DomainKey-Signature: a=rsa-sha1; c=nofws; d=sourceware.org; h=list-id
	:list-unsubscribe:list-subscribe:list-archive:list-post
	:list-help:sender:mime-version:from:date:message-id:subject:to
	:content-type; q=dns; s=default; b=bMXHWALDAnRtuOyfNb+MCe/gOZU8R
	Kgqn7N2gN1CCMr7hnhwgZZ03k7MRBGjG0Npn3KtB2N3Gfm3THTbrMHswoIRuYoVs
	aiHHR976EBTpo4TVZ355zGthrGHg5MU6DVRFk+JBp0l5aRgHqrl/9r3evhQzTMVd
	Uz3YRoZ/0vW3Ps=
Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm
Sender: cygwin-owner AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
MIME-Version: 1.0
From: =?UTF-8?Q?Jo=C3=A3o_Eiras?= <joao DOT eiras AT gmail DOT com>
Date: Wed, 6 Jun 2018 18:23:08 +0200
Message-ID: <CAJ+sA04wDBEWrvGoXrHYbpzNePked5HEKE93kodJNM7q2TM=Jg@mail.gmail.com>
Subject: Multiple attempts at opening same pipe for writing fail
To: cygwin AT cygwin DOT com
Content-Type: text/plain; charset="UTF-8"

Howdy.

I've prepared a testcase for a bug or inconsistency with linux that is
affecting one of my projects.

Basically I have a background process that keeps some state always
running with a pipe open for reading requests.

Foreground processes that try to open that pipe for writting (to send
a request) then block/fail on the second and subsequent tries.

The whole setup can be simplified with a small testcase that runs in a
single process.

https://github.com/joaoe/cygwin-pipe-testcase

I left a README that explains things in a bit more detail.

To summarize here, consider the following order of events:
* open pipe for reading with NONBLOCK into FD1 -> OK
* poll() FD1 -> empty event list, OK
* read from FD1 -> zero length read, OK

* open pipe for writting into FD2, write something -> OK
* poll() FD1 -> POLLIN and POLLPRI, POLLPRI is unexpected here, FAIL
* read from FD1 -> got stuff that was written

* close FD2
* poll() FD1 -> POLLIN and POLLPRI, wrong, should report POLLHUP, FAIL
* read from FD1 -> zero length read, OK

* open pipe for writting into FD2 again -> got error ENXIO, FAIL

So, known bug ? Do you confirm my observations ?

Any pointers so I could perhaps look into the relevant code ? It seems
like Cygwin still keeps some dirty state somewhere that tells the pipe
is still open by a writer somewhere.

Thank you !

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple