On 5/30/2016 4:02 PM, Ken Brown wrote:
> On 5/26/2016 8:31 AM, Ken Brown wrote:
>> This looks like a bug I reported several years ago; it actually has
>> nothing to do with emacs:
>>
>>   https://www.cygwin.com/ml/cygwin/2011-10/msg00445.html
>>
>> It was fixed but then reappeared some months later:
>>
>>   https://www.cygwin.com/ml/cygwin/2012-05/msg00049.html
>>
>> It was fixed again but apparently is back.  The test case from the
>> original report still exhibits the problem.
>
> I've tracked down the commits that fixed this bug in the past.  The first fix was
>
> commit 31d2bedc585420092eb53895c5f5646651f13215
> Author: Christopher Faylor <me@cgf.cx>
> Date:   Sun Oct 23 19:01:47 2011 +0000
>
>     * fhandler_tty.cc (fhandler_pty_slave::read): Use consistent way for testing
>     ReadFile return.
>     * pipe.cc (fhandler_pipe::create_selectable): Open the write side of the pipe
>     in message-mode to force writing as "chunks".  Explain why.
>
> The second was
>
> commit fb9d631817cc2b5d83c2a6ff928851d201c992bf
> Author: Christopher Faylor <me@cgf.cx>
> Date:   Fri May 4 03:00:43 2012 +0000
>
>     * DevNotes: Add entry cgf-000002.
>     * fhandler_tty.cc (bytes_available): Revert to previous Oct-2011 behavior where
>     a dummy buffer is used to determine how many bytes will be read.
>     (fhandler_pty_master::ioctl): Correct coercion in assignment.
>
> A few weeks later the code was simplified:
>
> commit bd8afa5eb160b56715b805befe850a5ba2131d28
> Author: Christopher Faylor <me@cgf.cx>
> Date:   Wed May 16 01:56:41 2012 +0000
>
>     * DevNotes: Add entry cgf-000008.
>     * fhandler_tty.cc (bytes_available): Simplify by returning the number of bytes
>     available in the message unless that is zero.
>
> I've looked at the current code in the vicinity of these changes, and nothing jumps out at me as obviously breaking the previous fixes; but that doesn't mean much.

FWIW, if I take my original test case, as modified by cgf, and uncomment 
the line "sleep (1);", then the program finishes successfully.  I don't 
know if that provides a clue.

For convenience, I'm attaching the modified test case.

Ken