Files
hermesandadmin 99f8b73654
Merge main to all / rebase (push) Failing after 31s
Python package / build (3.12.3) (push) Failing after 2m28s
Python package / build (3.13) (push) Failing after 2m29s
Python package / build (3.14) (push) Failing after 2m38s
Python package / build (3.12.3) (pull_request) Failing after 2m17s
Python package / build (3.13) (pull_request) Canceled after 55s
Python package / build (3.14) (pull_request) Canceled after 0s
feat(git): per-remote branch prefix filtering for multi-remote GitClone (#31)
## Summary

`GitClone`'s top-level `branchNamePrefix` argument was silently ignored when constructed with the explicit `remotes=[...]` form (verified in `backend/steps/git.py:106`):

```python
if remote_name == "origin" and self.branchNamePrefix != "":
    # apply refspec
```

The prefix only fires when a remote is named `"origin"`. The `remotes=` form passes names through verbatim, so configs like the open PR #28 set the prefix but never apply it — all branches from all remotes are fetched.

This PR extends the API: each remote entry now optionally accepts a `prefix` field. When set, the fetch refspec for that remote becomes scoped to `<prefix>/*` branches. Backwards-compatible — remotes without `prefix` keep the existing behavior, and `url=` shorthand with `branchNamePrefix` still works.

## API

```python
clone = GitClone(remotes=[
    {"name": "github",       "url": "...", "prefix": "ansible"},
    {"name": "gitea_hermes", "url": "...", "prefix": "ansible"},
    {"name": "gitea_admin",  "url": "...", "prefix": "ansible"},
])
```

Each remote is independent — some can have `prefix`, others not.

## Changes

### `backend/steps/git.py`

- `__init__` (`remotes=` branch): validate `prefix` (string if set, reject non-string with `ValueError`), persist into the normalized entry. Empty string / None → no prefix applied (existing behavior).
- `progress()` fetch loop: replace the single `if remote_name == "origin"` check with per-entry logic. Effective prefix = `entry["prefix"] or (self.branchNamePrefix if remote_name == "origin" else "")`. Apply refspec per remote when effective prefix is non-empty.

### `tests/test_git_integration.py`

3 new tests in `TestGitIntegration`:

- `test_git_clone_remotes_per_remote_prefix` — both remotes prefixed; only `ansible/*` branches visible, `master` is filtered out.
- `test_git_clone_remotes_mixed_prefix` — one remote prefixed, one not; each behaves independently.
- `test_git_clone_remotes_prefix_validation` — `prefix=123` / `prefix=["ansible"]` raise `ValueError`; `prefix=""` and missing-`prefix` are accepted.

Existing `test_git_clone_named_remotes` continues to pass (verified — no behavior change for unprefixed remotes).

## Test results

```
29 passed in ~13s
```

(18 in `test_git_integration.py`, 11 in `test_dummy.py`/`test_caching_step.py`/`test_cached_env.py`.)

## Migration path

After merge:

- Open migration PRs (#26 gmail_mcp, #28 hermes, planned #19 gitea.py) — switch from `branchNamePrefix="ansible"` to `"prefix": "ansible"` on each remote.
- `hermes_infra.py` after PR #8 already drops `branchNamePrefix`, no change needed.
- Existing `url=` configs (immich, gmail_mcp-old, hermes-old, registry, gitea-old) — unaffected.

## Unblocks

PR #28 — once this lands, hermes.py can drop the (currently misleading) top-level `branchNamePrefix="ansible"` and add `"prefix": "ansible"` per remote.

Refs #30
Blocks #28
Part of #18 (parent migration ticket)

---------

Co-authored-by: admin <rudolfss13@gmail.com>
Co-authored-by: brencher_bot <rudolfss13@gmail.com>
Reviewed-on: http://192.168.1.96/admin/brencher/pulls/31
Co-authored-by: hermes <rudolfss13+hermes@gmail.com>
2026-09-05 19:20:44 +02:00
..
2026-08-27 10:58:55 +02:00
2026-05-09 17:31:50 +02:00
2026-05-09 17:31:50 +02:00
2026-08-12 21:09:07 +02:00