Howdy, we encountered a bug with the

--table_delay

flag which in turn uses the

sleepFor()

function. I wanted to double-check and get your opinions before opening a bug. Specifically, the documentation says that

--table_delay

should "Add a microsecond delay between multiple table calls". In virtual_table.cpp, it's called like this

sleepFor(FLAGS_table_delay);

but then the actual

sleepFor()

function expects milliseconds - not microseconds.

inline void sleepFor(size_t msec) {
  std::chrono::milliseconds mduration(msec);
  std::this_thread::sleep_for(mduration);
}

This means that

table_delay

is off by a factor of 1000. We had this flag set to 1000 assuming it would be a 1ms delay between scans but instead it's a 1sec delay between scans. With 300 records (E.G. - processes) joined against another table (E.G. - hashes), this causes a 5min delay during query execution which hangs the scheduler pretty badly. My question is - is the bug that

sleepFor(FLAGS_table_delay);

should be

sleepFor(FLAGS_table_delay/1000);

or is the bug that

std::chrono::milliseconds

should be

std::chrono::microseconds

?