test(server): add xfail repros for cancel leaving no terminal state - #1171

…cel scenarios

The two cancel-terminal-state scenarios were taken from @technicalpickles' strict-xfail repro in a2aproject#1171 and folded in here with the markers dropped; the code comment previously cited only issue a2aproject#1170, which lost that provenance. Restore the attribution in the comment and record it as a Co-authored-by trailer.

Signed-off-by: AlgoVoi <chopmob@gmail.com>
Co-authored-by: Josh Nichols <joshua.nichols@gmail.com>

…ess (a2aproject#1170)

DefaultRequestHandlerV2 could report a successful cancel while leaving the
task with no terminal state, so a client that cancelled had no way to know
it could stop polling.

Two paths reached that:

- Mid-run, ActiveTask.cancel cancelled the producer before awaiting the
  executor's cancel. The producer is the only component that writes the
  task's terminal state, so killing it first (and a cleanup-only
  executor.cancel() that writes nothing) left the task WORKING forever.
  Await the executor's cancel first, mirroring the V1 handler ordering, so
  a terminal event can still be written to the still-open event queue.

- A task parked in a non-terminal state (e.g. input-required) has no
  running producer, so cancel returned it untouched. After the task
  settles, close it out as CANCELED when it is still non-terminal, so the
  outcome is always visible to the caller, mirroring V1's guard that made
  a non-cancelled outcome visible instead of a silent no-op.

Adds two regression scenarios (converted from the xfail repros in a2aproject#1171)
covering both paths with an empty executor cancel().

Signed-off-by: AlgoVoi <chopmob@gmail.com>

…cel scenarios

The two cancel-terminal-state scenarios were taken from @technicalpickles' strict-xfail repro in a2aproject#1171 and folded in here with the markers dropped; the code comment previously cited only issue a2aproject#1170, which lost that provenance. Restore the attribution in the comment and record it as a Co-authored-by trailer.

Signed-off-by: AlgoVoi <chopmob@gmail.com>
Co-authored-by: Josh Nichols <joshua.nichols@gmail.com>
Signed-off-by: AlgoVoi <chopmob@gmail.com>

…1159, #1170) (#1172)

## Summary

Two related fixes to `DefaultRequestHandlerV2`, shipped together because
the first is a prerequisite for the second to be safe.

**Fixes #1159 (owner-scoping, the security base).** `on_cancel_task` and
`on_subscribe_to_task` resolved a live task via
`ActiveTaskRegistry.get_or_create(task_id)` by `task_id` alone, skipping
the owner-aware `task_store.get(task_id, context)` that `on_get_task`
and the send path already perform. A caller who knows another tenant's
`task_id` could subscribe to its live stream and cancel it (CWE-639).
Both handlers now consult the owner-aware store first and fail closed
with `TaskNotFoundError` (masking existence, matching `on_get_task`).

**Fixes #1170 (cancel writes a terminal state).** `ActiveTask.cancel`
cancelled the producer before `AgentExecutor.cancel`, so the component
that owns the terminal state could not write it; and a task parked
non-terminal (e.g. `input-required`) reported cancel-success without
transitioning. The executor cancel now runs before the producer cancel
(producer still cancelled on the error path), and cancellation that
leaves the task non-terminal is closed out as `CANCELED`.

## Why one PR

#1170 alone makes #1159 worse: for a parked task, a cross-tenant cancel
via the #1159 bypass changes from a harmless no-op into an actual
cross-tenant `CANCELED` write. Landing #1159's owner check together (as
the base commit) removes that window. Merging them together, or #1159
before #1170, is safe; #1170 must not land first.

## Tests

- Owner-scope regressions: non-owner cancel and non-owner subscribe are
rejected (`TaskNotFoundError`), including a parked task; the owner still
succeeds.
- Cancel-terminal regressions: mid-run cancel reaches a terminal state;
cancel of a parked task does not silently succeed (the maintainer repros
from #1171, converted from xfail to passing).

## Validation

- Owner-scope tests: reverting the #1159 handler change fails all three;
restore passes. Full request-handler module: 62 passed.
- Cancel-terminal tests: reverting the #1170 change fails both
scenario_19 tests; restore passes. Full integration module: no
regression.
- Combined: with both applied, a cross-tenant cancel of a parked task is
rejected rather than writing `CANCELED`.
- `ruff` clean at the pinned version.

---------

Signed-off-by: AlgoVoi <chopmob@gmail.com>
Co-authored-by: Josh Nichols <joshua.nichols@gmail.com>