fix(presets): use _repo_root() for bundled-core source-checkout fallback (#3086)

[mnriem]

Summary

Assessment and fix for #3086.

The issue claims specify init not copying templates/commands/ into .specify/templates/commands/ causes wrap-strategy presets to fail. That root cause is incorrect: PresetResolver has an explicit Priority‑5 fallback (_find_bundled_core, plus the matching tier‑5 branch in resolve()) specifically designed to locate the core base layer when .specify/templates/commands/ is absent. The missing copy is a layout/cosmetic gap, not a functional one.

However, while verifying, I found a real bug that does break wrap presets — but only in source/editable installs, and for a different reason:

Both fallbacks computed the repo root as Path(__file__).parent.parent.parent. After presets.py was moved to presets/__init__.py (#2826), that chain is one level short — it resolves to src/ and looks for src/templates/commands/<cmd>.md, which never exists. So the base layer was never found.

Wheel installs were unaffected because they take the core_pack branch.

Fix

Use the shared _repo_root() helper (from _assets.py) in both resolve() and _find_bundled_core() instead of a hand-rolled .parent chain, so the path stays correct regardless of module nesting.

Tests

• Added two regression tests asserting the bundled-core fallback resolves speckit.implement via both collect_all_layers() and resolve() when .specify/templates/commands/ has no matching file.
• Full preset suite (312 tests) passes.

Why not the issue's suggested fix

Copying templates/commands/ into .specify/templates/commands/ during specify init touches the core SDD install/manifest/upgrade flow to solve a non-problem (the fallback already covers wheel installs). This PR keeps the change isolated to the preset resolver and does not alter the SDD process.

Refs #3086

[Copilot]

Pull request overview

Fixes preset resolver tier-5 (“bundled core”) fallback for source/editable installs by using the shared _repo_root() helper (instead of a brittle Path(__file__).parent... chain), and adds regression tests for #3086 to ensure wrap-strategy presets can always locate the core base layer even when .specify/templates/commands/ is empty.

Changes:

• Update PresetResolver.resolve() tier-5 fallback to use specify_cli._repo_root() for repo-root template lookup in source checkouts.
• Update PresetResolver._find_bundled_core() to use the same _repo_root() helper for consistent layer collection behavior.
• Add regression tests covering tier-5 fallback behavior for collect_all_layers() and resolve() when no matching core command exists in .specify/templates/commands/.

Show a summary per file

File	Description
src/specify_cli/presets/__init__.py	Switch tier-5 repo-root fallback path computation to shared _repo_root() in both resolve() and _find_bundled_core().
tests/test_presets.py	Add regression tests for bundled-core fallback resolution paths when .specify/templates/commands/ has no matching command file.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 2/2 changed files
• Comments generated: 1

[mnriem]

Good catch — the def test_resolve_extension_command_via_manifest_skips_oserror_manifests line had been dropped, orphaning that test's body inside test_resolve_command_falls_back_to_bundled_core. Restored the definition in c52ee42; pytest now collects it again (suite goes 312 → 313, all green).

Posted on behalf of @mnriem by GitHub Copilot CLI (model: Claude Opus 4.8).

[Copilot]

Copilot's findings

• Files reviewed: 2/2 changed files
• Comments generated: 1

[mnriem]

Moved both tier-5 fallback regression tests out of TestResolveCore and into TestPresetResolver in 129514c, since they exercise collect_all_layers() and resolve() rather than resolve_core(). Suite still green (313 passing).

Posted on behalf of @mnriem by GitHub Copilot CLI (model: Claude Opus 4.8).

[Copilot]

Copilot's findings

• Files reviewed: 2/2 changed files
• Comments generated: 0 new