fix(extensions): treat an unreadable staged backup as a conflict

[marcelsafin]

Description

The rescue-retry loop in install_from_directory() reads each staged backup with bare stat()/read_bytes() calls, so a staged config that cannot be read crashes the reinstall with a raw OSError. Every sibling read in this path — the live twin four lines below, the packaged baseline check, the mode sidecar — already catches OSError.

Fix: treat an unreadable staged file like an uncomparable live config: add it to the conflict set so both copies are preserved and the retry aborts with the existing resolution guidance while dest_dir is still untouched.

Testing

• Tested locally with uv run specify --help
• Ran existing tests with uv sync && uv run pytest (6,310 passed, 176 skipped)
• New regression test test_retry_with_unreadable_staged_config_aborts_and_preserves_both (fails on main with raw PermissionError, passes with fix)
• ruff check src tests clean

AI Disclosure

• I did not use AI assistance for this contribution
• I did use AI assistance (describe below)

Implemented autonomously by GitHub Copilot CLI (model: Claude Fable 5) under human direction; TDD (failing test first), full suite and lint verified locally. Commit includes Assisted-by/Co-authored-by trailers.

[Copilot]

Pull request overview

Handles unreadable staged extension backups safely during reinstall retries.

Changes:

• Converts staged-file OSError failures into preserved-config conflicts.
• Adds regression coverage verifying both copies remain intact.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/specify_cli/extensions/__init__.py	Handles staged backup read/stat failures.
tests/test_extensions.py	Tests unreadable staged backup recovery.

---

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

[Copilot]

Review details

Suppressed comments (1)

src/specify_cli/extensions/init.py:2106

• The stat() handler is reached only after _recognized_config_names(rescue_staging_dir) has already called entry.is_file() on the same staged path (lines 2014–2021). On supported Python 3.11–3.13, a PermissionError/other non-missing-path OSError from that implicit stat propagates there, so the retry still crashes before this new try; on runtimes that suppress the error, the name is omitted and the rescued config can be treated as absent. Please make the staged-name scan report matching entries whose metadata cannot be read as conflicts (and add a Path.stat regression case), rather than relying on this later stat boundary.

                try:
                    staged_stat = staged_file.stat()
                    staged_bytes = staged_file.read_bytes()
                except OSError:

• Files reviewed: 2/2 changed files
• Comments generated: 0 new
• Review effort level: Balanced

[mnriem]

Thank you!