builtins.getFlake fixes

[edolstra]

Motivation

Fixes #302:

• Devirtualize strings in builtins.getFlake.
• Propagate string context in builtins.flakeRefToString.
• Force attributes in builtins.flakeRefToString to fix unevaluated thunk errors.
• Make fetchTree return a lazily computed NAR hash for backward compatibility.

Note: allowing builtins.getFlake to fetch from store paths is probably a bad idea, since it's ambiguous when using chroot stores. So it prints a warning now.

Context

Summary by CodeRabbit

• New Features

  • Added an internal NAR-hash computation and a lazy fallback for missing narHash values.

• Bug Fixes

  • Improved string/context handling for flake references and added a deprecation warning emitted during evaluation.

• Tests

  • Added a functional test covering flake reference evaluation and the new deprecation warning.

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds an internal __narHash primop to compute NAR SHA-256 (SRI format) for a path, uses it as a lazy fallback when narHash is absent during tree emission, introduces contextual string handling in flake reference evaluation, and adds a functional regression test for parent-flake access from a subflake.

Changes

Cohort / File(s)	Summary
Core NAR hash computation src/libexpr/primops.cc	Added internal __narHash primop (__narHash(p)) that realises a path, performs a dry-run fetch-to-store to compute the NAR SHA-256, returns the SRI-formatted hash as a Nix string.
Lazy hash fallback in tree emission src/libexpr/primops/fetchTree.cc	When input.getNarHash() is missing, populate narHash by invoking the internal __narHash primop on the store path alias, enabling lazy computation for backwards compatibility.
Flake reference context handling src/libflake/flake-primops.cc	Introduced NixStringContext usage and switched to forceString with context in getFlake, prim_flakeRefToString, and flakeRefToString; forces attribute values before string extraction and emits warnings if context is present.
Tests - flakes suite tests/functional/flakes/get-flake.sh, tests/functional/flakes/meson.build	Added a functional test that stages a subflake referencing its parent via builtins.getFlake/builtins.flakeRefToString, checks for a deprecation warning and correct evaluated value; added test to meson build list.

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Possibly related PRs

• Support .gitattributes in subdirectories #335 — Related changes touching NAR-hash computation and fallback usage during fetch/path handling.
• fetchTree: Implicitly set __final = true when a narHash is supplied #297 — Related modifications to fetchTree's narHash semantics and lazy-tree handling.
• Make path inputs lazy #312 — Related work on making NAR hash computation for path/tree inputs lazy.

Suggested labels

flake-regression-test

Suggested reviewers

• cole-h
• grahamc

Poem

🐰 Through lazy trees I bound and dash,
I found a missing NAR and made a hash.
A tiny primop hummed its tune,
Parent flakes now greet the moon,
Hooray — no phantom store-path crash!

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% 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
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'builtins.getFlake fixes' directly relates to the main objective of the PR, which is to fix bugs in builtins.getFlake and related flake functionality.
Linked Issues check	✅ Passed	The PR implements all required fixes: devirtualizes strings in builtins.getFlake, propagates string context in builtins.flakeRefToString, forces attributes to prevent thunk errors, and adds lazy NAR hash computation for backward compatibility.
Out of Scope Changes check	✅ Passed	All changes directly address issue #302: the primop addition, context handling, attribute forcing, and test cases are all in scope and necessary to fix the lazy-trees parent flake reference error.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix-302

---

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

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@src/libexpr/primops/fetchTree.cc`:
- Around line 33-42: The lookup key for the internal primop is wrong: replace
the call that uses get(state.internalPrimOps, "narHash") in fetchTree.cc so it
requests the registered name "__narHash" instead; ensure the code uses the same
string used when the primop was registered (internalPrimOps.emplace(primOp.name,
...)) so get(state.internalPrimOps, "__narHash") returns the correct primop for
the mkApp call that builds the lazy NAR-hash.

[github-actions]

🚀 Deployed on https://698603b4ff6bd115fd80685c--determinate-nix.netlify.app