Mailing-List: contact dev-help@harmony.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@harmony.apache.org
Received-SPF: pass (athena.apache.org: domain of aleksey.shipilev@gmail.com
 designates 64.233.162.226 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=received:message-id:date:from:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references;
        b=d8w0h2Q2SCo1exHsxypxUXirQkXKd61wdstv8xlv7jCrxyrJMVgVo5NbJrp+dGmtnMxXgLEyQsfsBf6vrreigDm1wY4+XqFOgohdIoQBnxMgX1Fi73A+R5CB7Wkhk/wBTlAzm9KxGFiT2Gx7x7OnchC+gXBsSigdK633qEJYjNc=
Message-ID: <4bebff790711300452l6824f2fawd09adbf4d5f0d17f@mail.gmail.com>
Date: Fri, 30 Nov 2007 15:52:18 +0300
From: "Aleksey Shipilev" <aleksey.shipilev@gmail.com>
To: dev@harmony.apache.org
Subject: Re: [classlib][nio] jetty corrupts big static files
In-Reply-To: <4bebff790711290750n51322cdbg99e76a206d019842@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
References: <2c9597b90711290606l48fd13bfj240ca335ce067b5a@mail.gmail.com>
	 <4bebff790711290611if824866k76c019f750cba0b@mail.gmail.com>
	 <4bebff790711290612r12123989mae71b4653adbb4c7@mail.gmail.com>
	 <2c9597b90711290640t207fc92du5bc350ed98f23dba@mail.gmail.com>
	 <4bebff790711290643m1a626933qf0c1bc86f1d0b207@mail.gmail.com>
	 <2c9597b90711290650s658a53c8y2fe035021a4fbf40@mail.gmail.com>
	 <4bebff790711290750n51322cdbg99e76a206d019842@mail.gmail.com>

Alexei,

It seems like your issue is not reproduced on my RHEL4 systems, while
on Debian I see exact buggy behavior you described. I'm having a quick
look now...

Thanks,
Aleksey.

On Nov 29, 2007 6:50 PM, Aleksey Shipilev <aleksey.shipilev@gmail.com> wrote:
> Alexei,
>
> I have failed to reproduce the issue with r599456 32bit build on both
> Linux32/64 with both Selector providers and Jetty 6.1.2pre1.
> Stress tests worked out fine 5 of 5 runs each with 1000 iterations.
>
> I've used:
> Apache Harmony Launcher : (c) Copyright 1991, 2007 The Apache Software
> Foundation or its licensors, as applicable.
> java version "1.5.0"
> pre-alpha : not complete or compatible
> svn = r599456, (Nov 29 2007), Linux/ia32/gcc 3.4.6, release build
> http://harmony.apache.org
>
> On:
>  - Linux linux64 2.6.9-55.ELlargesmp #1 SMP Fri Apr 20 16:46:56 EDT
> 2007 x86_64 x86_64 x86_64 GNU/Linux
>  - Linux linux32 2.6.9-55.ELsmp #1 SMP Fri Apr 20 17:03:35 EDT 2007
> i686 i686 i386 GNU/Linux
>
> Thanks,
> Aleksey.
>
>
> On Nov 29, 2007 5:50 PM, Alexei Zakharov <alexei.zakharov@gmail.com> wrote:
> > Try it on Linux64, I was able to reproduce it very easily there.
> >
> > Thanks for doing this,
> >
> > Alexei
> >
> > 2007/11/29, Aleksey Shipilev <aleksey.shipilev@gmail.com>:
> > > I have one wild thought about the origin of this issue, so I'm going
> > > to spend some time for trying.
> > > BTW, I failed to reproduce the crash even on legacy selector on my
> > > r588236 workspace, now I'm updating it to more recent one.
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Nov 29, 2007 5:40 PM, Alexei Zakharov <alexei.zakharov@gmail.com> wrote:
> > > > Hi Aleksey,
> > > >
> > > > Your provider works better than the default one - VM crash disappears.
> > > > VM runs into the dead loop instead :) BTW it would be great if you can
> > > > have a look at it...
> > > >
> > > > Thanks,
> > > > Alexei
> > > >
> > > > 2007/11/29, Aleksey Shipilev <aleksey.shipilev@gmail.com>:
> > > >
> > > > > Oh sorry, forget it, I should have read JIRA carefully.
> > > > >
> > > > > On Nov 29, 2007 5:11 PM, Aleksey Shipilev <aleksey.shipilev@gmail.com> wrote:
> > > > > > Alexei,
> > > > > >
> > > > > > Can you try epoll()-based Selector [1] as fallback? Please specify
> > > > > > this option on cmdline:
> > > > > > -Djava.nio.channels.spi.SelectorProvider=org.apache.harmony.nio.internal.EpollSelectorProviderImpl
> > > > > >
> > > > > > [1] https://issues.apache.org/jira/browse/HARMONY-4879
> > > > > >
> > > > > >
> > > > > > On Nov 29, 2007 5:06 PM, Alexei Zakharov <alexei.zakharov@gmail.com> wrote:
> > > > > > > Hi all,
> > > > > > >
> > > > > > > If you keep track of CC reports (either at [1] or by reading
> > > > > > > alerts@harmony.apache.org) then you probably noticed that we have
> > > > > > > intermittent failures of Jetty scenario on Linux'es for quite long
> > > > > > > period of time already. I've spent some time on investigation of this
> > > > > > > failure and found that Jetty-on-Harmony has issues with handling of
> > > > > > > big static files. In most cases it simply returns corrupted data.
> > > > > > > However, sometimes it hangs or even crashes. I've put the detailed
> > > > > > > description of this problem here [2]. In brief I think we have bugs in
> > > > > > > Harmony implementation of NIO selectors. Jetty works ok if its default
> > > > > > > SelectChannelConnector is replaced with any other of its connectors
> > > > > > > like SockerConnector (no NIO) or BlockingChannelConnector (NIO, but no
> > > > > > > selectors). Stack traces also point to problems somewhere in
> > > > > > > Java_org_apache_harmony_luni_platform_OSNetworkSystem_selectImpl .
> > > > > > >
> > > > > > > IMO such issue is rather critical because it prevents default
> > > > > > > configuration of Jetty from working correctly on Harmony.
> > > > > > > Unfortunately I'm not very familiar with NIO code by myself. So I
> > > > > > > think it would be nice if our NIO experts can take a look at it. And
> > > > > > > IMO it would be good to have it fixed till M4.
> > > > > > >
> > > > > > > [1] http://people.apache.org/~mloenko/snapshot_testing/script/snapshots_summary.html
> > > > > > > [2] https://issues.apache.org/jira/browse/HARMONY-5219
> > > > > > >
> > > > > > > Thanks,
> > > > > > > Alexei
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>