Commit graph

3 commits

Author SHA1 Message Date
Roland
1a28a5900a
repos: git inherits only an allow-listed environment
git_env copied the whole process environment, and git consults its
environment ahead of `-c`: GIT_ALLOW_PROTOCOL=file re-enabled file://
despite protocol.allow=never, and GIT_CONFIG_COUNT/KEY_*/VALUE_* could
add an `insteadOf` that rewrites an approved URL to another host.  Only
PATH, HOME, TMPDIR, TZ and the SSL certificate paths are inherited now.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-06 17:32:59 -06:00
Roland
f49288b8ec
repos: validate branch names and repo directories
Branch names are passed to git as arguments and come from the API, the
config, a `/tree/<branch>` URL, or the remote itself (its default
branch).  A name starting with `-` would be read as an option, so every
branch must pass git's ref-name rules, including in the downloader just
before use.

A repo URL with a `.` or `..` path segment (or a `..` name) could format
a directory outside its template (e.g. `repos/<tag>/..` is the whole
repos directory).  Such URLs and names are rejected, and every repo
directory, including one from repos.yaml, must be strictly inside the
media directory with no `..`.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-06 16:56:09 -06:00
Roland
b3255e715c
Add a Repos module that mirrors git repositories (backend)
A Repo is a Collection (kind='repo') plus a `repository` row, kept up to
date by a recurring Download handled by the new `git` downloader.

* Full history by default (all branches and tags, no --prune; a force
  push saves the old HEAD under refs/wrolpi/backup/); snapshot mode
  (--depth 1) and a pinned branch are optional.  `/tree/<branch>` URLs
  select a branch.
* git runs HTTPS-only, never prompts, never runs hooks, and checks out
  symlinks as plain files.  Upstream disappearing defers the download
  with git's error; the local copy is kept.
* Repos are cloned to `repos_destination` (repos/%(repo_tag)s/%(repo_name)s)
  and the `repos` directory is always ignored, so repo files are never
  indexed.
* repos.yaml is the source of truth (import, dump, backup restore).
* /api/repos: list, create, get, edit, delete (files kept unless
  delete_files), update now, and a per-directory file tree.  Tagging uses
  the collections tag endpoint.

Also:
* run_command/process_runner can merge stderr into stdout (git writes
  progress to stderr).
* move_collection deletes an old directory that only holds empty
  directories (the move left an emptied .git behind).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-06 16:56:09 -06:00