Petr Gusev

@gleb-cloudius @nyh please re-review/merge

@nyh up

@nyh ping

@nyh fixed, please recheck/merge

> Nice, can you turn it into a unit test, so we don't regress this again when somebody decides to switch the backtrace implementation yet again? Yes, I though about...

tried to run this on ARM/debug/ASAN - no crash: ``` SEASTAR_THREAD_TEST_CASE(test_backtrace) { std::atomic reports{}; std::atomic frames{}; temporary_stall_detector_settings tsds(10ms, [&]() { ++reports; backtrace([&](auto) { ++frames; }); }); auto body = []...

copy from [a slack tread](https://scylla-db.slack.com/archives/C08FTDLLRD2/p1749638326616889?thread_ts=1749633608.044899&cid=C08FTDLLRD2): > ### Rust Driver > In the current implementation, all connection's workers: > writer > reader > keepaliver > orphaner (stream-id related thing) > are...

> Current implementation of TCP connection does not actually close local side when we call shutdown_input, it just sends eof to the reading side, the other side will not get...