aqua.rs	install_package_version_candidates selects the lock tag when present or registry-derived version-prefix candidates otherwise; a unit test covers inherited prefixes and formats.
Install option application src/backend/aqua.rs	install_version_ resolves the registry package and applies options through package_with_options_for_pkg; the previous helper is removed.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• jdx/mise#11267: Preserves v-prefixed HTTP lock URL resolution through shared version-candidate handling.
• jdx/mise#11482: Refactors Aqua version-candidate resolution and preserves version_prefix metadata.

Suggested reviewers: risu729

Poem

I’m a rabbit, hopping through tags,
Prefixes tucked in version bags.
HTTP paths now bloom just right,
Formats carried through the night.
Aqua installs, neat and bright!

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 37.50% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly matches the main change: preserving locked HTTP override metadata in Aqua installs.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[greptile-apps]

Greptile Summary

Preserves prefix-scoped Aqua metadata for locked HTTP artifacts while retaining exact-tag selection for GitHub release URLs.

• Uses registry-aware version candidates when a locked URL has no GitHub release tag.
• Keeps exact release tags as the sole candidate when one can be recovered from the URL.
• Adds regression coverage for inherited archive-format overrides.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
src/backend/aqua.rs	Updates install-time Aqua override resolution and adds focused regression coverage; no eligible follow-up defect was identified.

_{Reviews (2): Last reviewed commit: "test(aqua): cover locked HTTP install op..." | Re-trigger Greptile}

[coderabbitai]

🧹 Nitpick comments (1)

src/backend/aqua.rs (1)

3690-3714: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Cover the install wiring, not only candidate computation.

This test exercises install_package_version_candidates and AquaPackage::with_version directly, but not package_with_options_for_pkg from install_version_. Add a focused boundary or locked-HTTP integration test asserting the resolved version_prefix and format.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/backend/aqua.rs` around lines 3690 - 3714, The existing test only
verifies candidate generation and AquaPackage construction; add coverage through
the install wiring entry point package_with_options_for_pkg in install_version_.
Create a focused boundary or locked-HTTP integration test using the inherited
prefix and format override, then assert that the resolved package options
preserve version_prefix and format.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@src/backend/aqua.rs`:
- Around line 3690-3714: The existing test only verifies candidate generation
and AquaPackage construction; add coverage through the install wiring entry
point package_with_options_for_pkg in install_version_. Create a focused
boundary or locked-HTTP integration test using the inherited prefix and format
override, then assert that the resolved package options preserve version_prefix
and format.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Central YAML (base), Organization UI (inherited)

Review profile: CHILL

Plan: Pro Plus

Run ID: e3295d91-ce74-4f1c-b9a0-2ff99ad7231e

📥 Commits

Reviewing files that changed from the base of the PR and between ee3e917 and bc5b102.

📒 Files selected for processing (1)

• src/backend/aqua.rs

[github-actions]

Instruction counts

benchmark	trend	instructions	Δ	wall (min)	Δ
env	▇█▂▄▃▁▂▇▃▃▄▃▄▄▁▅▂▂▂	84,635,676 → 84,631,665	-0.00%	32.92 → 21.45ms	-34.84%
hook-env	▄▄▁▁▅▄▆▆▆▇▇▆▇█▅▇▂▄▄	87,290,603 → 87,291,524	+0.00%	52.06 → 24.20ms	-53.52%
ls	▇▇▃▅▇▆▆█▆▄▇▄▆▇▂█▁▅▆	76,725,097 → 76,754,311	+0.04%	22.42 → 22.01ms	-1.86%
registry	▅▄▁▂▂▂▅▅▅▄▅▃▅▅▄█▁▂▅	50,181,864 → 50,237,269	+0.11%	16.35 → 14.00ms	-14.39%
startup	▇█▃▅▅▄▆█▆▅▅▁▅▆▅█▃█▅	16,089,749 → 16,068,355	-0.13%	8.52 → 10.11ms	+18.63%

No instruction-count regression above 1%.

_{Only instruction counts gate. Wall clock is shown for context — on identical hardware it moves 4-20% run to run.}

_{Measured by tak — instruction-counted CLI benchmarks, stored in this repository's git notes.}

_{4708fe78a99f vs ee3e9171b164 · measured on the runner, not pushed to the history.}