DMARC-Filter: OpenDMARC Filter v1.3.2 sourceware.org 8126F385DC00
Date: Mon, 3 Aug 2020 12:50:04 +0200
From: Corinna Vinschen <corinna-cygwin@cygwin.com>
To: cygwin@cygwin.com
Subject: Re: Synchronization problem with posix_spawn
Message-ID: <20200803105004.GK460314@calimero.vinschen.de>
Mail-Followup-To: cygwin@cygwin.com
References: <b1992e8b-d2e8-9c44-8f93-a270d5a879ed@cornell.edu>
 <864b3031-9fc8-beb3-ba7c-1ade4c31a288@cornell.edu>
 <20200730115913.GL4206@calimero.vinschen.de>
 <20200730171723.GA460314@calimero.vinschen.de>
 <86051625-646d-065a-8543-1c3086411d3d@cornell.edu>
 <20200731081025.GB460314@calimero.vinschen.de>
 <9c44f4351d459a2ba8d27c65bf71679208cb13d6.camel@tdcadsl.dk>
MIME-Version: 1.0
Content-Disposition: inline
In-Reply-To: <9c44f4351d459a2ba8d27c65bf71679208cb13d6.camel@tdcadsl.dk>
Precedence: list
Reply-To: cygwin@cygwin.com
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Errors-To: cygwin-bounces@cygwin.com
Sender: "Cygwin" <cygwin-bounces@cygwin.com>

On Aug  3 11:10, Peter Dons Tychsen via Cygwin wrote:
> Hi all,
> 
> On Fri, 2020-07-31 at 10:10 +0200, Corinna Vinschen wrote:
> > Oh well.  I did a quick test with your new testcase (thanks for
> > that!)
> > and it seems to be a bit more complicated than I anticipated
> > yesterday.
> > The parent-child relationship between the processes is broken.  I
> > have
> > to think a while about this problem, stay tuned.
> 
> I also have seen this problem. I propose a different solution however.
> Why no get rid of the call to fork() all together. One of the things
> bogging down performance on larger setups is the calls to fork() are is
> inherently slow and should be avoided at all costs. Instead why don't
> we just call spawn(vpe) instead, which is quite stable and fast.
> 
> This would give a _huge_ boost to e.g. larger build jobs. Make supports
> posix_spawn, but it does not help on cygwin at it just calls back into
> fork().

spawn alone doesn't cut it, due to the requirement to support the
additional file actions and spawn atributes POSIX defines.  This
would require a revamp of Cygwin's spawn functionality, which is
already quite complicated.  So this is something I'm only willing
to do in homeopathic doses.


Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
--
Problem reports:      https://cygwin.com/problems.html
FAQ:                  https://cygwin.com/faq/
Documentation:        https://cygwin.com/docs.html
Unsubscribe info:     https://cygwin.com/ml/#unsubscribe-simple