test: add coverage for hook plugin system (hooks.py)

[mnadzam]

Description

src/fromager/hooks.py has 28% unit test coverage with no dedicated unit test file (97% combined coverage including e2e tests). Hooks are the primary extension mechanism, using stevedore for plugin discovery.

The three run_*_hooks() functions have no error handling - exceptions from plugins propagate unpredictably to callers. Error handling varies by call site: bootstrapper.py wraps its call in try/except, but build.py does not.

_get_hooks() uses @with_thread_lock() for thread-safe caching of the hook manager.

Scope

Function	Lines	Callers
log_hooks()	40-56	1
run_post_build_hooks()	67-91	1
run_post_bootstrap_hooks()	94-118	1
run_prebuilt_wheel_hooks()	121-141	1
_get_hooks() (private)	23-37	3 (internal)

Acceptance Criteria

• _get_hooks() - test stevedore manager creation and caching behavior
• run_post_build_hooks() - test with mocked hooks, verify correct arguments are passed
• Exception propagation - verify behavior when a hook raises an exception
• log_hooks() - test logging output for loaded plugins

[tiran]

How are you measuring coverage? Are you combining code coverage? Most of the code should be covered by e2e test cases.

[mnadzam]

I'm sorry, I see now that I was only looking at the unit test coverage and the combined coverage is already at 97% (84% for metrics.py). Should I close the issues and PRs or do the unit tests still add some value?

[LalatenduMohanty]

@mnadzam you should add in the description that you are talking about unit test coverage. Also lets include e2e code coverage and analyze which areas lacks code coverage.

Unit tests are light weight and can run quickly where as e2e are heavier and takes more time. I will still suggest that we add unit tests if possible where the coverage is less.

[LalatenduMohanty]

Here is how last coverage report was generated

## Current Coverage Summary (2026-03-10)

**Overall: 74%** (8962 statements, 1909 missed, branch coverage enabled)

To regenerate: `source .venv/bin/activate && coverage erase && python -m coverage run -m pytest tests/ && coverage combine && coverage report`

### Per-Module Coverage

| Module | Cover |
|--------|-------|
| `commands/lint.py` | 11% |
| `vendor_rust.py` | 13% |
| `commands/build.py` | 20% |
| `commands/build_order.py` | 23% |
| `commands/find_updates.py` | 27% |
| `commands/list_versions.py` | 27% |
| `server.py` | 27% |
| `commands/download_sequence.py` | 28% |
| `hooks.py` | 28% |
| `progress.py` | 35% |
| `sources.py` | 38% |
| `gitutils.py` | 40% |
| `commands/step.py` | 42% |
| `bootstrapper.py` | 46% |
| `wheels.py` | 51% |
| `metrics.py` | 52% |
| `candidate.py` | 55% |
| `commands/graph.py` | 55% |
| `commands/canonicalize.py` | 56% |
| `__main__.py` | 61% |
| `build_environment.py` | 62% |
| `requirement_resolver.py` | 62% |
| `commands/bootstrap.py` | 63% |
| `commands/server.py` | 64% |
| `finders.py` | 64% |
| `context.py` | 66% |
| `overrides.py` | 66% |
| `clickext.py` | 70% |
| `commands/__init__.py` | 71% |
| `commands/stats.py` | 74% |
| `resolver.py` | 76% |
| `extras_provider.py` | 80% |
| `commands/minimize.py` | 81% |
| `commands/list_overrides.py` | 81% |
| `read.py` | 83% |
| `threading_utils.py` | 83% |
| `commands/lint_requirements.py` | 84% |
| `log.py` | 88% |
| `http_retry.py` | 89% |
| `dependencies.py` | 90% |
| `commands/migrate_config.py` | 91% |
| `external_commands.py` | 92% |
| `pkgmetadata/pep753.py` | 93% |
| `packagesettings.py` | 94% |
| `pyproject.py` | 96% |
| `tarballs.py` | 96% |
| `dependency_graph.py` | 96% |
| `pkgmetadata/pep639.py` | 97% |
| `constraints.py` | 98% |

[LalatenduMohanty]

The project already collects coverage from both unit tests (tests/) and e2e tests (e2e/). CI combines them in
.github/workflows/test.yaml:148-189. But there's no convenient local workflow. Which we can fix.

[mnadzam]

I updated the descriptions of both this and the metrics.py issue. I can also add the local workflow as a new issue.