Files
kyleandClaude Opus 5.5 0a59dc0b22 M5c-1 plan: task 02, loopd logs the speaker and sends the turn's number
loopd copies the Turn's speaker onto the User record it logs, as it
came (not checked, not neutralised), and never into a request; after a
finished turn it sends SessionUpdated with the turn's number, the count
of User records in the session's log (Session::turns), so the first
turn is 1, a failed turn that was logged counts, and the count goes on
after a reopen or a restart. Failed turns are still not told.

Given: speaker_turn.rs (8 tests) and new versions of loopd's memory.rs,
turn_broker.rs, end_to_end.rs and support/turn.rs (run_turn gains
speaker, MemorySource::updated gains turn). Task 01 now copies
memory.rs-task01, its own version.

Checks: the reference passes make gate with 1138 tests (1130 + 8); the
task replayed from its text on a fresh ref-01 worktree with its own
target is byte-identical to the reference and passes the gate; /tmp
unchanged around both gate runs; 15 one-line mutations (speaker not
logged or dropped by the channel or neutralised, speaker in the request
at log time or at replay, turn 0-based or +1, counting Assistant
records, all records or only speakers' turns, a global or per-process
counter, told before logging or on failure, MemoryPort sending 0) all
caught; speaker_turn run 60 times under load, no failure; no
chat-template tag in the plan or given files; task lines within 100
characters but tables.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-27 23:49:35 -07:00

271 lines
8.0 KiB
Rust

//! Tests for what the turn loop does with the broker's answers: denials, pending approvals, and
//! a port that misbehaves. Do not edit.
mod support;
#[path = "support/turn.rs"]
mod turn_support;
use bxcore::{DataClass, DenyReason, LogRecord, Timestamp, ToolResponse, TurnEvent};
use support::{Reply, ReturnsPendingPort, ok_result, pending};
use turn_support::{setup, types};
const CHAT: &str = "/v1/chat/completions";
/// The tool events of a turn, in order, as short strings.
fn tool_events(events: &[TurnEvent]) -> Vec<String> {
events
.iter()
.filter_map(|e| match e {
TurnEvent::ToolCallStarted { name } => Some(format!("started {name}")),
TurnEvent::ApprovalPending { approval, tool, .. } => {
Some(format!("pending {approval} {tool}"))
}
TurnEvent::ToolDenied { name, reason } => Some(format!("denied {name} {reason:?}")),
TurnEvent::ToolResult { name, .. } => Some(format!("result {name}")),
_ => None,
})
.collect()
}
#[test]
fn a_denial_is_a_fixed_sentence_for_the_model_and_an_event_for_the_owner() {
let s = setup(vec![ToolResponse::Denied {
reason: DenyReason::TaintTooHigh,
detail: None,
}]);
s.server.route(
CHAT,
vec![Reply::fixture("tool_call"), Reply::fixture("plain")],
);
let mut session = s.session("a");
let (result, events) = s.turn(&mut session, "x");
assert!(
result.is_ok(),
"the turn goes on after a denial: {result:?}"
);
match &session.records()[4] {
LogRecord::ToolResult {
content,
class,
untrusted,
truncated,
..
} => {
assert_eq!(
content,
"Denied: this session has seen data too sensitive for this call."
);
assert_eq!(
(*class, *untrusted, *truncated),
(DataClass::Public, false, false)
);
}
other => panic!("{other:?}"),
}
assert_eq!(
tool_events(&events),
[
"started read_file",
"denied read_file TaintTooHigh",
"result read_file"
],
"the denial comes before the result"
);
// The model reads the sentence in the next request.
let m2 = s.server.requests_to(CHAT)[1].json();
assert_eq!(
m2["messages"][3]["content"],
"Denied: this session has seen data too sensitive for this call."
);
}
/// The broker's detail reaches the model, in the log and in the next request; the owner's event
/// is the same as for any denial.
#[test]
fn an_invalid_call_tells_the_model_why() {
let s = setup(vec![ToolResponse::Denied {
reason: DenyReason::InvalidArguments,
detail: Some("invalid path: n/a".to_string()),
}]);
s.server.route(
CHAT,
vec![Reply::fixture("tool_call"), Reply::fixture("plain")],
);
let mut session = s.session("a");
let (result, events) = s.turn(&mut session, "x");
assert!(result.is_ok(), "{result:?}");
let want = "Denied: the arguments are not valid for this tool: invalid path: n/a.";
match &session.records()[4] {
LogRecord::ToolResult {
content,
class,
untrusted,
truncated,
..
} => {
assert_eq!(content, want);
assert_eq!(
(*class, *untrusted, *truncated),
(DataClass::Public, false, false)
);
}
other => panic!("{other:?}"),
}
assert_eq!(
tool_events(&events),
[
"started read_file",
"denied read_file InvalidArguments",
"result read_file"
]
);
let m2 = s.server.requests_to(CHAT)[1].json();
assert_eq!(m2["messages"][3]["content"], want);
}
#[test]
fn a_denied_call_tool_names_the_target_tool_in_the_denial() {
let s = setup(vec![ToolResponse::Denied {
reason: DenyReason::NoGrant,
detail: None,
}]);
s.server.route(
CHAT,
vec![Reply::fixture("call_tool"), Reply::fixture("plain")],
);
let mut session = s.session("a");
let (result, events) = s.turn(&mut session, "echo box");
assert!(result.is_ok(), "{result:?}");
assert_eq!(
tool_events(&events),
[
"started call_tool",
"denied echo NoGrant",
"result call_tool"
],
"the owner writes grants for `echo`, not for `call_tool`"
);
assert!(
matches!(&session.records()[4], LogRecord::ToolResult { content, .. } if content == "Denied: no grant allows this call.")
);
}
#[test]
fn a_pending_approval_is_an_event_and_the_answer_after_it_is_the_result() {
let s = setup(vec![
pending(41, "2026-09-18T12:15:00.000Z"),
ok_result("straylight\n"),
]);
s.server.route(
CHAT,
vec![Reply::fixture("tool_call"), Reply::fixture("plain")],
);
let mut session = s.session("a");
let (result, events) = s.turn(&mut session, "x");
assert!(result.is_ok(), "{result:?}");
assert_eq!(
tool_events(&events),
[
"started read_file",
"pending 41 read_file",
"result read_file"
]
);
let expires = events.iter().find_map(|e| match e {
TurnEvent::ApprovalPending { expires, .. } => Some(*expires),
_ => None,
});
assert_eq!(
expires,
Some(Timestamp::parse("2026-09-18T12:15:00.000Z").unwrap())
);
assert_eq!(s.port.calls().len(), 1, "one call, however long it waited");
assert_eq!(
types(session.records()),
[
"start",
"user",
"assistant",
"usage",
"tool_result",
"assistant",
"usage"
],
"waiting writes nothing to the log"
);
assert!(
matches!(&session.records()[4], LogRecord::ToolResult { content, .. } if content == "straylight\n")
);
}
#[test]
fn a_pending_approval_that_ends_in_a_refusal() {
let s = setup(vec![
pending(7, "2026-09-18T12:15:00.000Z"),
ToolResponse::Denied {
reason: DenyReason::ApprovalRefused,
detail: None,
},
]);
s.server.route(
CHAT,
vec![Reply::fixture("tool_call"), Reply::fixture("plain")],
);
let mut session = s.session("a");
let (result, events) = s.turn(&mut session, "x");
assert!(result.is_ok(), "{result:?}");
assert_eq!(
tool_events(&events),
[
"started read_file",
"pending 7 read_file",
"denied read_file ApprovalRefused",
"result read_file"
]
);
assert!(
matches!(&session.records()[4], LogRecord::ToolResult { content, .. } if content == "Denied: the owner refused this call.")
);
}
#[test]
fn a_port_that_returns_a_pending_frame_as_its_answer_is_a_failure_not_a_decision() {
let s = setup(vec![]);
s.server.route(
CHAT,
vec![Reply::fixture("tool_call"), Reply::fixture("plain")],
);
let mut session = s.session("a");
let runtime = loopd::turn::Runtime {
cfg: &s.cfg,
client: &s.client,
port: &ReturnsPendingPort,
registry: &s.registry,
};
let mut events = Vec::new();
let result = loopd::turn::run_turn(&mut session, &runtime, "x", None, &mut |e| {
events.push(e.clone())
});
assert!(result.is_ok(), "{result:?}");
match &session.records()[4] {
LogRecord::ToolResult {
content,
class,
untrusted,
..
} => {
assert_eq!(
content,
"The tool failed: the tool broker gave no final answer"
);
assert_eq!((*class, *untrusted), (DataClass::Public, false));
}
other => panic!("{other:?}"),
}
assert_eq!(
tool_events(&events),
["started read_file", "result read_file"],
"neither pending nor denied: the port said neither"
);
}