Seemingly redundant flush operation in ansicolor_sink

[ghost]

It looks like fflush is already called after each message being logged. Why does the flush policy, if enabled, trigger an additional fflush operation?

template <typename ConsoleMutex>
SPDLOG_INLINE void ansicolor_sink<ConsoleMutex>::log(const details::log_msg &msg) {
    std::lock_guard<mutex_t> lock(mutex_);
    [...]
    {
        print_range_(formatted, 0, formatted.size());
    }
    fflush(target_file_);
}

template <typename ConsoleMutex>
SPDLOG_INLINE void ansicolor_sink<ConsoleMutex>::flush() {
    std::lock_guard<mutex_t> lock(mutex_);
    fflush(target_file_);
}

[tt4g]

I did not implement ansicolor_sink, but I found the maintainer's comment: #2321 (comment)

The stdout/stderr sinks flush on each log regardless of the flush level. This is by design since normally users expect to see the screen output immediately, and also due to the fact that under windows the WriteConsoleA() flush directly to the console anyway.