M11-PT-001: wildlife steering stays within the intent length limit
Foundation / checks (checks) (push) Successful in 45m24s
Foundation / checks (deploy) (push) Successful in 4m8s
Foundation / checks (ecs) (push) Successful in 2m18s
Foundation / checks (network) (push) Successful in 19m39s
Foundation / checks (postgres) (push) Successful in 44m48s
Foundation / checks (tsan) (push) Successful in 11m26s
Foundation / checks (client) (push) Successful in 17m4s
Foundation / postgres-tools (push) Successful in 21s
Foundation / server-image (push) Successful in 3m10s
Foundation / image (push) Successful in 3m18s
Foundation / checks (checks) (push) Successful in 45m24s
Foundation / checks (deploy) (push) Successful in 4m8s
Foundation / checks (ecs) (push) Successful in 2m18s
Foundation / checks (network) (push) Successful in 19m39s
Foundation / checks (postgres) (push) Successful in 44m48s
Foundation / checks (tsan) (push) Successful in 11m26s
Foundation / checks (client) (push) Successful in 17m4s
Foundation / postgres-tools (push) Successful in 21s
Foundation / server-image (push) Successful in 3m10s
Foundation / image (push) Successful in 3m18s
DEV region 6 stopped at tick 7144 (stall log: append error=13 lives=10, region_step).
The wildlife controller steered toward food by scaling the larger component to 1000, so
a diagonal heading such as {1000, 800} was longer than a cellular intent may be; the
owner refused it, failing the tick, and the deterministic controller asked again every
tick. Earlier tests only had buds already touching food, so none moved.
wildlife_toward scales each component by the exact length, truncates toward zero and
pulls in any rounding excess. A regression test reproduces the refusal and accepts the
steered intent; another checks length, direction and speed over a grid of offsets.
No history changes: the refused ticks were never committed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
@@ -82,6 +82,23 @@ lengths can never coincide, because 40 bytes is not a whole 48-byte patch.
|
||||
This explains the reproduced case only. The first report predates d98d71e, and its
|
||||
cause remains unconfirmed.
|
||||
|
||||
### Region 6 stall (2026-10-16)
|
||||
|
||||
Region 6 stopped committing at tick 7144. It refused every later tick, and the
|
||||
returning owner again saw "Existing cell is already bound" with no view.
|
||||
|
||||
- **Evidence:** the stall log added in 44e748f printed `at=append error=13 lives=10`,
|
||||
i.e. the region step refused the tick.
|
||||
- **Cause:** the wildlife controller steered a bud toward food by scaling the larger
|
||||
component to 1000. A diagonal intent such as `{1000, 800}` is longer than a cellular
|
||||
intent may be. The owner refused it, and the deterministic controller asked again
|
||||
every tick.
|
||||
- **Why it slipped through:** the earlier tests only had buds that already touched
|
||||
food, so they never moved.
|
||||
|
||||
Steering now keeps the intent's length within the limit. A regression test reproduces
|
||||
the refusal. No history changed, because the refused ticks were never committed.
|
||||
|
||||
## Retest backlog
|
||||
|
||||
| Local issue | Priority / scope | Acceptance evidence |
|
||||
|
||||
Reference in New Issue
Block a user