Repository navigation
T26 — Retained boxes managed, end to end - #3273
Conversation
One test renames a box, copies a file out of it after it stops, replays and follows its events, lists it in a prune dry run with the disk it holds, and prunes it, which ends the follow. It also restarts a long-running box under the same id and name. A failure prints each command it ran with its output and exit code.
|
Important Review skippedBot user detected. To trigger a single review, invoke the
|
|
Approved: no findings at Nothing to do. Verdict APPROVE, posted as a comment: the App cannot review its own pull request; the check and the label carry it. Clobber: none: the diff only appends. |
There was a problem hiding this comment.
arch-review: PASS · the full ruling
The test conforms to the design. The rename keeps the box id (line 487). The copy reads from the stopped box, and the events replay and follow (line 502). The dry run lists only the stopped box with its disk (line 502). The restart keeps the id and name (line 471). The test does not prove a prune event in the span, as the PR body says. The design does not require that span, so it is not a condition here.
Ruled at 212e281 against gominimal/arch@15b603e6afb8 by claude-opus-5-5.
T26 — Retained boxes managed, end to end
slice: S6
kind: integration
covers: none
size: S
est: 20 code, 100 tests
executor: agent
difficulty: routine
model: light
diagnostics: when the end-to-end test fails, it prints each command it ran with that command's stdout, stderr and exit code
observability: one span covers the rename, the copy, the events read and the prune of the same box
layers:
- tests:
- crates/minimal/src/tests.rs: the slice's end-to-end test over a test daemon
test: cargo nextest run -p minimal retained_box_management_end_to_end
produces: none
consumes: none
proof: Rename a box, prune old boxes with a dry run that shows the disk they hold, copy a file out of a stopped box, replay or follow a box's events, and restart a long-running box.
depends on:
- blocked by #2907
- blocked by #2915
- blocked by #2908
- blocked by #2909
Proof
In one test over a harness daemon, a person can see a box renamed from web to front, a file copied out of it after it stops, its events replayed and followed, and a prune dry run that lists it with the disk it holds and removes nothing. Then a prune removes it and the follow ends. A long-running box restarts under the same id and name. A failure prints each command run so far, with its stdout, stderr and exit code.
Notes from the agent: The harness daemon cannot start a box, so the restart answer is scripted through cmd::restart_box_on_hosts_with, the same as box_restart_until_stopped_restarts_same_id_and_name. The test asserts that the box rename, box cp and box events events are logged in the retained_box_end_to_end span with the box id. The prune runs inside the span too, but nothing in the log proves it: prune_hosts logs no client event on success, because cmd_box_prune logs the verb after prune_hosts returns. The daemon-side RPC parent_span_id does not help either, because it is the same inside and outside the span. If the observability line needs a logged prune event in the span, prune_hosts in crates/minimal/src/cmd/boxes/lifecycle.rs needs a tracing event, and that file is outside this task's layers.
Tests
cargo nextest run -p minimal retained_box_management_end_to_endVerify
min task run agent-verifyexited 100 under rustc 1.99.0, w 8000 hich the repository does not pin (1.99.1 in rust-toolchain.toml); the PR is not a draft on that alone — CI, which honours the pin, is the gate.Closes #2920