[tontester] Add wrappers for running control queries

[DanShaders]

The only implemented control query is engine.validator.getActorTextStats. One can run it from inside a script via

actor_stats = await full_node.engine_console.get_actor_stats()

Adding support for more queries is as easy as adding them to a switch in is_engine_console_query in tonlib/tonlib/EngineConsoleClient.cpp and writing a wrapper in test/tontester/src/tonlib/engine_console.py.

[github-actions]

@codex review

[chatgpt-codex-connector]

To use Codex here, create a Codex account and connect to github.

[DanShaders]

Just an ordinary self-review, nothing to see here.

[DanShaders]

• Extra ninja -C build install recompiles files because of git watcher

[github-actions]

Findings:

• tdutils/td/utils/MpscPollableQueue.h:27 now uses std::lock_guard/std::mutex, but the header still only includes <utility> and the old td/utils/SpinLock.h. Translation units that include this header directly (for example tonlib/test/online.cpp) will see 'std::mutex' was not declared' compile errors. Please add <mutex> (and drop the unused SpinLock include) so the header is self‑contained.
• test/tonlib/test-ffi-event-loop.cpp:1-3 declares only <chrono>, <thread>, <vector> but later uses std::set (…#209). Without including <set> this test will fail to compile. Add the missing header alongside the other standard includes.
• test/tonlib/test-ffi-awaitable.cpp:13-15 defines static constexpr Tag tags[2]; and immediately takes &tags[2] for continuation_1. Dereferencing the third element of a two-element array is undefined behaviour, even if the pointer is only used for identity comparisons. Use tags + 1 or &tags[1] to stay within bounds.

[github-actions]

Findings

• tonlib_event_loop_wait flattens std::optional<Continuation> into a raw pointer and returns nullptr both for “no event” and for an actual Continuation{nullptr} (tonlib/tonlib/tonlib_engine_console.cpp:24-35). FFIEventLoop explicitly supports nullptr continuations (see the NullptrContinuation test), so any caller that ever passes a null continuation to tonlib_response_await_suspend will never see it delivered—the event is indistinguishable from a timeout/cancel and gets dropped. Consider either disallowing null continuations (and validating in tonlib_response_await_suspend) or returning an explicit status so a null continuation can still be observed.

[github-actions]

• tonlib/tonlib/EngineConsoleClient.cpp:60-69 – Possible hang on the first query: if AdnlExtClient::on_ready fires between create(...) and the ready_promise being queued, ready_ becomes true but the promise never gets resolved. The coroutine then awaits ready_awaiter indefinitely (until a disconnect triggers on_stop_ready), so initial requests can stall. Consider rechecking ready_ after creating the client and resolving the promise immediately if already ready, or registering the promise before/inside the creation path to avoid this race.

No other blocking issues found in this diff.

[github-actions]

• Missing headers break the new C++ test: test/tonlib/test-ffi-event-loop.cpp:1-3 uses std::set (line 209) and std::atomic (line 227) but only includes <chrono>, <thread>, and <vector>. Builds will fail with “std::set/atomic is not a member of std.” Add <set> and <atomic> includes.