Re: [PATCH fifo 0/2] Add support for duplex FIFOs

[Ken Brown <kbrown at cornell dot edu>]

On 4/29/2019 3:06 PM, Corinna Vinschen wrote:
> On Apr 29 18:42, Ken Brown wrote:
>> Actually, the answer might be obvious.  Looking at MSDN, it now seems
>> clear to me that you can't do I/O on the server side of a pipe until
>> the pipe connects to a client.  So I'll have to rethink how to deal
>> with the O_RDWR case.
> 
> Sorry being late, but yeah, STATUS_PIPE_LISTENING means no client is
> connected.  This is one of the more ugly implementation details of
> Windows pipes.  There just isn't a generic buffer which can be filled
> even if no one is listening yet on the other side :-P

This should be fixed now, along with another serious bug that was exposed after 
I fixed this one.  (I had some fork/exec tests that were succeeding by accident 
because of the duplex bug.)

Ken