Hello Steve, that's great you are working on this!

I've ran through your patch and I have the following remarks:

• Since wide chars have two bytes, there may be problem when someone wants to read or write odd number of bytes. If the number is > 1, it's ok since the code may read or write less bytes, but when the number is exactly 1, the code should maybe raise some exception.

• WriteConsoleW always fails with ERROR_NOT_ENOUGH_MEMORY (8) if we try to write more than a certain number of bytes. For me, the number is something like 41000. Unfortunately, it depends on actual heap usage of the console process. I do len = min(len, 32767) in write. The the value chosen comes from issue11395 .

• If someone types something like ^Zfoo, the standard sys.stdin returns '' -- it ignores everything after EOF if it is the first byte read. I reproduce the bahaviour in win_unicode_console to be compatible.

• There may be some issue when someone hits Ctrl-C on input. It seems that in that case, ReadConsoleW fails with ERROR_OPERATION_ABORTED (995) and some signal is asynchronously fired. It may happen that the corresponding KeyboardInterrupt exception occurs later that it should. In my Python/ctypes situation I do an ugly hack – I detect ERROR_OPERATION_ABORTED and in that case I sleep for 0.1 seconds to wait for the exception. I understand that the situation may me different in C.