fix(pkg-host): the default endpoint names the store that exists #263

Merged
binjovi-bot merged 1 commit from sean/pkg-host-default-endpoint into trunk 2026-09-10 09:36:05 +00:00 AGit
Owner

defaultEndpoint still named the MinIO tenant deleted on 2026-09-10. The Deployment always sets MINIO_ENDPOINT so the constant was inert, which is what makes it a trap. The variable keeps its name -- it is the contract with the Deployment, like Forgejo's minio driver. A Go test pins the value, refuses minio and headless names, and requires it to pass newPackageServer's validation; RUN go test ./... makes that a build gate. Both controls confirmed red. VERSION 3. https://claude.ai/code/session_01KZoQin34jeyt6nDGqvJA76

defaultEndpoint still named the MinIO tenant deleted on 2026-09-10. The Deployment always sets MINIO_ENDPOINT so the constant was inert, which is what makes it a trap. The variable keeps its name -- it is the contract with the Deployment, like Forgejo's minio driver. A Go test pins the value, refuses minio and headless names, and requires it to pass newPackageServer's validation; RUN go test ./... makes that a build gate. Both controls confirmed red. VERSION 3. https://claude.ai/code/session_01KZoQin34jeyt6nDGqvJA76
fix(pkg-host): the default endpoint names the store that exists
All checks were successful
binjovi/ci Binjovi completed the frozen plan
36752a9fb0
`defaultEndpoint` still pointed at `http://minio.minio.svc.cluster.local`.
The Deployment always sets MINIO_ENDPOINT, so the constant was inert -- and
that is exactly what makes it a trap: it names a tenant deleted on
2026-09-10, so the first person to run this image without the variable gets
a connection to nothing and no hint why. Two migrations in this fleet have
already been slowed by a stale value that "could not matter".

The port is not optional: RustFS has no port-80 listener. The headless name
would be wrong for a different reason -- it hands out one server whose node
a roll can take away mid-request.

The VARIABLE keeps its name. MINIO_ENDPOINT is the contract this image has
with the Deployment that sets it, exactly as Forgejo's storage driver is
still called `minio` after the same move. Renaming it is a two-repository
change and not this one.

A Go test now pins the value, refuses any `minio` or headless name, and
requires the default to survive newPackageServer's own validation, so it
cannot be a string that merely looks right. `RUN go test ./...` in this
image's Dockerfile makes that a build gate. Both controls run and confirmed
red: the MinIO value restored, and the headless name.

VERSION 3 -- the image changed, so the version must.

Claude-Session: https://claude.ai/code/session_01KZoQin34jeyt6nDGqvJA76
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
sean/builder-images!263
No description provided.