Path: csiph.com!fu-berlin.de!uni-berlin.de!not-for-mail
From: Chris Angelico <rosuav@gmail.com>
Newsgroups: comp.lang.python
Subject: Re: nosepipe error
Date: Wed, 24 Feb 2016 05:42:15 +1100
Lines: 23
Message-ID: <mailman.78.1456252938.20994.python-list@python.org>
References: <CAGGBd_rFDSg4QpGc34vZ+8ACbfSwhrJ+iLDSAXhCvhqrAW-arQ@mail.gmail.com> <CAPTjJmp=bCtM1ZUmoy2V48GrdAHtAoZd=JZNoHcJC85JuDKDmg@mail.gmail.com> <CAGGBd_oYNsfXdmbGVv0+pAt6TWegy7ztjcGfYx7Pf-p=WeQDGw@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
In-Reply-To: <CAGGBd_oYNsfXdmbGVv0+pAt6TWegy7ztjcGfYx7Pf-p=WeQDGw@mail.gmail.com>
Precedence: list
Xref: csiph.com comp.lang.python:103413

On Wed, Feb 24, 2016 at 5:11 AM, Dan Stromberg <drsalists@gmail.com> wrote:
> I'm seeing, in the nosepipe code:
>         # we use stdout for IPC, so block all other output
>         self._stream = sys.__stdout__
>
> I'm not sure using stdout/stdin for IPC is a great thing; a lot of
> tests write to stdout or stderr, whether intentionally or
> unintentionally.

Using standard streams for IPC is well recognized. If it's known in
advance, it shouldn't be a problem - writing to stdout/stderr would
simply result in a test failure.

> I'm attempting to replace the use of stdout/stdin with os.pipe(), but
> so far I'm not having much luck.  I get three write()'s, and then a
> hang.  So far, I'm ignoring the fact that os.read() and os.write()
> might split or aggregate chunks over a socket, since this is a
> socketpair communicating on the same machine.

Sounds like the buffer's full. You'll need to have another thread
reading before you start writing, to prevent that.

ChrisA