Mailing-List: contact dev-help@harmony.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@harmony.apache.org
Received-SPF: pass (herse.apache.org: domain of gcjhd-harmony-dev@m.gmane.org
 designates 80.91.229.2 as permitted sender)
To: dev@harmony.apache.org
From: Egor Pasko <egor.pasko@gmail.com>
Subject: Re: [general] Harmony M2 schedule
Date: 07 Jun 2007 14:35:44 +0400
Lines: 120
Message-ID: <0vqhcpkuk5b.fsf@gmail.com>
References: <906dd82e0706040123n6f821119y8a4f289ee0255aab@mail.gmail.com>
 <51d555c70706050853j227a187bt559c82749349fad3@mail.gmail.com>
 <906dd82e0706052154o77311488h631d3aa2253c373@mail.gmail.com>
 <469bff730706052334m631082fj56193d3f855692ad@mail.gmail.com>
 <0vqwsyhvaqd.fsf@gmail.com>
 <ecd185710706060249m7acba60eie1e67949eba0654c@mail.gmail.com>
 <ecd185710706060252u35977f88m62b02a59a9bfd06c@mail.gmail.com>
 <0vqsl95v1j4.fsf@gmail.com>
 <ecd185710706060529g9e3de7dq6d396fd6df2cc3b7@mail.gmail.com>
 <0vqodjtuuft.fsf@gmail.com>
 <906dd82e0706060745h29559abam57b1db14ce0b1b41@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.4
Sender: news <news@sea.gmane.org>

On the 0x2ED day of Apache Harmony Mikhail Loenko wrote:
> Hi Egor,
> 
> the point is the following: we put only those requirements for which
> there are volunteers. If for example Oleg's favorite debugger works on
> Windows he don't commit to any requirements on other platfroms.

That is not only a personal issue for volunteers to decide by
theselves, but also a strategic and political question. It remindss
me of some words of Geir that I support at 100%

Geir> Linux is a peer distro for this project.  It always has been.
Geir> 
Geir> To be an open source project that only distributes software for closed
Geir> source ecosystems like Windows is the sort of irony I'd prefer not to
Geir> be associated with :)

That's the first point.

Secondly, I can be responsible for reaching almost the same goals on
Linux (because I believe in modularity and that most of the issues are
OS-independant), BUT .. I need somebody to assist with running tests
and supporting the same kind of statistics on Linux/x86 and
Linux/x86_64 with regular updates. And creating corresponding JIRA-s
would be just great.

If there is no such volunteer, shame on you all :)

And.. I would set aside current performance issues on Linux for M2
knowing the current situation with slow thread local storage access
and probably other things.

As far as I understand, Oleg would *not* be very much burdened to run
exactly same tests on Linux as he does on Windows and collect
statictics and simplest evaluation. That does not require a debugger.

Oleg? What are your views on that? Is it difficult to run all the same
testsuites on Linux and support statistics? Are there
platform-specific things? In worst case we can exclude some unpleasant
tests from Linux runs.

> If you like to be responsible for meeting pass rates on some specific
> platform I think nobody will object to put this into requirements.
> So, do you?

I cannot run all the tests regularly. Lack of machines. Need
help. Possible?

> Thanks,
> Mikhail
> 
> 06 Jun 2007 16:41:10 +0400, Egor Pasko <egor.pasko@gmail.com>:
> > On the 0x2ED day of Apache Harmony Oleg Oleinik wrote:
> > > > I suppose, we would not be targeting this only platform. How about
> > > > others? Is the picture equally beautiful?
> > >
> > > Unfortunately, I have no information for other platforms.
> >
> > I would prefer if _stability_ targets for Linux/x86 were
> > same-or-better-than WindowsAny/x86. Does everybody feel the same?
> >
> > > On 06 Jun 2007 14:07:59 +0400, Egor Pasko <egor.pasko@gmail.com> wrote:
> > > > On the 0x2ED day of Apache Harmony Oleg Oleinik wrote:
> > > > > Clarification: DRL VM was used, Windows 2003.
> > > >
> > > > Oleg, BIG thank you!
> > > >
> > > > I suppose, we would not be targeting this only platform. How about
> > > > others? Is the picture equally beautiful?
> > > >
> > > > PS: 65% on stress tests make me worry a little :(
> > > >
> > > > > On 6/6/07, Oleg Oleinik <oleg.oleinik@gmail.com> wrote:
> > > > > > > > Thanks for capturing the requirements. IMO the requirements still lack
> > > > > > > > good tests to ensure stability and reliability. I propose to include
> > > > > > > > the following recently accepted tests:
> > > > > > > >
> > > > > > > > 1. VM validation test suite - http://issues.apache.org/jira/browse/HARMONY-3206
> > > > > > > > 2. Functional test suite - http://issues.apache.org/jira/browse/HARMONY-3528
> > > > > > > > 3. Stress tests - http://issues.apache.org/jira/browse/HARMONY-3536
> > > > > > > > 4. Reliability test - https://issues.apache.org/jira/browse/HARMONY-2918
> > > > > > > >
> > > > > > > > Olegs seems proposed a reasonable target for the tests-  98% pass rate
> > > > > > > > of all valid tests from these test suites.
> > > > > > >
> > > > > > > what is the % today? where can I find these numbers to make sure 98% is real?
> > > > > >
> > > > > > Here are some approximate numbers I have (without details with some
> > > > > > probability of mistake) to estimate how far we are from 98% pass
> > > > > > rates:
> > > > > >
> > > > > > Reliability test suite: valid 83, pass 59 (0,5 hour per test), fail
> > > > > > 24, rate: 71%
> > > > > > Functional: valid 1328, pass 1254, fail 74, rate 94%
> > > > > > Stress: valid 376, pass 245, fail 131, rate 65%
> > > > > > VM validation: valid 4346, pass 4263, fail 83, rate 98%
> > > > > >
> > > > > > TPTP profiler tests: valid 174, pass 174, rate 100%
> > > > > > Geronimo unit tests: valid 616, pass 603, fail 13, rate 97,8%
> > > > > > Eclipse unit tests (wrong test suites patched): valid 36642, pass
> > > > > > 35598, fail 1044, rate 97%
> > > > > >
> > > > > > Harmony class library tests: total ~24000, fail ~300, rate 98,8%
> > > > > >
> > > > >
> > > >
> > > > --
> > > > Egor Pasko
> > > >
> > > >
> > >
> >
> > --
> > Egor Pasko
> >
> >
> 

-- 
Egor Pasko