Fix/887 self hosted provider resolution - #529
Conversation
|
665ff3d to
15c7709
Compare
|
Want your agent to iterate on Greptile's feedback? Try greploops. |
…on the same host gitHostOf reads the host of any remote or provider URL (https, scp-like or ssh:// SSH, with a user, a port or a path prefix); sameGitHost compares two of them. extractBaseUrl no longer returns a whole ssh:// remote or keeps a user, which created one CLI-managed provider per repository. Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The remote decides which providers are candidates, by host, and the vendor a CLI sends is only trusted when it sends no remote. A self-hosted remote now reaches the token provider an admin configured instead of always falling back to a CLI-managed one, and that fallback warns when a token provider of the host exists. Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…n provider Adoption compares hosts instead of sources, since the CLI records a self-hosted instance under an 'unknown' provider. A CLI-managed provider emptied by an adoption is removed, so the repository only shows under the token connection. Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…emote Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…rded Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The remote parser kept only its last two path segments, so promyze/sandbox/quentin-nuxt became sandbox/quentin-nuxt and a repository tracked from the CLI never matched the one a token connection lists. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…sted adoption spec It imported @packmind/logger, which the package does not depend on, and asserting on logger output goes against the backend test standard. Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Checking for repositories and deleting were two calls, so a repository the CLI added in between was left on a deleted provider. Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Keeping the whole group path made a remote of https://host/gitlab/group/app read as owner gitlab/group, which the provider reports as group. The owner now drops the path prefix of a provider on the remote's host, and adding a repository adopts a row the CLI recorded with the prefix, under the owner the provider reports. Refs #887 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A group read without an installation prefix now only names repositories of the connection installed under it: tracking lookups, CLI resolution, target resolution and adoption no longer strip the prefix from another host's group, nor twice from a group that starts with it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… setups - A web remote of a prefixed instance is read without the prefix, an SSH one as is; moving a tracked branch gives the owner as such a remote carries it. - A prefix naming the GitLab API root (/api/v4) is not part of the groups. - A repository is found under any CLI-managed provider of its host, as older servers kept one per SSH remote. - GitHub providers always reach github.com, whatever URL they store. - A remote naming no host is refused instead of adding a provider each call. - Tracked and known-repository lookups ignore case, as adoption records the owner as the provider spells it; per-provider readings are queried scoped. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Older CLIs send `github.com/owner/repo`; the whole-path parser refused it and the distribution notification failed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With the remote known, tracking no longer finds a repository of another host at the same path. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
215aaad to
b54d30d
Compare
0.36.1 is published without the feature, so the registry leg ran every scenario against a binary that answers `*` once pnpm's release-age window let it install that version. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|



Explanation
Relates to #
Type of Change
Affected Components
Testing
Test Details:
TODO List
Reviewer Notes