Repository navigation
Merge remote-tracking branch 'origin/main' into agent/2894-boxes-list…
PASS
Baseline. This is a re-review of dbe84265 against the PASS on a40b9310 (this PR), on the same arch commit 7a86ec7b. I checked these facts. since.diff eq
3693
uals git diff a40b9310 dbe84265. pr.diff equals git diff 014fa3eb dbe84265, and 014fa3eb (#2985) is the merge base. The compare has one branch commit, 652fe117, and one merge, dbe84265, which brings #2985 from main. The merge is clean (git diff-tree --cc is empty). So the crates/minimald/src/net/answerer.rs hunks in since.diff belong to #2985 and are not part of this PR. pr.diff touches only crates/minimal/src/tests.rs.
Governs: architecture.md Glossary line 53 (for local, min is the provider and reads each native host over its own socket). min box lines 474-475 and 485 (the list, show, stop and rm grammar, and the fields show renders, type among them). Type nouns line 458 (stop keeps the record, rm deletes it). Line 499 (stopped is one of the five record states). Global flags line 350 (--force overrides the refusal to rm a running box).
Compare. (1) The test now asserts the TYPE cell of the listing and the Type: line of show. Both equal boxes::SESSION_BOX_TYPE. render_box_list (cmd/boxes/list.rs) and render_box_show (cmd/boxes/show.rs) print record.box_type, so the asserts check what line 475 lists. (2) The step for the second rm no longer writes Removed web. itself. It reuses the rm closure and asserts empty stdout, which matches the doc comment on rm_listed_box (cmd/boxes/lifecycle.rs). cmd_box_rm writes the line after its cleanup. This change makes the proof more honest (rubric rule 3): the earlier head asserted a line that the test itself wrote. The design does not mention that message, so its absence from the proof is not a deviation. (3) The span check now requires one box_verb{verb, box_id} span with its min box <verb> ran event for each run of a verb. Before, it required only at least one. This matches the task's observability: line more closely.
Ruling. Everything from the baseline still conforms. The compare only makes the asserts stricter.
Not checked. I did not compile or run the test, because the contract forbids a build. rm is passed by value two times, as show was already on the earlier head. Whether the compiler accepts that is a build fact, not a design fact. The PR body reports the targeted test as exit 0.
Ruled at dbe8426 against gominimal/arch@7a86ec7bb3c0 by claude-opus-5-5; arch_rulings 865, job df768bd9-adba-42e9-b726-2e3bf3e10f87.