Repository navigation
T9 — Tasks run and waited on, end to end - #3230
Conversation
One test drives a task through the run verbs, `min task logs -f`, `min box wait`, `min task list` and `min box prune`. A failure prints each command it ran with its stdout, stderr and exit.
|
Important Review skippedBot user detected. To trigger a single review, invoke the
|
|
Approved: every review thread resolved at @agent-137 resolved the last thread, which releases the approval this review had withheld. Nothing to do. |
There was a problem hiding this comment.
arch-review: PASS · the full ruling
The change adds one end-to-end test and no product code. Each assertion matches the design. Line 494 of architecture.md says the run returns the entrypoint's stdio and exit code. Line 485 says min box wait returns the code. Line 502 says the record stays until a prune. The record rewrite and the stand-in run follow the rulings on #3184 and #3087. CI must still run the nextest line, because the sandbox did not have cargo-nextest.
Ruled at 4260d21 against gominimal/arch@15b603e6afb8 by claude-opus-5-5.
…and-waited-on-end-to-end
The end-to-end test now files the code the run ended with, not a literal 3, and a second min box wait must return that code from the record. The comment names the minimald test that pins the daemon's filing of an entrypoint's exit code.
There was a problem hiding this comment.
arch-review: PASS · the full ruling
The new commit carries the run's exit code into the task record. A second min box wait must now return that code, and the list must show it. This matches lines 377 and 485 of architecture.md. The earlier PASS on 4260d21d stands for everything else. No conditions.
Ruled at d6bc341 against gominimal/arch@15b603e6afb8 by claude-opus-5-5.
|
Round 1 of 10 on the runtime's review findings, pushed as
Verify Converted to a draft: something above needs a person before review. Review threads the coder judged:
A person can read task_run_logs_wait_list_end_to_end and see one task run with Notes from the agent: The test line |
|
Undrafted by the BCLI babysitter. The draft comes only from the runtime's test lines failing with |
There was a problem hiding this comment.
minimal-review — approved
Verdict APPROVE — every review thread is resolved; @agent-137 closed the last one. The review of d6bc3417d3e0 withheld this approval on the threads open then. Posted as a comment: the App cannot approve its own pull request; the check and the label carry it.
T9 — Tasks run and waited on, end to end
slice: S2
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 run, the log follow and the wait of the same task
layers:
- tests:
- crates/minimal/src/tests.rs: the slice's end-to-end test over a test daemon
test: cargo nextest run -p minimal task_run_logs_wait_list_end_to_end
produces: none
consumes: none
proof: Run a task or an entry with
min task runormin box run, see its output and exit code, follow its logs, wait for it, and still see it inmin task listuntil a prune removes it.depends on:
- blocked by #2892
- blocked by #2896
- blocked by #2897
Proof
A person can run
cargo nextest run -p minimal task_run_logs_wait_list_end_to_endand see one task go through every step.min task runandmin box runboth parse to a run of the entry. The run gives back the entrypoint's stdout, stderr and exit code 3.min task logs check -fwrites the task's output in order and ends when the prune removes the task.min box wait checkreturns when the task stops, with the code its record keeps.min task listshows the task asexited (exit 3)untilmin box pruneremoves it.Notes from the agent: cargo-nextest is not installed here, so I ran the test line with
cargo test -p minimal --lib <name>. The runner must still run the nextest line as written. The run half goes through the existing stand-in SSH entrypoint (stand_in_daemon), under the task box's id. It does not use the harness daemon, because minimald's mock launcher is cfg(test) inside minimald and a test in this crate cannot start a real entrypoint. The other steps use a native harness daemon (native_box_daemon). The test rewrites the stopped task's record.json toexitedwith exit code 3, astask_list_shows_exited_until_pruneddoes, so that the list shows the code that the run returned. Diagnostics: every assertion message carries a transcript of each command run so far, with its stdout, stderr and exit. Observability: onetask_end_to_endtracing span, with the box id, wraps the run, the log follow and the wait. The spawned futures usein_current_span. rustfmt andcargo clippy -p minimal --all-targetsare clean. I did not runjust cior the whole crate suite.Tests
cargo nextest run -p minimal task_run_logs_wait_list_end_to_endVerify
min task run agent-verifyexited 101 under rustc 1.99.0, which 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 #2902