up: refuse, not just notice, an existing dev container that lacks the harness-pool runtime mount #111
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Follow-up to #102 and #107 found while reviewing the Phase 4 closure docs. On the devcontainer runtime, wraptool up requires devcontainer.json to declare the read-only runtime/ mount and refuses otherwise. But a container created before that declaration was added is reused as-is by devcontainer up, since a container's mounts are fixed at creation. Today devcontainerStaleRuntimeMountNotice in cmd/up.go only prints a notice to stderr and proceeds, so a container built against the unsupported wraptool:0 major keeps running the per-image binary with no pool generation involved, which contradicts the no-fallback rule every other path enforces. Proposal: when devcontainerRuntimeMountStale reports the live container has no mount at /opt/wraptool-harness/runtime, stop before devcontainer up with an error naming wraptool down --rm and wraptool up, the same way the missing-declaration case already does. Keep the notice text as the error body. Add a test on the stale path mirroring the missing-declaration tests. TODO.md keeps this as the one open Before 1.0 sub-item until it lands.