You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by Tony Wu <wu...@gmail.com> on 2008/01/04 03:23:12 UTC

Re: [classlib]remove the duplicate locale data

Hi,

I assume this patch is ok finally if no more reponse :)

On 12/26/07, Tony Wu <wu...@gmail.com> wrote:
> On 12/22/07, Alexey Petrenko <al...@gmail.com> wrote:
> > Tony, can we fix all the issues with ICU in the branch before
> > committing these changes to trunk?
>
> Thanks for your suggestion, I think I've got almost done on my
> machine.  I'll consider about branching when I face a big task again
> :)
>
> >
> > SY, Alexey
> >
> > 2007/12/21, Aleksey Shipilev <al...@gmail.com>:
> > > As I've concerned, SPECjbb2005 degradation is not gone: it is still 7%.
> > > I think we shouldn't commit this patch in main trunk.
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Dec 21, 2007 5:50 PM, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > Tony,
> > > >
> > > > I think it's better to check new patch performance before committing.
> > > > I'm going to check it in next several hours.
> > > >
> > > > Thanks,
> > > > Aleksey,
> > > > ESSD, Intel.
> > > >
> > > >
> > > > On Dec 21, 2007 12:01 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > Hi all,
> > > > > I'd like to recommit the patch if no one objects.
> > > > >
> > > > > My new patch on 5061 fixes
> > > > > 1. the performance issue addressed in 5129
> > > > > 2. the return type gap on 5085
> > > > > 3. the testcases in text module, now they can pass on both RI and Harmony
> > > > >
> > > > > Still this patch willl exclude following tests in luni
> > > > > +tests/api/java/util/CurrencyTest.java
> > > > > +tests/api/java/util/FormatterTest.java
> > > > > +tests/api/java/util/GregorianCalendarTest.java
> > > > > +tests/api/java/util/LocaleTest.java
> > > > > +tests/api/java/util/ScannerTest.java
> > > > >
> > > > > In total there are 1 error and 13 failures, all of which are caused by
> > > > > differences between RI and ICU. I'll record them as non-bug difference
> > > > > later.
> > > > >
> > > > >
> > > > > On 11/12/07, Vladimir Strigun <vs...@gmail.com> wrote:
> > > > > > Thanks Mark,
> > > > > >
> > > > > > All dacapo benches passed after your commit.
> > > > > >
> > > > > > Thanks.
> > > > > > Vladimir.
> > > > > >
> > > > > > On 11/12/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > I noticed Mark has committed your patch, Thank you and Mark :)
> > > > > > >
> > > > > > > On 11/10/07, Vladimir Strigun <vs...@gmail.com> wrote:
> > > > > > > > On 11/9/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > Hi Vladimir
> > > > > > > > > really busy this weekend. I've recorded it as JIRA 5100, high
> > > > > > > > > appreciate if you could provide a simple testcase.
> > > > > > > >
> > > > > > > > Hi Tony,
> > > > > > > >
> > > > > > > > It seems we've faced with first isues in new ICU version. ICU returns
> > > > > > > > null instead of correct string for time zone. Could you please check
> > > > > > > > the patch attached to JIRA and apply it to fix the regression?
> > > > > > > >
> > > > > > > > Thanks.
> > > > > > > > Vladimir.
> > > > > > > >
> > > > > > > > > On 11/7/07, Vladimir Strigun <vs...@gmail.com> wrote:
> > > > > > > > > > Hi Tony,
> > > > > > > > > >
> > > > > > > > > > It seems your commit r592434 introduce regression for Dacapo.jython
> > > > > > > > > > (the bench passed on r592433). Here is the execution log:
> > > > > > > > > > ===== DaCapo jython starting =====
> > > > > > > > > > -------------------------------------------------------------------------------
> > > > > > > > > > PYBENCH 2.0
> > > > > > > > > > -------------------------------------------------------------------------------
> > > > > > > > > > * using Python 2.2a1
> > > > > > > > > > * Python version doesn't support garbage collection
> > > > > > > > > > * system check interval set to maximum: 2147483647
> > > > > > > > > > * using timer: time.time
> > > > > > > > > >
> > > > > > > > > > Traceback (innermost last):
> > > > > > > > > >  File ".\scratch\jython\pybench\pybench.py", line 946, in ?
> > > > > > > > > >  File "Z:\UBS\Storage\.\scratch\jython\pybench\CommandLine.py", line
> > > > > > > > > > 346, in __init__
> > > > > > > > > >  File ".\scratch\jython\pybench\pybench.py", line 903, in main
> > > > > > > > > >  File ".\scratch\jython\pybench\pybench.py", line 425, in __init__
> > > > > > > > > > java.lang.IllegalArgumentException: Cannot create PyString from null!
> > > > > > > > > >        at org.python.core.PyString.<init>(Unknown Source)
> > > > > > > > > >        at org.python.modules.time.classDictInit(Unknown Source)
> > > > > > > > > >        at java.lang.reflect.VMReflection.invokeMethod(VMReflection.java)
> > > > > > > > > >        at java.lang.reflect.Method.invoke(Method.java:317)
> > > > > > > > > >        at org.python.core.PyJavaClass.initialize(Unknown Source)
> > > > > > > > > >        at org.python.core.PyJavaClass.lookupGivingClass(Unknown Source)
> > > > > > > > > >        at org.python.core.PyClass.lookup(Unknown Source)
> > > > > > > > > >        at org.python.core.PyJavaClass.__findattr__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyObject.__getattr__(Unknown Source)
> > > > > > > > > >        at org.python.pycode._pyx1.__init__$14(.\scratch\jython\pybench\pybench.py:425)
> > > > > > > > > >        at org.python.pycode._pyx1.call_function(.\scratch\jython\pybench\pybench.py)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyFunction.__call__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyInstance.__init__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyClass.__call__(Unknown Source)
> > > > > > > > > >        at org.python.pycode._pyx1.main$25(.\scratch\jython\pybench\pybench.py:903)
> > > > > > > > > >        at org.python.pycode._pyx1.call_function(.\scratch\jython\pybench\pybench.py)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyFunction.__call__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyMethod.__call__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyObject.__call__(Unknown Source)
> > > > > > > > > >        at CommandLine$py.__init__$15(Z:\UBS\Storage\.\scratch\jython\pybench\CommandLine.py:346)
> > > > > > > > > >        at CommandLine$py.call_function(Z:\UBS\Storage\.\scratch\jython\pybench\CommandLine.py)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyFunction.__call__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyInstance.__init__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyClass.__call__(Unknown Source)
> > > > > > > > > >        at org.python.core.PyObject.__call__(Unknown Source)
> > > > > > > > > >        at org.python.pycode._pyx1.f$0(.\scratch\jython\pybench\pybench.py:946)
> > > > > > > > > >        at org.python.pycode._pyx1.call_function(.\scratch\jython\pybench\pybench.py)
> > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.PyCode.call(Unknown Source)
> > > > > > > > > >        at org.python.core.Py.runCode(Unknown Source)
> > > > > > > > > >        at org.python.core.__builtin__.execfile_flags(Unknown Source)
> > > > > > > > > >        at org.python.util.PythonInterpreter.execfile(Unknown Source)
> > > > > > > > > >        at org.python.util.jython.main(Unknown Source)
> > > > > > > > > >        at dacapo.jython.JythonHarness.iterate(JythonHarness.java:38)
> > > > > > > > > >        at dacapo.Benchmark.run(Benchmark.java:126)
> > > > > > > > > >        at dacapo.TestHarness.runBenchmark(TestHarness.java:302)
> > > > > > > > > >        at dacapo.TestHarness.main(TestHarness.java:242)
> > > > > > > > > >        at Harness.main(Harness.java:5)
> > > > > > > > > >        at java.lang.reflect.VMReflection.invokeMethod(VMReflection.java)
> > > > > > > > > >        at java.lang.reflect.Method.invoke(Method.java:317)
> > > > > > > > > >        at org.apache.harmony.vm.JarRunner.main(JarRunner.java:80)
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > java.lang.IllegalArgumentException:
> > > > > > > > > > java.lang.IllegalArgumentException: Cannot create PyString from null!
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > Thanks.
> > > > > > > > > > Vladimir.
> > > > > > > > > >
> > > > > > > > > > On 11/6/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > > I've committed the patch at r592434.
> > > > > > > > > > > the excluded classes are,
> > > > > > > > > > >
> > > > > > > > > > > tests.api.java.util.CurrencyTest
> > > > > > > > > > > tests.api.java.util.FormatterTest
> > > > > > > > > > > tests.api.java.util.GregorianCalendarTest
> > > > > > > > > > > tests.api.java.util.LocaleTest
> > > > > > > > > > > tests.api.java.util.ScannerTest
> > > > > > > > > > >
> > > > > > > > > > > org.apache.harmony.text.tests.java.text.DecimalFormatSymbolsTest
> > > > > > > > > > > org.apache.harmony.text.tests.java.text.NumberFormatTest
> > > > > > > > > > > org.apache.harmony.text.tests.java.text.SimpleDateFormatTest
> > > > > > > > > > >
> > > > > > > > > > > pls kindly let me know if you find any problem, thanks.
> > > > > > > > > > >
> > > > > > > > > > > On 11/6/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > > > On 11/6/07, Tim Ellison <t....@gmail.com> wrote:
> > > > > > > > > > > > > Tony Wu wrote:
> > > > > > > > > > > > > > I've raised a JIRA[1] for migrating the dependencies of locale related
> > > > > > > > > > > > > > data to icu4j and remove the data in
> > > > > > > > > > > > > > luni/src/main/java/org/apache/harmony/luni/internal/locale/. Currently
> > > > > > > > > > > > > > I have delegated all the locale related classes to corresponding
> > > > > > > > > > > > > > classes in icu4j and successfully removed the dependencies of harmony
> > > > > > > > > > > > > > resource bundles.
> > > > > > > > > > > > >
> > > > > > > > > > > > > Cool -- this is good because it removes the code duplication we have at
> > > > > > > > > > > > > present, it shifts the maintenance of the locale data into the right
> > > > > > > > > > > > > place (IMHO) that being the ICU project, and gives us the opportunity to
> > > > > > > > > > > > > use the ICU tools to customize/refresh/etc the data for Harmony each
> > > > > > > > > > > > > time the USA decide to introduce another daylight savings change ;-)
> > > > > > > > > > > >
> > > > > > > > > > > > exactly.
> > > > > > > > > > > > >
> > > > > > > > > > > > > > As expected, there are around 20 failures of harmony test. Many of
> > > > > > > > > > > > > > them are caused by data difference. I've listed all of them on harmony
> > > > > > > > > > > > > > wiki[2] and will raise them to icu soon.
> > > > > > > > > > > > >
> > > > > > > > > > > > > Ack.  I believe that (pretty much) everyone gets the data from the CLDR,
> > > > > > > > > > > > > so I would expect any differences to be tests that have out of date
> > > > > > > > > > > > > assumptions about the data.
> > > > > > > > > > > >
> > > > > > > > > > > > Yes, I'll check it based on CLDR.
> > > > > > > > > > > > >
> > > > > > > > > > > > > > By applying this patch, we can get 4.2 mega bytes decrease in harmony
> > > > > > > > > > > > > > source code, including the svn data.
> > > > > > > > > > > > >
> > > > > > > > > > > > > ooh, say that again!  Removing 4.2Mb of redundant source code is excellent.
> > > > > > > > > > > > >
> > > > > > > > > > > > > > My proposal is to apply the patch on 5061 first and exclude these
> > > > > > > > > > > > > > tests which fail on different locale data. Then I'll take follow up
> > > > > > > > > > > > > > action to contact icu team and get solution to move them out.
> > > > > > > > > > > > > > Otherwise I'm afraid my fix will be outdated for waiting until icu
> > > > > > > > > > > > > > team get ready to fix the data. Do you have any objection?
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Later, I'll take care of the performance issue of this delegation.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Thanks in advance for your comments
> > > > > > > > > > > > >
> > > > > > > > > > > > > Sounds like a good plan.
> > > > > > > > > > > > >
> > > > > > > > > > > > > Regards,
> > > > > > > > > > > > > Tim
> > > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > > --
> > > > > > > > > > > > Tony Wu
> > > > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > --
> > > > > > > > > > > Tony Wu
> > > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > --
> > > > > > > > > Tony Wu
> > > > > > > > > China Software Development Lab, IBM
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > > Tony Wu
> > > > > > > China Software Development Lab, IBM
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > >
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Tim Ellison <t....@gmail.com>.
Tony Wu wrote:
> I assume this patch is ok finally if no more reponse :)

Sounds good to me.

Regards,
Tim


Re: [classlib]remove the duplicate locale data

Posted by Tim Ellison <t....@gmail.com>.
Aleksey Shipilev wrote:
> Tim, I had already applied Tony's patch and build it OK, so no
> modification needed I guess.

Ok (thanks Tony).

Tim

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Tim, I had already applied Tony's patch and build it OK, so no
modification needed I guess.

Thanks,
Aleksey.

On Jan 14, 2008 3:12 PM, Tim Ellison <t....@gmail.com> wrote:
> Sorry Tony, that's likely me restructuring the LUNI tests to the package
> naming convention.  I'll help restructure your diff if you want.

Re: [classlib]remove the duplicate locale data

Posted by Tim Ellison <t....@gmail.com>.
Tony Wu wrote:
> Oh yes, thank you Aleksay. I hadn't commit this patch yet.
> 
> My thinkpad got something wrong these days, so I just get back to work
> today. I double checked the patch on HARMONY-5061, seems it is
> outdated because of the test structure change in luni. I've provided a
> new one named "allinone0114.diff". would you please verify it? thanks
> a lot!

Sorry Tony, that's likely me restructuring the LUNI tests to the package 
naming convention.  I'll help restructure your diff if you want.

Regards,
Tim

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
On 2/12/08, Pavel Pervov <pm...@gmail.com> wrote:
> Tony,
>
> <SNIP>
>
> > Hi, Pavel and all
> > my solution below
> >
> > > https://issues.apache.org/jira/browse/HARMONY-5459
> > Fixed, thanks for your patch.
> Everything works fine now, thanks.
> > > https://issues.apache.org/jira/browse/HARMONY-5461
> > Can not reproduce
> Closed as invalid.
> > > https://issues.apache.org/jira/browse/HARMONY-5465
> > Reported to ICU before. ICU refused to fix but we can work around it
> > if necessary. Need your suggestion.
> I'd suggest to workaround it in Harmony.

worked around at r620735
> > > https://issues.apache.org/jira/browse/HARMONY-5466
> > non-bug difference, due to CLDR data.
> Should we close it now?

closed
> > > https://issues.apache.org/jira/browse/HARMONY-5467
> > non-bug difference, due to CLDR data.
> Same question here.
> > > https://issues.apache.org/jira/browse/HARMONY-5468
> > report to http://bugs.icu-project.org/trac/ticket/6174
> It is targeted to 4.0 now.
> > > https://issues.apache.org/jira/browse/HARMONY-5469
> > I suppose it is another CLDR data difference, need your commets.
> I've attached the test there. Please, look now.

This is a non-bug difference. In the current CLDR, the SHORT pattern
is "dd/MM/yyyy" in UK. Closed too.
>
> <SNIP>
>
>
> Thanks you,
>    Pavel.
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Pavel Pervov <pm...@gmail.com>.
Tony,

<SNIP>

> Hi, Pavel and all
> my solution below
>
> > https://issues.apache.org/jira/browse/HARMONY-5459
> Fixed, thanks for your patch.
Everything works fine now, thanks.
> > https://issues.apache.org/jira/browse/HARMONY-5461
> Can not reproduce
Closed as invalid.
> > https://issues.apache.org/jira/browse/HARMONY-5465
> Reported to ICU before. ICU refused to fix but we can work around it
> if necessary. Need your suggestion.
I'd suggest to workaround it in Harmony.
> > https://issues.apache.org/jira/browse/HARMONY-5466
> non-bug difference, due to CLDR data.
Should we close it now?
> > https://issues.apache.org/jira/browse/HARMONY-5467
> non-bug difference, due to CLDR data.
Same question here.
> > https://issues.apache.org/jira/browse/HARMONY-5468
> report to http://bugs.icu-project.org/trac/ticket/6174
It is targeted to 4.0 now.
> > https://issues.apache.org/jira/browse/HARMONY-5469
> I suppose it is another CLDR data difference, need your commets.
I've attached the test there. Please, look now.

<SNIP>


Thanks you,
    Pavel.

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
On 2/10/08, Tim Ellison <t....@gmail.com> wrote:
> Good work Tony.  Perhaps a wiki page or umbrella JIRA would help us
> track these?  I think they may get lost in mail discussion.

That's right. Update to wiki page
http://wiki.apache.org/harmony/MigrateToICU

>
> Just my 2c.
> Tim
>
> Tony Wu wrote:
> > On 2/7/08, Pavel Pervov <pm...@gmail.com> wrote:
> >> Tony,
> >>
> >> Unfortunately, I have a number of JIRAs for you. They all are
> >> regressions after removing duplicate locale data.
> >>
> >> Here they go:
> >>
> > Hi, Pavel and all
> > my solution below
> >
> >> https://issues.apache.org/jira/browse/HARMONY-5459
> > Fixed, thanks for your patch.
> >> https://issues.apache.org/jira/browse/HARMONY-5461
> > Can not reproduce
> >> https://issues.apache.org/jira/browse/HARMONY-5465
> > Reported to ICU before. ICU refused to fix but we can work around it
> > if necessary. Need your suggestion.
> >> https://issues.apache.org/jira/browse/HARMONY-5466
> > non-bug difference, due to CLDR data.
> >> https://issues.apache.org/jira/browse/HARMONY-5467
> > non-bug difference, due to CLDR data.
> >> https://issues.apache.org/jira/browse/HARMONY-5468
> > report to http://bugs.icu-project.org/trac/ticket/6174
> >> https://issues.apache.org/jira/browse/HARMONY-5469
> > I suppose it is another CLDR data difference, need your commets.
> >
> >
> >> I would like to know your opinion on whether they should be fixed of
> >> resolved some other way.
> >>
> >> Thanks,
> >>    Pavel.
> >>
> >> P.S. There is also HARMONY-5013 which was introduced with moving to
> >> ICU4J 3.8. Could you, please, look at that issue too?
> >>
> >> On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
> >>> Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
> >>>
> >>> On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> >>>> Hi Tony,
> >>>>
> >>>> The good thing is, after the commit I have compared revisions 610727
> >>>> (before your commit) and 612866 (after your commit) and haven't
> >>>> noticed degradation while running SPECjbb2005 in 2-VM configuration.
> >>>> That's definitely good news :)
> >>>>
> >>>> Thanks,
> >>>> Aleksey.
> >>>>
> >>>> On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> >>>>> Thanks, Aleksey and all
> >>>>>
> >>>>> patch committed at r612718. Then I'm going to deal with the non-bug
> >>>>> difference. Hopefully many legcy bugs/differences can be fixed this
> >>>>> time.
> >>>>>
> >>>>>
> >>>>> On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> >>>>>> Hi, guys!
> >>>>>>
> >>>>>> Sure, correctness is important, but performance is important too. I
> >>>>>> had profiled both versions (clean and patched) and see no significant
> >>>>>> difference there: there are more GC happens what I believe connected
> >>>>>> to internal ICU object creation and such. So, there are no obvious way
> >>>>>> to maintain the same performance level, and may be we will try again
> >>>>>> to eliminate ICU usage from the hotpath of frequently used workloads -
> >>>>>> but previous investigation shows it's not that simple.
> >>>>>>
> >>>>>> Tony, please go ahead with committing this patch, we will deal with
> >>>>>> ICU performance issues a little later.
> >>>>>>
> >>>>>> Thanks,
> >>>>>> Aleksey.
> >>>>>>
> >>>>>> On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> >>>>>>> Hi Aleksey,
> >>>>>>>
> >>>>>>> Thanks for you help.
> >>>>>>>
> >>>>>>> I've proved there is no performance degradation on my machine
> >>>>>>> mentioned in my previous mail. I suppose the different result on your
> >>>>>>> machine is caused by different options to SpecJBB. Anyway, my POV is
> >>>>>>> that we should be willing to pay for the adoption of ICU if we have
> >>>>>>> to. All of us should be positive on this point. I'd like to clarify
> >>>>>>> the factors I'm facing below.
> >>>>>>>
> >>>>>>> Firstly, the original implementation of harmony is faster but
> >>>>>>> incorrect. It is not reasonable to keep the code as is and refuse to
> >>>>>>> correct it just because the bad version has better performance. IMHO
> >>>>>>> performance is nothing if there is no correctness.
> >>>>>>>
> >>>>>>> Secondly, we adopt ICU through delegation which involves extra method
> >>>>>>> calls than we implement it by ourselves, it does harm to performance
> >>>>>>> and can not be worked around. But please do not forget we benefit from
> >>>>>>> ICU in bug fixing, maintenance, smaller code size and so on.
> >>>>>>>
> >>>>>>> Lastly, branching does not make sense to me. My fix is very separate
> >>>>>>> in luni and text, I can not guarantee that there is no modification
> >>>>>>> during my work, so the synchronization between HEAD and my branch is
> >>>>>>> required. Actually only I myself am working on the development work
> >>>>>>> (Surely, Aleksey is very helpful on testing), this synchronization
> >>>>>>> might be a nightmare to me. Furthermore, the code on branch will not
> >>>>>>> be automatically tested by continuous integration system and BTI, I do
> >>>>>>> not want to work without collaboration, that's not an open source
> >>>>>>> style, right? Will you ask a contributor to create a branch and play
> >>>>>>> with himself whenever he wants to contribute?
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>> On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> >>>>>>>> On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> >>>>>>>>> Aleksey Shipilev wrote:
> >>>>>>>>>> Yep, Tim, you're right. I believe that new implementation fixes a
> >>>>>>>>>> number of bugs and will try to get it not degrading. I just want to
> >>>>>>>>>> maintain the performance level of current trunk on the same level,
> >>>>>>>>>> gradually fixing functional bugs. I don't like to sacrifice
> >>>>>>>>>> performance of HEAD revision for non-critical bugfix. That is, I want
> >>>>>>>>>> to see HEAD changes like this:
> >>>>>>>>>>
> >>>>>>>>>> "high performance, minor bug -> high performance, no bugs"
> >>>>>>>>>>
> >>>>>>>>>> rather than
> >>>>>>>>>>
> >>>>>>>>>> "high performance, minor bug -> low performance, no bugs -> high
> >>>>>>>>>> performance, no bugs"
> >>>>>>>>>>
> >>>>>>>>>> ...because anyone could get the HEAD Harmony revision for
> >>>>>>>>>> performance measurements at any time.
> >>>>>>>> Couldn't someone also get the HEAD Harmony revision and suffer from the
> >>>>>>>> known/fixable-with-Tony's-patch bugs at any time?
> >>>>>>>>
> >>>>>>>>>> What do you think?
> >>>>>>>>> For sure, improving performance and fixing the bugs is the most
> >>>>>>>>> desirable state.  I actually don't mind some minor performance
> >>>>>>>>> regression on HEAD between releases provided it is an area being
> >>>>>>>>> actively worked upon.
> >>>>>>>> +1 especially if it fixes bugs
> >>>>>>>>
> >>>>>>>>> I'd also like to get to 4.2Mb source code reduction too ;-)
> >>>>>>>> Me too.
> >>>>>>>>
> >>>>>>>>> If you and Tony are happy to work on the patch to get it perfect then
> >>>>>>>>> go ahead.  I hope it is not too troublesome to keep it in synch.  You
> >>>>>>>>> could also consider a branch in SVN.
> >>>>>>>> This bothers me too.  Firstly, while it is being developed in patch
> >>>>>>>> on JIRA, it is likely only Tony and Aleksey will really look at it.
> >>>>>>>> Secondly, that progress will be slow because of the cost of keeping in
> >>>>>>>> sync - this applies to an SVN branch too.
> >>>>>>>>
> >>>>>>>> I can't help thinking we'll make more progress if we apply the patch to
> >>>>>>>> HEAD now.  We'll get wider visibility of problems with the new code -
> >>>>>>>> and there are likely issues beyond the performance problems that have
> >>>>>>>> been the focus so far - and more people will see the benefit of Tony
> >>>>>>>> (and Aleksey's) hard work in getting us to this point.
> >>>>>>>>
> >>>>>>>> I certainly don't want all this work to be completed outside svn HEAD
> >>>>>>>> and committed a week or two before the freeze for next milestone.
> >>>>>>>>
> >>>>>>>> Regards,
> >>>>>>>>  Mark.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>
> >>>>>>> --
> >>>>>>>
> >>>>>>> Tony Wu
> >>>>>>> China Software Development Lab, IBM
> >>>>>>>
> >>>>>
> >>>>> --
> >>>>>
> >>>>> Tony Wu
> >>>>> China Software Development Lab, IBM
> >>>>>
> >>>
> >>> --
> >>> Tony Wu
> >>> China Software Development Lab, IBM
> >>>
> >>
> >> --
> >> Pavel Pervov,
> >> Intel Enterprise Solutions Software Division
> >>
> >
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Tim Ellison <t....@gmail.com>.
Good work Tony.  Perhaps a wiki page or umbrella JIRA would help us 
track these?  I think they may get lost in mail discussion.

Just my 2c.
Tim

Tony Wu wrote:
> On 2/7/08, Pavel Pervov <pm...@gmail.com> wrote:
>> Tony,
>>
>> Unfortunately, I have a number of JIRAs for you. They all are
>> regressions after removing duplicate locale data.
>>
>> Here they go:
>>
> Hi, Pavel and all
> my solution below
> 
>> https://issues.apache.org/jira/browse/HARMONY-5459
> Fixed, thanks for your patch.
>> https://issues.apache.org/jira/browse/HARMONY-5461
> Can not reproduce
>> https://issues.apache.org/jira/browse/HARMONY-5465
> Reported to ICU before. ICU refused to fix but we can work around it
> if necessary. Need your suggestion.
>> https://issues.apache.org/jira/browse/HARMONY-5466
> non-bug difference, due to CLDR data.
>> https://issues.apache.org/jira/browse/HARMONY-5467
> non-bug difference, due to CLDR data.
>> https://issues.apache.org/jira/browse/HARMONY-5468
> report to http://bugs.icu-project.org/trac/ticket/6174
>> https://issues.apache.org/jira/browse/HARMONY-5469
> I suppose it is another CLDR data difference, need your commets.
> 
> 
>> I would like to know your opinion on whether they should be fixed of
>> resolved some other way.
>>
>> Thanks,
>>    Pavel.
>>
>> P.S. There is also HARMONY-5013 which was introduced with moving to
>> ICU4J 3.8. Could you, please, look at that issue too?
>>
>> On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
>>> Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
>>>
>>> On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
>>>> Hi Tony,
>>>>
>>>> The good thing is, after the commit I have compared revisions 610727
>>>> (before your commit) and 612866 (after your commit) and haven't
>>>> noticed degradation while running SPECjbb2005 in 2-VM configuration.
>>>> That's definitely good news :)
>>>>
>>>> Thanks,
>>>> Aleksey.
>>>>
>>>> On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
>>>>> Thanks, Aleksey and all
>>>>>
>>>>> patch committed at r612718. Then I'm going to deal with the non-bug
>>>>> difference. Hopefully many legcy bugs/differences can be fixed this
>>>>> time.
>>>>>
>>>>>
>>>>> On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
>>>>>> Hi, guys!
>>>>>>
>>>>>> Sure, correctness is important, but performance is important too. I
>>>>>> had profiled both versions (clean and patched) and see no significant
>>>>>> difference there: there are more GC happens what I believe connected
>>>>>> to internal ICU object creation and such. So, there are no obvious way
>>>>>> to maintain the same performance level, and may be we will try again
>>>>>> to eliminate ICU usage from the hotpath of frequently used workloads -
>>>>>> but previous investigation shows it's not that simple.
>>>>>>
>>>>>> Tony, please go ahead with committing this patch, we will deal with
>>>>>> ICU performance issues a little later.
>>>>>>
>>>>>> Thanks,
>>>>>> Aleksey.
>>>>>>
>>>>>> On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
>>>>>>> Hi Aleksey,
>>>>>>>
>>>>>>> Thanks for you help.
>>>>>>>
>>>>>>> I've proved there is no performance degradation on my machine
>>>>>>> mentioned in my previous mail. I suppose the different result on your
>>>>>>> machine is caused by different options to SpecJBB. Anyway, my POV is
>>>>>>> that we should be willing to pay for the adoption of ICU if we have
>>>>>>> to. All of us should be positive on this point. I'd like to clarify
>>>>>>> the factors I'm facing below.
>>>>>>>
>>>>>>> Firstly, the original implementation of harmony is faster but
>>>>>>> incorrect. It is not reasonable to keep the code as is and refuse to
>>>>>>> correct it just because the bad version has better performance. IMHO
>>>>>>> performance is nothing if there is no correctness.
>>>>>>>
>>>>>>> Secondly, we adopt ICU through delegation which involves extra method
>>>>>>> calls than we implement it by ourselves, it does harm to performance
>>>>>>> and can not be worked around. But please do not forget we benefit from
>>>>>>> ICU in bug fixing, maintenance, smaller code size and so on.
>>>>>>>
>>>>>>> Lastly, branching does not make sense to me. My fix is very separate
>>>>>>> in luni and text, I can not guarantee that there is no modification
>>>>>>> during my work, so the synchronization between HEAD and my branch is
>>>>>>> required. Actually only I myself am working on the development work
>>>>>>> (Surely, Aleksey is very helpful on testing), this synchronization
>>>>>>> might be a nightmare to me. Furthermore, the code on branch will not
>>>>>>> be automatically tested by continuous integration system and BTI, I do
>>>>>>> not want to work without collaboration, that's not an open source
>>>>>>> style, right? Will you ask a contributor to create a branch and play
>>>>>>> with himself whenever he wants to contribute?
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
>>>>>>>> On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
>>>>>>>>> Aleksey Shipilev wrote:
>>>>>>>>>> Yep, Tim, you're right. I believe that new implementation fixes a
>>>>>>>>>> number of bugs and will try to get it not degrading. I just want to
>>>>>>>>>> maintain the performance level of current trunk on the same level,
>>>>>>>>>> gradually fixing functional bugs. I don't like to sacrifice
>>>>>>>>>> performance of HEAD revision for non-critical bugfix. That is, I want
>>>>>>>>>> to see HEAD changes like this:
>>>>>>>>>>
>>>>>>>>>> "high performance, minor bug -> high performance, no bugs"
>>>>>>>>>>
>>>>>>>>>> rather than
>>>>>>>>>>
>>>>>>>>>> "high performance, minor bug -> low performance, no bugs -> high
>>>>>>>>>> performance, no bugs"
>>>>>>>>>>
>>>>>>>>>> ...because anyone could get the HEAD Harmony revision for
>>>>>>>>>> performance measurements at any time.
>>>>>>>> Couldn't someone also get the HEAD Harmony revision and suffer from the
>>>>>>>> known/fixable-with-Tony's-patch bugs at any time?
>>>>>>>>
>>>>>>>>>> What do you think?
>>>>>>>>> For sure, improving performance and fixing the bugs is the most
>>>>>>>>> desirable state.  I actually don't mind some minor performance
>>>>>>>>> regression on HEAD between releases provided it is an area being
>>>>>>>>> actively worked upon.
>>>>>>>> +1 especially if it fixes bugs
>>>>>>>>
>>>>>>>>> I'd also like to get to 4.2Mb source code reduction too ;-)
>>>>>>>> Me too.
>>>>>>>>
>>>>>>>>> If you and Tony are happy to work on the patch to get it perfect then
>>>>>>>>> go ahead.  I hope it is not too troublesome to keep it in synch.  You
>>>>>>>>> could also consider a branch in SVN.
>>>>>>>> This bothers me too.  Firstly, while it is being developed in patch
>>>>>>>> on JIRA, it is likely only Tony and Aleksey will really look at it.
>>>>>>>> Secondly, that progress will be slow because of the cost of keeping in
>>>>>>>> sync - this applies to an SVN branch too.
>>>>>>>>
>>>>>>>> I can't help thinking we'll make more progress if we apply the patch to
>>>>>>>> HEAD now.  We'll get wider visibility of problems with the new code -
>>>>>>>> and there are likely issues beyond the performance problems that have
>>>>>>>> been the focus so far - and more people will see the benefit of Tony
>>>>>>>> (and Aleksey's) hard work in getting us to this point.
>>>>>>>>
>>>>>>>> I certainly don't want all this work to be completed outside svn HEAD
>>>>>>>> and committed a week or two before the freeze for next milestone.
>>>>>>>>
>>>>>>>> Regards,
>>>>>>>>  Mark.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>>
>>>>>>> Tony Wu
>>>>>>> China Software Development Lab, IBM
>>>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> Tony Wu
>>>>> China Software Development Lab, IBM
>>>>>
>>>
>>> --
>>> Tony Wu
>>> China Software Development Lab, IBM
>>>
>>
>> --
>> Pavel Pervov,
>> Intel Enterprise Solutions Software Division
>>
> 
> 

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
On 2/7/08, Pavel Pervov <pm...@gmail.com> wrote:
> Tony,
>
> Unfortunately, I have a number of JIRAs for you. They all are
> regressions after removing duplicate locale data.
>
> Here they go:
>
Hi, Pavel and all
my solution below

> https://issues.apache.org/jira/browse/HARMONY-5459
Fixed, thanks for your patch.
> https://issues.apache.org/jira/browse/HARMONY-5461
Can not reproduce
> https://issues.apache.org/jira/browse/HARMONY-5465
Reported to ICU before. ICU refused to fix but we can work around it
if necessary. Need your suggestion.
> https://issues.apache.org/jira/browse/HARMONY-5466
non-bug difference, due to CLDR data.
> https://issues.apache.org/jira/browse/HARMONY-5467
non-bug difference, due to CLDR data.
> https://issues.apache.org/jira/browse/HARMONY-5468
report to http://bugs.icu-project.org/trac/ticket/6174
> https://issues.apache.org/jira/browse/HARMONY-5469
I suppose it is another CLDR data difference, need your commets.


>
> I would like to know your opinion on whether they should be fixed of
> resolved some other way.
>
> Thanks,
>    Pavel.
>
> P.S. There is also HARMONY-5013 which was introduced with moving to
> ICU4J 3.8. Could you, please, look at that issue too?
>
> On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
> > Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
> >
> > On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > Hi Tony,
> > >
> > > The good thing is, after the commit I have compared revisions 610727
> > > (before your commit) and 612866 (after your commit) and haven't
> > > noticed degradation while running SPECjbb2005 in 2-VM configuration.
> > > That's definitely good news :)
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > Thanks, Aleksey and all
> > > >
> > > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > > time.
> > > >
> > > >
> > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > Hi, guys!
> > > > >
> > > > > Sure, correctness is important, but performance is important too. I
> > > > > had profiled both versions (clean and patched) and see no significant
> > > > > difference there: there are more GC happens what I believe connected
> > > > > to internal ICU object creation and such. So, there are no obvious way
> > > > > to maintain the same performance level, and may be we will try again
> > > > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > > > but previous investigation shows it's not that simple.
> > > > >
> > > > > Tony, please go ahead with committing this patch, we will deal with
> > > > > ICU performance issues a little later.
> > > > >
> > > > > Thanks,
> > > > > Aleksey.
> > > > >
> > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > Hi Aleksey,
> > > > > >
> > > > > > Thanks for you help.
> > > > > >
> > > > > > I've proved there is no performance degradation on my machine
> > > > > > mentioned in my previous mail. I suppose the different result on your
> > > > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > > > that we should be willing to pay for the adoption of ICU if we have
> > > > > > to. All of us should be positive on this point. I'd like to clarify
> > > > > > the factors I'm facing below.
> > > > > >
> > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > > > correct it just because the bad version has better performance. IMHO
> > > > > > performance is nothing if there is no correctness.
> > > > > >
> > > > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > > > calls than we implement it by ourselves, it does harm to performance
> > > > > > and can not be worked around. But please do not forget we benefit from
> > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > >
> > > > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > > > in luni and text, I can not guarantee that there is no modification
> > > > > > during my work, so the synchronization between HEAD and my branch is
> > > > > > required. Actually only I myself am working on the development work
> > > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > > > be automatically tested by continuous integration system and BTI, I do
> > > > > > not want to work without collaboration, that's not an open source
> > > > > > style, right? Will you ask a contributor to create a branch and play
> > > > > > with himself whenever he wants to contribute?
> > > > > >
> > > > > >
> > > > > >
> > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > >
> > > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > > > > maintain the performance level of current trunk on the same level,
> > > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > > > > to see HEAD changes like this:
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > > >
> > > > > > > > > rather than
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > > > performance, no bugs"
> > > > > > > > >
> > > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > > performance measurements at any time.
> > > > > > >
> > > > > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > >
> > > > > > > > > What do you think?
> > > > > > >
> > > > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > > regression on HEAD between releases provided it is an area being
> > > > > > > > actively worked upon.
> > > > > > >
> > > > > > > +1 especially if it fixes bugs
> > > > > > >
> > > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > > >
> > > > > > > Me too.
> > > > > > >
> > > > > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > > > > could also consider a branch in SVN.
> > > > > > >
> > > > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > > > > sync - this applies to an SVN branch too.
> > > > > > >
> > > > > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > > > > and there are likely issues beyond the performance problems that have
> > > > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > >
> > > > > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > > > > and committed a week or two before the freeze for next milestone.
> > > > > > >
> > > > > > > Regards,
> > > > > > >  Mark.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > >
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > >
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>
>
> --
> Pavel Pervov,
> Intel Enterprise Solutions Software Division
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Pavel,

Thanks for reporting them to JIRA.

I just did a quick look, most of them are the differences between ICU
and RI. For some problem like 5467, we have an agreement on dealing
with them as non-bug differences, others are the difference of
SimpleDateFormat which we are disucssing in another thread.

On 2/7/08, Pavel Pervov <pm...@gmail.com> wrote:
> Tony,
>
> Unfortunately, I have a number of JIRAs for you. They all are
> regressions after removing duplicate locale data.
>
> Here they go:
>
> https://issues.apache.org/jira/browse/HARMONY-5459
> https://issues.apache.org/jira/browse/HARMONY-5461
> https://issues.apache.org/jira/browse/HARMONY-5465
> https://issues.apache.org/jira/browse/HARMONY-5466
> https://issues.apache.org/jira/browse/HARMONY-5467
> https://issues.apache.org/jira/browse/HARMONY-5468
> https://issues.apache.org/jira/browse/HARMONY-5469
>
> I would like to know your opinion on whether they should be fixed of
> resolved some other way.
>
> Thanks,
>    Pavel.
>
> P.S. There is also HARMONY-5013 which was introduced with moving to
> ICU4J 3.8. Could you, please, look at that issue too?
>
> On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
> > Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
> >
> > On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > Hi Tony,
> > >
> > > The good thing is, after the commit I have compared revisions 610727
> > > (before your commit) and 612866 (after your commit) and haven't
> > > noticed degradation while running SPECjbb2005 in 2-VM configuration.
> > > That's definitely good news :)
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > Thanks, Aleksey and all
> > > >
> > > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > > time.
> > > >
> > > >
> > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > Hi, guys!
> > > > >
> > > > > Sure, correctness is important, but performance is important too. I
> > > > > had profiled both versions (clean and patched) and see no significant
> > > > > difference there: there are more GC happens what I believe connected
> > > > > to internal ICU object creation and such. So, there are no obvious way
> > > > > to maintain the same performance level, and may be we will try again
> > > > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > > > but previous investigation shows it's not that simple.
> > > > >
> > > > > Tony, please go ahead with committing this patch, we will deal with
> > > > > ICU performance issues a little later.
> > > > >
> > > > > Thanks,
> > > > > Aleksey.
> > > > >
> > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > Hi Aleksey,
> > > > > >
> > > > > > Thanks for you help.
> > > > > >
> > > > > > I've proved there is no performance degradation on my machine
> > > > > > mentioned in my previous mail. I suppose the different result on your
> > > > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > > > that we should be willing to pay for the adoption of ICU if we have
> > > > > > to. All of us should be positive on this point. I'd like to clarify
> > > > > > the factors I'm facing below.
> > > > > >
> > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > > > correct it just because the bad version has better performance. IMHO
> > > > > > performance is nothing if there is no correctness.
> > > > > >
> > > > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > > > calls than we implement it by ourselves, it does harm to performance
> > > > > > and can not be worked around. But please do not forget we benefit from
> > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > >
> > > > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > > > in luni and text, I can not guarantee that there is no modification
> > > > > > during my work, so the synchronization between HEAD and my branch is
> > > > > > required. Actually only I myself am working on the development work
> > > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > > > be automatically tested by continuous integration system and BTI, I do
> > > > > > not want to work without collaboration, that's not an open source
> > > > > > style, right? Will you ask a contributor to create a branch and play
> > > > > > with himself whenever he wants to contribute?
> > > > > >
> > > > > >
> > > > > >
> > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > >
> > > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > > > > maintain the performance level of current trunk on the same level,
> > > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > > > > to see HEAD changes like this:
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > > >
> > > > > > > > > rather than
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > > > performance, no bugs"
> > > > > > > > >
> > > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > > performance measurements at any time.
> > > > > > >
> > > > > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > >
> > > > > > > > > What do you think?
> > > > > > >
> > > > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > > regression on HEAD between releases provided it is an area being
> > > > > > > > actively worked upon.
> > > > > > >
> > > > > > > +1 especially if it fixes bugs
> > > > > > >
> > > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > > >
> > > > > > > Me too.
> > > > > > >
> > > > > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > > > > could also consider a branch in SVN.
> > > > > > >
> > > > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > > > > sync - this applies to an SVN branch too.
> > > > > > >
> > > > > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > > > > and there are likely issues beyond the performance problems that have
> > > > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > >
> > > > > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > > > > and committed a week or two before the freeze for next milestone.
> > > > > > >
> > > > > > > Regards,
> > > > > > >  Mark.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > >
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > >
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>
>
> --
> Pavel Pervov,
> Intel Enterprise Solutions Software Division
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Alexey Petrenko <al...@gmail.com>.
BTW, I would move this discussion to separate thread to get more
community attention and opinions.

SY, Alexey

2008/2/9, Alexey Petrenko <al...@gmail.com>:
> 2008/2/7, Mark Hindess <ma...@googlemail.com>:
> >
> > On 7 February 2008 at 15:48, "Alexey Petrenko" <al...@gmail.com>
> > wrote:
> > > 2008/2/6, Tony Wu <wu...@gmail.com>:
> > > > On 2/7/08, Alexey Petrenko <al...@gmail.com> wrote:
> > > > > I would suggest to mark these issues as "must be fixed before M5"
> > > > > since Harmony M5 should not be worse then M4.
> > > > Why do you think these differences are so important?
> > > Because these issues are regressions :)
> > Alexey, I don't think that is fair.
> Do not agree with this :)
>
> But
> > Regressions are changes that affect
> > things negatively and improvements are changes that affect things
> > positively.  Some of these may turn out to fall in the middle but many
> > of them will turn out to be improvements - i.e. because they are up to
> > date and the RI is outdated.
> I'm fully agree with this.
>
> We have applications or internal/external test suite failures. And all
> of this failures could be regressions or not. And we can not be fully
> sure until we will investigate each case.
> So if our next milestone build has new failures, does not matter on
> applications or test suites, anyone  can say that M5 is worse then M4.
> And we could not reply anything on this because we just do not know
> are these failures regressions, non-bug-differencies or even
> improvements.
>
> Thus I vote to investigate all the new discovered issues caused by
> this or any other commit and resolve them.
>
> SY, Alexey
>

Re: [classlib]remove the duplicate locale data

Posted by Tim Ellison <t....@gmail.com>.
Alexey Petrenko wrote:
> 2008/2/7, Mark Hindess <ma...@googlemail.com>:
>> On 7 February 2008 at 15:48, "Alexey Petrenko" <al...@gmail.com>
>> wrote:
>>> 2008/2/6, Tony Wu <wu...@gmail.com>:
>>>> On 2/7/08, Alexey Petrenko <al...@gmail.com> wrote:
>>>>> I would suggest to mark these issues as "must be fixed before M5"
>>>>> since Harmony M5 should not be worse then M4.
>>>> Why do you think these differences are so important?
>>> Because these issues are regressions :)
>> Alexey, I don't think that is fair.
> Do not agree with this :)
> 
> But
>> Regressions are changes that affect
>> things negatively and improvements are changes that affect things
>> positively.  Some of these may turn out to fall in the middle but many
>> of them will turn out to be improvements - i.e. because they are up to
>> date and the RI is outdated.
> I'm fully agree with this.
> 
> We have applications or internal/external test suite failures. And all
> of this failures could be regressions or not. And we can not be fully
> sure until we will investigate each case.
> So if our next milestone build has new failures, does not matter on
> applications or test suites, anyone  can say that M5 is worse then M4.
> And we could not reply anything on this because we just do not know
> are these failures regressions, non-bug-differencies or even
> improvements.
> 
> Thus I vote to investigate all the new discovered issues caused by
> this or any other commit and resolve them.

Yes, they should all be investigated so we can make a determination for M5.

Regards,
Tim

Re: [classlib]remove the duplicate locale data

Posted by Alexey Petrenko <al...@gmail.com>.
2008/2/7, Mark Hindess <ma...@googlemail.com>:
>
> On 7 February 2008 at 15:48, "Alexey Petrenko" <al...@gmail.com>
> wrote:
> > 2008/2/6, Tony Wu <wu...@gmail.com>:
> > > On 2/7/08, Alexey Petrenko <al...@gmail.com> wrote:
> > > > I would suggest to mark these issues as "must be fixed before M5"
> > > > since Harmony M5 should not be worse then M4.
> > > Why do you think these differences are so important?
> > Because these issues are regressions :)
> Alexey, I don't think that is fair.
Do not agree with this :)

But
> Regressions are changes that affect
> things negatively and improvements are changes that affect things
> positively.  Some of these may turn out to fall in the middle but many
> of them will turn out to be improvements - i.e. because they are up to
> date and the RI is outdated.
I'm fully agree with this.

We have applications or internal/external test suite failures. And all
of this failures could be regressions or not. And we can not be fully
sure until we will investigate each case.
So if our next milestone build has new failures, does not matter on
applications or test suites, anyone  can say that M5 is worse then M4.
And we could not reply anything on this because we just do not know
are these failures regressions, non-bug-differencies or even
improvements.

Thus I vote to investigate all the new discovered issues caused by
this or any other commit and resolve them.

SY, Alexey

Re: [classlib]remove the duplicate locale data

Posted by Mark Hindess <ma...@googlemail.com>.
On 7 February 2008 at 15:48, "Alexey Petrenko" <al...@gmail.com> 
wrote:
> 2008/2/6, Tony Wu <wu...@gmail.com>:
> > On 2/7/08, Alexey Petrenko <al...@gmail.com> wrote:
> > > I would suggest to mark these issues as "must be fixed before M5"
> > > since Harmony M5 should not be worse then M4.
> > Why do you think these differences are so important?
> Because these issues are regressions :)

Alexey, I don't think that is fair.  Regressions are changes that affect
things negatively and improvements are changes that affect things
positively.  Some of these may turn out to fall in the middle but many
of them will turn out to be improvements - i.e. because they are up to
date and the RI is outdated.

To me, Tony's question about why these are important is crucial.

If the issues were found with test cases, then why aren't these tests
checked in?  (If they were Tony would have had a chance to discover them
before making his changes in trunk.)

If the issues were found by application errors, then these are fair more
critical and I'm sure Tony was asking so that he might focus on these.

Personally, if these issues are application failures then I am +1 that
they "must be fixed before M5" but if they are test case regressions
then it is not so clear cut.  I'd have to weigh up the positive and
negative effects of the ICU changes.

Regards,
 Mark.



Re: [classlib]remove the duplicate locale data

Posted by Alexey Petrenko <al...@gmail.com>.
2008/2/6, Tony Wu <wu...@gmail.com>:
> On 2/7/08, Alexey Petrenko <al...@gmail.com> wrote:
> > I would suggest to mark these issues as "must be fixed before M5"
> > since Harmony M5 should not be worse then M4.
> Why do you think these differences are so important?
Because these issues are regressions :)

If some of them will be proved as non-bug-difference and everyone will
agree with this I do not have any objection about this kind of
resolution.

SY, Alexey

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
On 2/7/08, Alexey Petrenko <al...@gmail.com> wrote:
> I would suggest to mark these issues as "must be fixed before M5"
> since Harmony M5 should not be worse then M4.
>
> SY, Alexey

Hi, Alexey

Why do you think these differences are so important, do you have some
application which is blocked by some of them. I want to know which one
has higher priority.

I don't think they are so significant as to be considered as a
regradation. Actually they are different form of represtation rather
than incorrect behavior. And please do not think all the differences
against RI are bugs, I've claimed that ICU3.8 uses the CLDR 5.0 which
is much newer than RI. That is to say, the contry code, timezone name,
offset, etc. is up to date in Harmony and outdated in RI. And the
outdated data in this area does no make any sense.

As last, I have to say sorry for that I can not take action
immediately since today is Spring Festival in China, I must spend my
time with family. Don't worry, I can give you my word that I can fix
all of them before M5.

>
> 2008/2/6, Pavel Pervov <pm...@gmail.com>:
> > Tony,
> >
> > Unfortunately, I have a number of JIRAs for you. They all are
> > regressions after removing duplicate locale data.
> >
> > Here they go:
> >
> > https://issues.apache.org/jira/browse/HARMONY-5459
> > https://issues.apache.org/jira/browse/HARMONY-5461
> > https://issues.apache.org/jira/browse/HARMONY-5465
> > https://issues.apache.org/jira/browse/HARMONY-5466
> > https://issues.apache.org/jira/browse/HARMONY-5467
> > https://issues.apache.org/jira/browse/HARMONY-5468
> > https://issues.apache.org/jira/browse/HARMONY-5469
> >
> > I would like to know your opinion on whether they should be fixed of
> > resolved some other way.
> >
> > Thanks,
> >     Pavel.
> >
> > P.S. There is also HARMONY-5013 which was introduced with moving to
> > ICU4J 3.8. Could you, please, look at that issue too?
> >
> > On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
> > > Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
> > >
> > > On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > Hi Tony,
> > > >
> > > > The good thing is, after the commit I have compared revisions 610727
> > > > (before your commit) and 612866 (after your commit) and haven't
> > > > noticed degradation while running SPECjbb2005 in 2-VM configuration.
> > > > That's definitely good news :)
> > > >
> > > > Thanks,
> > > > Aleksey.
> > > >
> > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > Thanks, Aleksey and all
> > > > >
> > > > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > > > time.
> > > > >
> > > > >
> > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > Hi, guys!
> > > > > >
> > > > > > Sure, correctness is important, but performance is important too. I
> > > > > > had profiled both versions (clean and patched) and see no significant
> > > > > > difference there: there are more GC happens what I believe connected
> > > > > > to internal ICU object creation and such. So, there are no obvious way
> > > > > > to maintain the same performance level, and may be we will try again
> > > > > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > > > > but previous investigation shows it's not that simple.
> > > > > >
> > > > > > Tony, please go ahead with committing this patch, we will deal with
> > > > > > ICU performance issues a little later.
> > > > > >
> > > > > > Thanks,
> > > > > > Aleksey.
> > > > > >
> > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > Hi Aleksey,
> > > > > > >
> > > > > > > Thanks for you help.
> > > > > > >
> > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > mentioned in my previous mail. I suppose the different result on your
> > > > > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > > > > that we should be willing to pay for the adoption of ICU if we have
> > > > > > > to. All of us should be positive on this point. I'd like to clarify
> > > > > > > the factors I'm facing below.
> > > > > > >
> > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > > > > correct it just because the bad version has better performance. IMHO
> > > > > > > performance is nothing if there is no correctness.
> > > > > > >
> > > > > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > > > > calls than we implement it by ourselves, it does harm to performance
> > > > > > > and can not be worked around. But please do not forget we benefit from
> > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > >
> > > > > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > > > > in luni and text, I can not guarantee that there is no modification
> > > > > > > during my work, so the synchronization between HEAD and my branch is
> > > > > > > required. Actually only I myself am working on the development work
> > > > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > > > > be automatically tested by continuous integration system and BTI, I do
> > > > > > > not want to work without collaboration, that's not an open source
> > > > > > > style, right? Will you ask a contributor to create a branch and play
> > > > > > > with himself whenever he wants to contribute?
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > >
> > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > > > > > maintain the performance level of current trunk on the same level,
> > > > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > > > >
> > > > > > > > > > rather than
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > > > > performance, no bugs"
> > > > > > > > > >
> > > > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > > > performance measurements at any time.
> > > > > > > >
> > > > > > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > >
> > > > > > > > > > What do you think?
> > > > > > > >
> > > > > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > > > regression on HEAD between releases provided it is an area being
> > > > > > > > > actively worked upon.
> > > > > > > >
> > > > > > > > +1 especially if it fixes bugs
> > > > > > > >
> > > > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > > > >
> > > > > > > > Me too.
> > > > > > > >
> > > > > > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > > > > > could also consider a branch in SVN.
> > > > > > > >
> > > > > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > >
> > > > > > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > > > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > > > > > and there are likely issues beyond the performance problems that have
> > > > > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > >
> > > > > > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > > > > > and committed a week or two before the freeze for next milestone.
> > > > > > > >
> > > > > > > > Regards,
> > > > > > > >  Mark.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > >
> > > > > > > Tony Wu
> > > > > > > China Software Development Lab, IBM
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > >
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> > >
> > > --
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
> >
> > --
> > Pavel Pervov,
> > Intel Enterprise Solutions Software Division
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Pavel Pervov <pm...@gmail.com>.
Alexey,

I agree. Marked them with "fix version" 5.0M5.

Pavel.
On 2/6/08, Alexey Petrenko <al...@gmail.com> wrote:
> I would suggest to mark these issues as "must be fixed before M5"
> since Harmony M5 should not be worse then M4.
>
> SY, Alexey
>
> 2008/2/6, Pavel Pervov <pm...@gmail.com>:
> > Tony,
> >
> > Unfortunately, I have a number of JIRAs for you. They all are
> > regressions after removing duplicate locale data.
> >
> > Here they go:
> >
> > https://issues.apache.org/jira/browse/HARMONY-5459
> > https://issues.apache.org/jira/browse/HARMONY-5461
> > https://issues.apache.org/jira/browse/HARMONY-5465
> > https://issues.apache.org/jira/browse/HARMONY-5466
> > https://issues.apache.org/jira/browse/HARMONY-5467
> > https://issues.apache.org/jira/browse/HARMONY-5468
> > https://issues.apache.org/jira/browse/HARMONY-5469
> >
> > I would like to know your opinion on whether they should be fixed of
> > resolved some other way.
> >
> > Thanks,
> >     Pavel.
> >
> > P.S. There is also HARMONY-5013 which was introduced with moving to
> > ICU4J 3.8. Could you, please, look at that issue too?
> >
> > On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
> > > Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
> > >
> > > On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > Hi Tony,
> > > >
> > > > The good thing is, after the commit I have compared revisions 610727
> > > > (before your commit) and 612866 (after your commit) and haven't
> > > > noticed degradation while running SPECjbb2005 in 2-VM configuration.
> > > > That's definitely good news :)
> > > >
> > > > Thanks,
> > > > Aleksey.
> > > >
> > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > Thanks, Aleksey and all
> > > > >
> > > > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > > > time.
> > > > >
> > > > >
> > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > Hi, guys!
> > > > > >
> > > > > > Sure, correctness is important, but performance is important too. I
> > > > > > had profiled both versions (clean and patched) and see no significant
> > > > > > difference there: there are more GC happens what I believe connected
> > > > > > to internal ICU object creation and such. So, there are no obvious way
> > > > > > to maintain the same performance level, and may be we will try again
> > > > > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > > > > but previous investigation shows it's not that simple.
> > > > > >
> > > > > > Tony, please go ahead with committing this patch, we will deal with
> > > > > > ICU performance issues a little later.
> > > > > >
> > > > > > Thanks,
> > > > > > Aleksey.
> > > > > >
> > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > Hi Aleksey,
> > > > > > >
> > > > > > > Thanks for you help.
> > > > > > >
> > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > mentioned in my previous mail. I suppose the different result on your
> > > > > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > > > > that we should be willing to pay for the adoption of ICU if we have
> > > > > > > to. All of us should be positive on this point. I'd like to clarify
> > > > > > > the factors I'm facing below.
> > > > > > >
> > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > > > > correct it just because the bad version has better performance. IMHO
> > > > > > > performance is nothing if there is no correctness.
> > > > > > >
> > > > > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > > > > calls than we implement it by ourselves, it does harm to performance
> > > > > > > and can not be worked around. But please do not forget we benefit from
> > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > >
> > > > > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > > > > in luni and text, I can not guarantee that there is no modification
> > > > > > > during my work, so the synchronization between HEAD and my branch is
> > > > > > > required. Actually only I myself am working on the development work
> > > > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > > > > be automatically tested by continuous integration system and BTI, I do
> > > > > > > not want to work without collaboration, that's not an open source
> > > > > > > style, right? Will you ask a contributor to create a branch and play
> > > > > > > with himself whenever he wants to contribute?
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > >
> > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > > > > > maintain the performance level of current trunk on the same level,
> > > > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > > > >
> > > > > > > > > > rather than
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > > > > performance, no bugs"
> > > > > > > > > >
> > > > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > > > performance measurements at any time.
> > > > > > > >
> > > > > > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > >
> > > > > > > > > > What do you think?
> > > > > > > >
> > > > > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > > > regression on HEAD between releases provided it is an area being
> > > > > > > > > actively worked upon.
> > > > > > > >
> > > > > > > > +1 especially if it fixes bugs
> > > > > > > >
> > > > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > > > >
> > > > > > > > Me too.
> > > > > > > >
> > > > > > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > > > > > could also consider a branch in SVN.
> > > > > > > >
> > > > > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > >
> > > > > > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > > > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > > > > > and there are likely issues beyond the performance problems that have
> > > > > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > >
> > > > > > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > > > > > and committed a week or two before the freeze for next milestone.
> > > > > > > >
> > > > > > > > Regards,
> > > > > > > >  Mark.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > >
> > > > > > > Tony Wu
> > > > > > > China Software Development Lab, IBM
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > >
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> > >
> > > --
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
> >
> > --
> > Pavel Pervov,
> > Intel Enterprise Solutions Software Division
> >
>


-- 
Pavel Pervov,
Intel Enterprise Solutions Software Division

Re: [classlib]remove the duplicate locale data

Posted by Alexey Petrenko <al...@gmail.com>.
I would suggest to mark these issues as "must be fixed before M5"
since Harmony M5 should not be worse then M4.

SY, Alexey

2008/2/6, Pavel Pervov <pm...@gmail.com>:
> Tony,
>
> Unfortunately, I have a number of JIRAs for you. They all are
> regressions after removing duplicate locale data.
>
> Here they go:
>
> https://issues.apache.org/jira/browse/HARMONY-5459
> https://issues.apache.org/jira/browse/HARMONY-5461
> https://issues.apache.org/jira/browse/HARMONY-5465
> https://issues.apache.org/jira/browse/HARMONY-5466
> https://issues.apache.org/jira/browse/HARMONY-5467
> https://issues.apache.org/jira/browse/HARMONY-5468
> https://issues.apache.org/jira/browse/HARMONY-5469
>
> I would like to know your opinion on whether they should be fixed of
> resolved some other way.
>
> Thanks,
>     Pavel.
>
> P.S. There is also HARMONY-5013 which was introduced with moving to
> ICU4J 3.8. Could you, please, look at that issue too?
>
> On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
> > Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
> >
> > On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > Hi Tony,
> > >
> > > The good thing is, after the commit I have compared revisions 610727
> > > (before your commit) and 612866 (after your commit) and haven't
> > > noticed degradation while running SPECjbb2005 in 2-VM configuration.
> > > That's definitely good news :)
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > Thanks, Aleksey and all
> > > >
> > > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > > time.
> > > >
> > > >
> > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > Hi, guys!
> > > > >
> > > > > Sure, correctness is important, but performance is important too. I
> > > > > had profiled both versions (clean and patched) and see no significant
> > > > > difference there: there are more GC happens what I believe connected
> > > > > to internal ICU object creation and such. So, there are no obvious way
> > > > > to maintain the same performance level, and may be we will try again
> > > > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > > > but previous investigation shows it's not that simple.
> > > > >
> > > > > Tony, please go ahead with committing this patch, we will deal with
> > > > > ICU performance issues a little later.
> > > > >
> > > > > Thanks,
> > > > > Aleksey.
> > > > >
> > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > Hi Aleksey,
> > > > > >
> > > > > > Thanks for you help.
> > > > > >
> > > > > > I've proved there is no performance degradation on my machine
> > > > > > mentioned in my previous mail. I suppose the different result on your
> > > > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > > > that we should be willing to pay for the adoption of ICU if we have
> > > > > > to. All of us should be positive on this point. I'd like to clarify
> > > > > > the factors I'm facing below.
> > > > > >
> > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > > > correct it just because the bad version has better performance. IMHO
> > > > > > performance is nothing if there is no correctness.
> > > > > >
> > > > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > > > calls than we implement it by ourselves, it does harm to performance
> > > > > > and can not be worked around. But please do not forget we benefit from
> > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > >
> > > > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > > > in luni and text, I can not guarantee that there is no modification
> > > > > > during my work, so the synchronization between HEAD and my branch is
> > > > > > required. Actually only I myself am working on the development work
> > > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > > > be automatically tested by continuous integration system and BTI, I do
> > > > > > not want to work without collaboration, that's not an open source
> > > > > > style, right? Will you ask a contributor to create a branch and play
> > > > > > with himself whenever he wants to contribute?
> > > > > >
> > > > > >
> > > > > >
> > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > >
> > > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > > > > maintain the performance level of current trunk on the same level,
> > > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > > > > to see HEAD changes like this:
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > > >
> > > > > > > > > rather than
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > > > performance, no bugs"
> > > > > > > > >
> > > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > > performance measurements at any time.
> > > > > > >
> > > > > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > >
> > > > > > > > > What do you think?
> > > > > > >
> > > > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > > regression on HEAD between releases provided it is an area being
> > > > > > > > actively worked upon.
> > > > > > >
> > > > > > > +1 especially if it fixes bugs
> > > > > > >
> > > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > > >
> > > > > > > Me too.
> > > > > > >
> > > > > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > > > > could also consider a branch in SVN.
> > > > > > >
> > > > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > > > > sync - this applies to an SVN branch too.
> > > > > > >
> > > > > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > > > > and there are likely issues beyond the performance problems that have
> > > > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > >
> > > > > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > > > > and committed a week or two before the freeze for next milestone.
> > > > > > >
> > > > > > > Regards,
> > > > > > >  Mark.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > >
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > >
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>
>
> --
> Pavel Pervov,
> Intel Enterprise Solutions Software Division
>

Re: [classlib]remove the duplicate locale data

Posted by Pavel Pervov <pm...@gmail.com>.
Tony,

Unfortunately, I have a number of JIRAs for you. They all are
regressions after removing duplicate locale data.

Here they go:

https://issues.apache.org/jira/browse/HARMONY-5459
https://issues.apache.org/jira/browse/HARMONY-5461
https://issues.apache.org/jira/browse/HARMONY-5465
https://issues.apache.org/jira/browse/HARMONY-5466
https://issues.apache.org/jira/browse/HARMONY-5467
https://issues.apache.org/jira/browse/HARMONY-5468
https://issues.apache.org/jira/browse/HARMONY-5469

I would like to know your opinion on whether they should be fixed of
resolved some other way.

Thanks,
    Pavel.

P.S. There is also HARMONY-5013 which was introduced with moving to
ICU4J 3.8. Could you, please, look at that issue too?

On 1/23/08, Tony Wu <wu...@gmail.com> wrote:
> Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!
>
> On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > Hi Tony,
> >
> > The good thing is, after the commit I have compared revisions 610727
> > (before your commit) and 612866 (after your commit) and haven't
> > noticed degradation while running SPECjbb2005 in 2-VM configuration.
> > That's definitely good news :)
> >
> > Thanks,
> > Aleksey.
> >
> > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > Thanks, Aleksey and all
> > >
> > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > time.
> > >
> > >
> > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > Hi, guys!
> > > >
> > > > Sure, correctness is important, but performance is important too. I
> > > > had profiled both versions (clean and patched) and see no significant
> > > > difference there: there are more GC happens what I believe connected
> > > > to internal ICU object creation and such. So, there are no obvious way
> > > > to maintain the same performance level, and may be we will try again
> > > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > > but previous investigation shows it's not that simple.
> > > >
> > > > Tony, please go ahead with committing this patch, we will deal with
> > > > ICU performance issues a little later.
> > > >
> > > > Thanks,
> > > > Aleksey.
> > > >
> > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > Hi Aleksey,
> > > > >
> > > > > Thanks for you help.
> > > > >
> > > > > I've proved there is no performance degradation on my machine
> > > > > mentioned in my previous mail. I suppose the different result on your
> > > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > > that we should be willing to pay for the adoption of ICU if we have
> > > > > to. All of us should be positive on this point. I'd like to clarify
> > > > > the factors I'm facing below.
> > > > >
> > > > > Firstly, the original implementation of harmony is faster but
> > > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > > correct it just because the bad version has better performance. IMHO
> > > > > performance is nothing if there is no correctness.
> > > > >
> > > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > > calls than we implement it by ourselves, it does harm to performance
> > > > > and can not be worked around. But please do not forget we benefit from
> > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > >
> > > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > > in luni and text, I can not guarantee that there is no modification
> > > > > during my work, so the synchronization between HEAD and my branch is
> > > > > required. Actually only I myself am working on the development work
> > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > > be automatically tested by continuous integration system and BTI, I do
> > > > > not want to work without collaboration, that's not an open source
> > > > > style, right? Will you ask a contributor to create a branch and play
> > > > > with himself whenever he wants to contribute?
> > > > >
> > > > >
> > > > >
> > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > >
> > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > > > Aleksey Shipilev wrote:
> > > > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > > > maintain the performance level of current trunk on the same level,
> > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > > > to see HEAD changes like this:
> > > > > > > >
> > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > >
> > > > > > > > rather than
> > > > > > > >
> > > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > > performance, no bugs"
> > > > > > > >
> > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > performance measurements at any time.
> > > > > >
> > > > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > >
> > > > > > > > What do you think?
> > > > > >
> > > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > regression on HEAD between releases provided it is an area being
> > > > > > > actively worked upon.
> > > > > >
> > > > > > +1 especially if it fixes bugs
> > > > > >
> > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > >
> > > > > > Me too.
> > > > > >
> > > > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > > > could also consider a branch in SVN.
> > > > > >
> > > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > > > sync - this applies to an SVN branch too.
> > > > > >
> > > > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > > > and there are likely issues beyond the performance problems that have
> > > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > >
> > > > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > > > and committed a week or two before the freeze for next milestone.
> > > > > >
> > > > > > Regards,
> > > > > >  Mark.
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > >
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> > >
> > > --
> > >
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>


-- 
Pavel Pervov,
Intel Enterprise Solutions Software Division

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
I noticed that you have posted some patch for the CLDR difference.
Thank you very much. I'm really busy these days...

On 1/30/08, Sean Qiu <se...@gmail.com> wrote:
> Hi, Tony
>
> I volunteered to fix some of the differences for you, if you are busy.
>
> Best regards.
>
>
> 2008/1/28, Tony Wu <wu...@gmail.com>:
> >
> > Ok, investigating.
> >
> > On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > > Hi Tony,
> > >
> > > Please look at another issue connected with the ICU commit.
> > > https://issues.apache.org/jira/browse/HARMONY-5435
> > > It has not been investigated deeply yet.
> > >
> > > --Rustem
> > >
> > >
> > > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> > >
> > > > They are all caused by different implementation of DateFormat, which
> > > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > > some real product.  See following output, they have the same meaning
> > > > but don't exactly equal on text.
> > > >
> > > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > > Expected: 'text here 12:17:01 UTC PM and here'
> > > >
> > > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > > >
> > > > Result:   'text here 00-04-05 and here'
> > > > Expected: 'text here 05/04/00 and here'
> > > >
> > > > I'll open another thread to ask people's opinions on this difference.
> > > >
> > > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > > Tony,
> > > > >
> > > > > It looks like this commit caused a regression. Could you take a look
> > at
> > > > > HARMONY-5430, please?
> > > > >
> > > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > >
> > > > > > Thanks, Aleksey and all
> > > > > >
> > > > > > patch committed at r612718. Then I'm going to deal with the
> > non-bug
> > > > > > difference. Hopefully many legcy bugs/differences can be fixed
> > this
> > > > > > time.
> > > > > >
> > > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > > Hi, guys!
> > > > > > >
> > > > > > > Sure, correctness is important, but performance is important
> > too. I
> > > > > > > had profiled both versions (clean and patched) and see no
> > > > significant
> > > > > > > difference there: there are more GC happens what I believe
> > connected
> > > > > > > to internal ICU object creation and such. So, there are no
> > obvious
> > > > way
> > > > > > > to maintain the same performance level, and may be we will try
> > again
> > > > > > > to eliminate ICU usage from the hotpath of frequently used
> > workloads
> > > > -
> > > > > > > but previous investigation shows it's not that simple.
> > > > > > >
> > > > > > > Tony, please go ahead with committing this patch, we will deal
> > with
> > > > > > > ICU performance issues a little later.
> > > > > > >
> > > > > > > Thanks,
> > > > > > > Aleksey.
> > > > > > >
> > > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > Hi Aleksey,
> > > > > > > >
> > > > > > > > Thanks for you help.
> > > > > > > >
> > > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > > mentioned in my previous mail. I suppose the different result
> > on
> > > > your
> > > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> > POV
> > > > is
> > > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > > have
> > > > > > > > to. All of us should be positive on this point. I'd like to
> > > > clarify
> > > > > > > > the factors I'm facing below.
> > > > > > > >
> > > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > > incorrect. It is not reasonable to keep the code as is and
> > refuse
> > > > to
> > > > > > > > correct it just because the bad version has better
> > performance.
> > > > IMHO
> > > > > > > > performance is nothing if there is no correctness.
> > > > > > > >
> > > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > > method
> > > > > > > > calls than we implement it by ourselves, it does harm to
> > > > performance
> > > > > > > > and can not be worked around. But please do not forget we
> > benefit
> > > > from
> > > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > > >
> > > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > > separate
> > > > > > > > in luni and text, I can not guarantee that there is no
> > > > modification
> > > > > > > > during my work, so the synchronization between HEAD and my
> > branch
> > > > is
> > > > > > > > required. Actually only I myself am working on the development
> > > > work
> > > > > > > > (Surely, Aleksey is very helpful on testing), this
> > synchronization
> > > > > > > > might be a nightmare to me. Furthermore, the code on branch
> > will
> > > > not
> > > > > > > > be automatically tested by continuous integration system and
> > BTI,
> > > > I do
> > > > > > > > not want to work without collaboration, that's not an open
> > source
> > > > > > > > style, right? Will you ask a contributor to create a branch
> > and
> > > > play
> > > > > > > > with himself whenever he wants to contribute?
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > > >
> > > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> > t.p.ellison@gmail.com>
> > > > > > wrote:
> > > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > > Yep, Tim, you're right. I believe that new
> > implementation
> > > > fixes
> > > > > > a
> > > > > > > > > > > number of bugs and will try to get it not degrading. I
> > just
> > > > want
> > > > > > to
> > > > > > > > > > > maintain the performance level of current trunk on the
> > same
> > > > > > level,
> > > > > > > > > > > gradually fixing functional bugs. I don't like to
> > sacrifice
> > > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> > That
> > > > is, I
> > > > > > want
> > > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > > >
> > > > > > > > > > > "high performance, minor bug -> high performance, no
> > bugs"
> > > > > > > > > > >
> > > > > > > > > > > rather than
> > > > > > > > > > >
> > > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> > ->
> > > > high
> > > > > > > > > > > performance, no bugs"
> > > > > > > > > > >
> > > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> > for
> > > > > > > > > > > performance measurements at any time.
> > > > > > > > >
> > > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> > suffer
> > > > from
> > > > > > the
> > > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > > >
> > > > > > > > > > > What do you think?
> > > > > > > > >
> > > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > > most
> > > > > > > > > > desirable state.  I actually don't mind some minor
> > performance
> > > > > > > > > > regression on HEAD between releases provided it is an area
> > > > being
> > > > > > > > > > actively worked upon.
> > > > > > > > >
> > > > > > > > > +1 especially if it fixes bugs
> > > > > > > > >
> > > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> > ;-)
> > > > > > > > >
> > > > > > > > > Me too.
> > > > > > > > >
> > > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > > perfect
> > > > > > then
> > > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > > synch.
> > > > > >  You
> > > > > > > > > > could also consider a branch in SVN.
> > > > > > > > >
> > > > > > > > > This bothers me too.  Firstly, while it is being developed
> > in
> > > > patch
> > > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> > at
> > > > it.
> > > > > > > > > Secondly, that progress will be slow because of the cost of
> > > > keeping
> > > > > > in
> > > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > > >
> > > > > > > > > I can't help thinking we'll make more progress if we apply
> > the
> > > > patch
> > > > > > to
> > > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> > new
> > > > code
> > > > > > -
> > > > > > > > > and there are likely issues beyond the performance problems
> > that
> > > > > > have
> > > > > > > > > been the focus so far - and more people will see the benefit
> > of
> > > > Tony
> > > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > > >
> > > > > > > > > I certainly don't want all this work to be completed outside
> > svn
> > > > > > HEAD
> > > > > > > > > and committed a week or two before the freeze for next
> > > > milestone.
> > > > > > > > >
> > > > > > > > > Regards,
> > > > > > > > >  Mark.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > --
> > > > > > > >
> > > > > > > > Tony Wu
> > > > > > > > China Software Development Lab, IBM
> > > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>
>
>
> --
> Sean Qiu
> http://xiaoxia.turendui.com
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Sean Qiu <se...@gmail.com>.
Hi, Tony

I volunteered to fix some of the differences for you, if you are busy.

Best regards.


2008/1/28, Tony Wu <wu...@gmail.com>:
>
> Ok, investigating.
>
> On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > Hi Tony,
> >
> > Please look at another issue connected with the ICU commit.
> > https://issues.apache.org/jira/browse/HARMONY-5435
> > It has not been investigated deeply yet.
> >
> > --Rustem
> >
> >
> > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> >
> > > They are all caused by different implementation of DateFormat, which
> > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > some real product.  See following output, they have the same meaning
> > > but don't exactly equal on text.
> > >
> > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > Expected: 'text here 12:17:01 UTC PM and here'
> > >
> > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > >
> > > Result:   'text here 00-04-05 and here'
> > > Expected: 'text here 05/04/00 and here'
> > >
> > > I'll open another thread to ask people's opinions on this difference.
> > >
> > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > Tony,
> > > >
> > > > It looks like this commit caused a regression. Could you take a look
> at
> > > > HARMONY-5430, please?
> > > >
> > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > >
> > > > > Thanks, Aleksey and all
> > > > >
> > > > > patch committed at r612718. Then I'm going to deal with the
> non-bug
> > > > > difference. Hopefully many legcy bugs/differences can be fixed
> this
> > > > > time.
> > > > >
> > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > Hi, guys!
> > > > > >
> > > > > > Sure, correctness is important, but performance is important
> too. I
> > > > > > had profiled both versions (clean and patched) and see no
> > > significant
> > > > > > difference there: there are more GC happens what I believe
> connected
> > > > > > to internal ICU object creation and such. So, there are no
> obvious
> > > way
> > > > > > to maintain the same performance level, and may be we will try
> again
> > > > > > to eliminate ICU usage from the hotpath of frequently used
> workloads
> > > -
> > > > > > but previous investigation shows it's not that simple.
> > > > > >
> > > > > > Tony, please go ahead with committing this patch, we will deal
> with
> > > > > > ICU performance issues a little later.
> > > > > >
> > > > > > Thanks,
> > > > > > Aleksey.
> > > > > >
> > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > Hi Aleksey,
> > > > > > >
> > > > > > > Thanks for you help.
> > > > > > >
> > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > mentioned in my previous mail. I suppose the different result
> on
> > > your
> > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> POV
> > > is
> > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > have
> > > > > > > to. All of us should be positive on this point. I'd like to
> > > clarify
> > > > > > > the factors I'm facing below.
> > > > > > >
> > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > incorrect. It is not reasonable to keep the code as is and
> refuse
> > > to
> > > > > > > correct it just because the bad version has better
> performance.
> > > IMHO
> > > > > > > performance is nothing if there is no correctness.
> > > > > > >
> > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > method
> > > > > > > calls than we implement it by ourselves, it does harm to
> > > performance
> > > > > > > and can not be worked around. But please do not forget we
> benefit
> > > from
> > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > >
> > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > separate
> > > > > > > in luni and text, I can not guarantee that there is no
> > > modification
> > > > > > > during my work, so the synchronization between HEAD and my
> branch
> > > is
> > > > > > > required. Actually only I myself am working on the development
> > > work
> > > > > > > (Surely, Aleksey is very helpful on testing), this
> synchronization
> > > > > > > might be a nightmare to me. Furthermore, the code on branch
> will
> > > not
> > > > > > > be automatically tested by continuous integration system and
> BTI,
> > > I do
> > > > > > > not want to work without collaboration, that's not an open
> source
> > > > > > > style, right? Will you ask a contributor to create a branch
> and
> > > play
> > > > > > > with himself whenever he wants to contribute?
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > >
> > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> t.p.ellison@gmail.com>
> > > > > wrote:
> > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > Yep, Tim, you're right. I believe that new
> implementation
> > > fixes
> > > > > a
> > > > > > > > > > number of bugs and will try to get it not degrading. I
> just
> > > want
> > > > > to
> > > > > > > > > > maintain the performance level of current trunk on the
> same
> > > > > level,
> > > > > > > > > > gradually fixing functional bugs. I don't like to
> sacrifice
> > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> That
> > > is, I
> > > > > want
> > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> high performance, no
> bugs"
> > > > > > > > > >
> > > > > > > > > > rather than
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> ->
> > > high
> > > > > > > > > > performance, no bugs"
> > > > > > > > > >
> > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> for
> > > > > > > > > > performance measurements at any time.
> > > > > > > >
> > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> suffer
> > > from
> > > > > the
> > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > >
> > > > > > > > > > What do you think?
> > > > > > > >
> > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > most
> > > > > > > > > desirable state.  I actually don't mind some minor
> performance
> > > > > > > > > regression on HEAD between releases provided it is an area
> > > being
> > > > > > > > > actively worked upon.
> > > > > > > >
> > > > > > > > +1 especially if it fixes bugs
> > > > > > > >
> > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> ;-)
> > > > > > > >
> > > > > > > > Me too.
> > > > > > > >
> > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > perfect
> > > > > then
> > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > synch.
> > > > >  You
> > > > > > > > > could also consider a branch in SVN.
> > > > > > > >
> > > > > > > > This bothers me too.  Firstly, while it is being developed
> in
> > > patch
> > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> at
> > > it.
> > > > > > > > Secondly, that progress will be slow because of the cost of
> > > keeping
> > > > > in
> > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > >
> > > > > > > > I can't help thinking we'll make more progress if we apply
> the
> > > patch
> > > > > to
> > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> new
> > > code
> > > > > -
> > > > > > > > and there are likely issues beyond the performance problems
> that
> > > > > have
> > > > > > > > been the focus so far - and more people will see the benefit
> of
> > > Tony
> > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > >
> > > > > > > > I certainly don't want all this work to be completed outside
> svn
> > > > > HEAD
> > > > > > > > and committed a week or two before the freeze for next
> > > milestone.
> > > > > > > >
> > > > > > > > Regards,
> > > > > > > >  Mark.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > >
> > > > > > > Tony Wu
> > > > > > > China Software Development Lab, IBM
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> > >
> > > --
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>



-- 
Sean Qiu
http://xiaoxia.turendui.com

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
llya,

ah. I know what's going on here, fixed at r617008 with regression testcase.

On 1/31/08, Ilya Berezhniuk <il...@gmail.com> wrote:
> Hi Tony,
>
> Thank you for fixing HARMONY-5436; I checked - the failure have disappeared.
> Unfortunately, 19 new errors were introduced by r616678 commit in
> 'jdtcoremodel' EUT suite.
>
> All errors look similar:
> java.lang.NullPointerException
>    at java.text.MessageFormat.format(MessageFormat.java:451)
>    at org.eclipse.jdt.internal.core.util.Messages.bind(Messages.java:388)
>    at org.eclipse.jdt.internal.core.util.Messages.bind(Messages.java:354)
>
> I did not file an issue for this - I can file it if you like.
>
> Thanks,
> Ilya.
>
> 2008/1/30, Rustem Rafikov <r....@gmail.com>:
> >
> > Hi Tony,
> >
> >
> > > Frankly I've have no idea on the error log posted on Harmony-5435. I'm
> > > > not sure what kind of test/tool Rustem is running. Just want to
> > > > confirm with you, I've run specJBB2005 and get the report successfully
> > > > for 5 times. Highly appreciate if there is further infomation.
> >
> >
> > Actually I ran multi-vm configuration on Windows Server 2003. Could you
> > please try multi-vm in such way that each vm instance is assigned on 2
> > cores
> > (eq 2 vm on 4 cores)?
> > It is possible that command  line options also make sense for reproducing
> > the problem. We used rather complex one. I'll try to find simplest command
> > line reproducing the issue and let you know.
> >
> > --
> > Thanks
> > Rustem
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Ilya Berezhniuk <il...@gmail.com>.
Hi Tony,

Thank you for fixing HARMONY-5436; I checked - the failure have disappeared.
Unfortunately, 19 new errors were introduced by r616678 commit in
'jdtcoremodel' EUT suite.

All errors look similar:
java.lang.NullPointerException
    at java.text.MessageFormat.format(MessageFormat.java:451)
    at org.eclipse.jdt.internal.core.util.Messages.bind(Messages.java:388)
    at org.eclipse.jdt.internal.core.util.Messages.bind(Messages.java:354)

I did not file an issue for this - I can file it if you like.

Thanks,
Ilya.

2008/1/30, Rustem Rafikov <r....@gmail.com>:
>
> Hi Tony,
>
>
> > Frankly I've have no idea on the error log posted on Harmony-5435. I'm
> > > not sure what kind of test/tool Rustem is running. Just want to
> > > confirm with you, I've run specJBB2005 and get the report successfully
> > > for 5 times. Highly appreciate if there is further infomation.
>
>
> Actually I ran multi-vm configuration on Windows Server 2003. Could you
> please try multi-vm in such way that each vm instance is assigned on 2
> cores
> (eq 2 vm on 4 cores)?
> It is possible that command  line options also make sense for reproducing
> the problem. We used rather complex one. I'll try to find simplest command
> line reproducing the issue and let you know.
>
> --
> Thanks
> Rustem
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Thanks, Sergey
I think it's useful.

On 2/6/08, Sergey Dmitriev <se...@gmail.com> wrote:
> Hello
>
> Just thought it could be interesting for you guys: I have created a
> mini-test which demonstrates the stability regression at r612718. I
> didn't do much of analysis but it seems we have a synchronization
> issue in ICU.
>
> Please refer to HARMONY-5460.
>
>
> Thanks
> Sergey.
>
>
> On Jan 31, 2008 4:43 PM, Ilya Berezhniuk <il...@gmail.com> wrote:
> > Tony,
> >
> > Thanks a lot!
> > Now MessageFormat.format works perfectly, and EUT 'jdtcoremodel' passes.
> >
> > Thanks,
> > Ilya.
> >
> > 2008/1/31, Rustem Rafikov <r....@gmail.com>:
> >
> > >
> > > Tony,
> > > I'm confused too.
> > > I'll try to investigate and debug the problem from my side as well.
> > >
> > > --bwR
> > >
> > > On Jan 31, 2008 9:08 AM, Tony Wu <wu...@gmail.com> wrote:
> > >
> > > > Rustem,
> > > >
> > > > Do you mean this problem is a multi-vm problem and it does pass on
> > > > sigle vm? I'm confused, since my commit is for locale related classes,
> > > > it's hard for me to imagine the relationship to the number of
> > > > vm...Anyway, I'd like to try this configuration, let me find a
> > > > multi-core machine first.
> > > >
> > > >
> > > >
> > >
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Sergey Dmitriev <se...@gmail.com>.
Hello

Just thought it could be interesting for you guys: I have created a
mini-test which demonstrates the stability regression at r612718. I
didn't do much of analysis but it seems we have a synchronization
issue in ICU.

Please refer to HARMONY-5460.


Thanks
Sergey.


On Jan 31, 2008 4:43 PM, Ilya Berezhniuk <il...@gmail.com> wrote:
> Tony,
>
> Thanks a lot!
> Now MessageFormat.format works perfectly, and EUT 'jdtcoremodel' passes.
>
> Thanks,
> Ilya.
>
> 2008/1/31, Rustem Rafikov <r....@gmail.com>:
>
> >
> > Tony,
> > I'm confused too.
> > I'll try to investigate and debug the problem from my side as well.
> >
> > --bwR
> >
> > On Jan 31, 2008 9:08 AM, Tony Wu <wu...@gmail.com> wrote:
> >
> > > Rustem,
> > >
> > > Do you mean this problem is a multi-vm problem and it does pass on
> > > sigle vm? I'm confused, since my commit is for locale related classes,
> > > it's hard for me to imagine the relationship to the number of
> > > vm...Anyway, I'd like to try this configuration, let me find a
> > > multi-core machine first.
> > >
> > >
> > >
> >
>

Re: [classlib]remove the duplicate locale data

Posted by Ilya Berezhniuk <il...@gmail.com>.
Tony,

Thanks a lot!
Now MessageFormat.format works perfectly, and EUT 'jdtcoremodel' passes.

Thanks,
Ilya.

2008/1/31, Rustem Rafikov <r....@gmail.com>:
>
> Tony,
> I'm confused too.
> I'll try to investigate and debug the problem from my side as well.
>
> --bwR
>
> On Jan 31, 2008 9:08 AM, Tony Wu <wu...@gmail.com> wrote:
>
> > Rustem,
> >
> > Do you mean this problem is a multi-vm problem and it does pass on
> > sigle vm? I'm confused, since my commit is for locale related classes,
> > it's hard for me to imagine the relationship to the number of
> > vm...Anyway, I'd like to try this configuration, let me find a
> > multi-core machine first.
> >
> >
> >
>

Re: [classlib]remove the duplicate locale data

Posted by Rustem Rafikov <r....@gmail.com>.
Tony,
I'm confused too.
I'll try to investigate and debug the problem from my side as well.

--bwR

On Jan 31, 2008 9:08 AM, Tony Wu <wu...@gmail.com> wrote:

> Rustem,
>
> Do you mean this problem is a multi-vm problem and it does pass on
> sigle vm? I'm confused, since my commit is for locale related classes,
> it's hard for me to imagine the relationship to the number of
> vm...Anyway, I'd like to try this configuration, let me find a
> multi-core machine first.
>
>
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Rustem,

Do you mean this problem is a multi-vm problem and it does pass on
sigle vm? I'm confused, since my commit is for locale related classes,
it's hard for me to imagine the relationship to the number of
vm...Anyway, I'd like to try this configuration, let me find a
multi-core machine first.

On 1/31/08, Rustem Rafikov <r....@gmail.com> wrote:
> Hi Tony,
>
>
> > Frankly I've have no idea on the error log posted on Harmony-5435. I'm
> > > not sure what kind of test/tool Rustem is running. Just want to
> > > confirm with you, I've run specJBB2005 and get the report successfully
> > > for 5 times. Highly appreciate if there is further infomation.
>
>
> Actually I ran multi-vm configuration on Windows Server 2003. Could you
> please try multi-vm in such way that each vm instance is assigned on 2 cores
> (eq 2 vm on 4 cores)?
> It is possible that command  line options also make sense for reproducing
> the problem. We used rather complex one. I'll try to find simplest command
> line reproducing the issue and let you know.
>
> --
> Thanks
> Rustem
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Rustem Rafikov <r....@gmail.com>.
Hi Tony,


> Frankly I've have no idea on the error log posted on Harmony-5435. I'm
> > not sure what kind of test/tool Rustem is running. Just want to
> > confirm with you, I've run specJBB2005 and get the report successfully
> > for 5 times. Highly appreciate if there is further infomation.


Actually I ran multi-vm configuration on Windows Server 2003. Could you
please try multi-vm in such way that each vm instance is assigned on 2 cores
(eq 2 vm on 4 cores)?
It is possible that command  line options also make sense for reproducing
the problem. We used rather complex one. I'll try to find simplest command
line reproducing the issue and let you know.

--
Thanks
Rustem

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
On 1/30/08, Aleksey Shipilev <al...@gmail.com> wrote:
> Hi, Tony.
>
> On Jan 30, 2008 11:41 AM, Tony Wu <wu...@gmail.com> wrote:
> > Thanks. It will be helpful if you can give me some clue on this problem.
> I can provide more info on hang-up statistics before and after the
> commit, like Rustem does. BTW, 5-10% of Rustem's runs fails in the
> middle of the run after commit, while 0% fails before commit. That's
> the same thing I observe although I haven't exact statistics, but we
> could gather it in few days.
>
Frankly I've have no idea on the error log posted on Harmony-5435. I'm
not sure what kind of test/tool Rustem is running. Just want to
confirm with you, I've run specJBB2005 and get the report successfully
for 5 times. Highly appreciate if there is further infomation.

> > It's true that I can revert it this time and every time in future, but
> > it does not solve the problem. What I want is to find the root cause
> > and fix it rather than worm out of.
> Sure, you're right. But here's the question on what should reside in
> the mainline trunk, and what should reside in branch or any other
> patchset. Last time we discussed this we came to agreement that end
> user should not experience functional (and as such, stability)
> problems with mainline trunk. I also argue that committing buggy issue
> into mainline and then trying to solve the problems on-the-fly is not
> the best user-oriented practice. It is useful for
> this-feature-developers, right, but it's frustrating for developers of
> other components, when stability of overall build is affected by this
> feature.
>
I believe the trunk is for developer not for end user, correct me if
I'm wrong.  I can not imagine that an opensource project asks their
end user to try their trunk directly, just the other way, they may
claim that the trunk is not a stable version and suggest the user to
download the release version. Probably that's the different opinion
among us :)

> I think it's generally a bad practice not to split stable/non-stable
> branches, even though it brings overheads for syncing the branches.
>
> So, given the perspective we are dealing with one branch, what we should choose:
>  1. Consider mainline as stable branch? Then we should revert your
> commit and maintain it as patchset, while trying to get clue what's
> happening.
>  2. Consider mainline as unstable branch? Then we should keep
> "probably-buggy" commit in trunk and get the things over on-the-fly.
>
> Honestly, I know how to work in (1) scheme only. Scheme (2) naturally
> brings up the idea to create a stable branch, because unstable branch
> seem to be unusable for referencing.
>
> What do you think, what should we do?

I think we have been working with the second way. Otherwise, what does
the Milestone build stand for? We create a tag as M4 and claim it is a
stable verion, we may ask user to try M4 and continue our development
on trunk. That's wyh I happy to revert my patch last time for assuring
the stability of M4.

In the contrast, if we use the trunk as a stable verion. We may need
to create many branches to guarantee that. How to manage the
synchronization from trunk to branch and how to make sure that the
branch is stable enough to be merged back to trunk? Shall we set up a
BTI for every branch? What a mess if everything goes to be that way.

>
> Thanks,
> Aleksey.
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Hi, Tony.

On Jan 30, 2008 11:41 AM, Tony Wu <wu...@gmail.com> wrote:
> Thanks. It will be helpful if you can give me some clue on this problem.
I can provide more info on hang-up statistics before and after the
commit, like Rustem does. BTW, 5-10% of Rustem's runs fails in the
middle of the run after commit, while 0% fails before commit. That's
the same thing I observe although I haven't exact statistics, but we
could gather it in few days.

> It's true that I can revert it this time and every time in future, but
> it does not solve the problem. What I want is to find the root cause
> and fix it rather than worm out of.
Sure, you're right. But here's the question on what should reside in
the mainline trunk, and what should reside in branch or any other
patchset. Last time we discussed this we came to agreement that end
user should not experience functional (and as such, stability)
problems with mainline trunk. I also argue that committing buggy issue
into mainline and then trying to solve the problems on-the-fly is not
the best user-oriented practice. It is useful for
this-feature-developers, right, but it's frustrating for developers of
other components, when stability of overall build is affected by this
feature.

I think it's generally a bad practice not to split stable/non-stable
branches, even though it brings overheads for syncing the branches.

So, given the perspective we are dealing with one branch, what we should choose:
 1. Consider mainline as stable branch? Then we should revert your
commit and maintain it as patchset, while trying to get clue what's
happening.
 2. Consider mainline as unstable branch? Then we should keep
"probably-buggy" commit in trunk and get the things over on-the-fly.

Honestly, I know how to work in (1) scheme only. Scheme (2) naturally
brings up the idea to create a stable branch, because unstable branch
seem to be unusable for referencing.

What do you think, what should we do?

Thanks,
Aleksey.

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Thanks. It will be helpful if you can give me some clue on this problem.

It's true that I can revert it this time and every time in future, but
it does not solve the problem. What I want is to find the root cause
and fix it rather than worm out of.

On 1/30/08, Aleksey Shipilev <al...@gmail.com> wrote:
> Tony,
>
> Rustem had an input on stability earlier.  My runs show that there's
> different types of intermittent hang-ups on SPECjbb2005, presumably
> with this commit, since manual reverting of 612718 commit helps to
> solve this problem.
>
> More to follow.
>
> Thanks,
> Aleksey.
>
> On Jan 30, 2008 11:12 AM, Tony Wu <wu...@gmail.com> wrote:
> > Could you please give me more infomation on the stability? I'd like to
> > help on this.
> >
> >
> > On 1/30/08, Tony Wu <wu...@gmail.com> wrote:
> > > Hi,
> > >
> > > what do you mean about the great reduce on stability?
> > >
> > > I noticed 2 JIRAs here, HARMONY-5436 is a minor difference between RI
> > > and ICU's MessageFormat, I can work around it in 1 minute.
> > >
> > > And HARMONY-5435 is an  intermittent crash but I haven't got any crash
> > > on my winxp desktop when running specJBB for a whole day.
> > >
> > > On 1/30/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > Tony,
> > > >
> > > > Since this patch is known to greatly reduce stability, can we revert
> > > > it for a while?
> > > >
> > > > Thanks,
> > > > Aleksey.
> > > >
> > > > On Jan 28, 2008 7:34 PM, Ilya Berezhniuk <il...@gmail.com> wrote:
> > > > > Hi Tony,
> > > > >
> > > > > I've found yet another issue in EUT caused by r612718.
> > > > > ("Invalid name specified: {0}" is returned instead of "Invalid name
> > > > > specified: null")
> > > > >
> > > > > Issue is filed as HARMONY-5436.
> > > > >
> > > > > Thanks,
> > > > > Ilya.
> > > > >
> > > > > 2008/1/28, Tony Wu <wu...@gmail.com>:
> > > > >
> > > > > >
> > > > > > Ok, investigating.
> > > > > >
> > > > > > On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > > > > > > Hi Tony,
> > > > > > >
> > > > > > > Please look at another issue connected with the ICU commit.
> > > > > > > https://issues.apache.org/jira/browse/HARMONY-5435
> > > > > > > It has not been investigated deeply yet.
> > > > > > >
> > > > > > > --Rustem
> > > > > > >
> > > > > > >
> > > > > > > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > >
> > > > > > > > They are all caused by different implementation of DateFormat, which
> > > > > > > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > > > > > > some real product.  See following output, they have the same meaning
> > > > > > > > but don't exactly equal on text.
> > > > > > > >
> > > > > > > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > > > > > > Expected: 'text here 12:17:01 UTC PM and here'
> > > > > > > >
> > > > > > > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > > > > > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > > > > > > >
> > > > > > > > Result:   'text here 00-04-05 and here'
> > > > > > > > Expected: 'text here 05/04/00 and here'
> > > > > > > >
> > > > > > > > I'll open another thread to ask people's opinions on this difference.
> > > > > > > >
> > > > > > > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > > > > > > Tony,
> > > > > > > > >
> > > > > > > > > It looks like this commit caused a regression. Could you take a look
> > > > > > at
> > > > > > > > > HARMONY-5430, please?
> > > > > > > > >
> > > > > > > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > >
> > > > > > > > > > Thanks, Aleksey and all
> > > > > > > > > >
> > > > > > > > > > patch committed at r612718. Then I'm going to deal with the
> > > > > > non-bug
> > > > > > > > > > difference. Hopefully many legcy bugs/differences can be fixed
> > > > > > this
> > > > > > > > > > time.
> > > > > > > > > >
> > > > > > > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > > > > > > Hi, guys!
> > > > > > > > > > >
> > > > > > > > > > > Sure, correctness is important, but performance is important
> > > > > > too. I
> > > > > > > > > > > had profiled both versions (clean and patched) and see no
> > > > > > > > significant
> > > > > > > > > > > difference there: there are more GC happens what I believe
> > > > > > connected
> > > > > > > > > > > to internal ICU object creation and such. So, there are no
> > > > > > obvious
> > > > > > > > way
> > > > > > > > > > > to maintain the same performance level, and may be we will try
> > > > > > again
> > > > > > > > > > > to eliminate ICU usage from the hotpath of frequently used
> > > > > > workloads
> > > > > > > > -
> > > > > > > > > > > but previous investigation shows it's not that simple.
> > > > > > > > > > >
> > > > > > > > > > > Tony, please go ahead with committing this patch, we will deal
> > > > > > with
> > > > > > > > > > > ICU performance issues a little later.
> > > > > > > > > > >
> > > > > > > > > > > Thanks,
> > > > > > > > > > > Aleksey.
> > > > > > > > > > >
> > > > > > > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > > > Hi Aleksey,
> > > > > > > > > > > >
> > > > > > > > > > > > Thanks for you help.
> > > > > > > > > > > >
> > > > > > > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > > > > > > mentioned in my previous mail. I suppose the different result
> > > > > > on
> > > > > > > > your
> > > > > > > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> > > > > > POV
> > > > > > > > is
> > > > > > > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > > > > > > have
> > > > > > > > > > > > to. All of us should be positive on this point. I'd like to
> > > > > > > > clarify
> > > > > > > > > > > > the factors I'm facing below.
> > > > > > > > > > > >
> > > > > > > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > > > > > > incorrect. It is not reasonable to keep the code as is and
> > > > > > refuse
> > > > > > > > to
> > > > > > > > > > > > correct it just because the bad version has better
> > > > > > performance.
> > > > > > > > IMHO
> > > > > > > > > > > > performance is nothing if there is no correctness.
> > > > > > > > > > > >
> > > > > > > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > > > > > > method
> > > > > > > > > > > > calls than we implement it by ourselves, it does harm to
> > > > > > > > performance
> > > > > > > > > > > > and can not be worked around. But please do not forget we
> > > > > > benefit
> > > > > > > > from
> > > > > > > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > > > > > > >
> > > > > > > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > > > > > > separate
> > > > > > > > > > > > in luni and text, I can not guarantee that there is no
> > > > > > > > modification
> > > > > > > > > > > > during my work, so the synchronization between HEAD and my
> > > > > > branch
> > > > > > > > is
> > > > > > > > > > > > required. Actually only I myself am working on the development
> > > > > > > > work
> > > > > > > > > > > > (Surely, Aleksey is very helpful on testing), this
> > > > > > synchronization
> > > > > > > > > > > > might be a nightmare to me. Furthermore, the code on branch
> > > > > > will
> > > > > > > > not
> > > > > > > > > > > > be automatically tested by continuous integration system and
> > > > > > BTI,
> > > > > > > > I do
> > > > > > > > > > > > not want to work without collaboration, that's not an open
> > > > > > source
> > > > > > > > > > > > style, right? Will you ask a contributor to create a branch
> > > > > > and
> > > > > > > > play
> > > > > > > > > > > > with himself whenever he wants to contribute?
> > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > > > > > > >
> > > > > > > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> > > > > > t.p.ellison@gmail.com>
> > > > > > > > > > wrote:
> > > > > > > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > > > > > > Yep, Tim, you're right. I believe that new
> > > > > > implementation
> > > > > > > > fixes
> > > > > > > > > > a
> > > > > > > > > > > > > > > number of bugs and will try to get it not degrading. I
> > > > > > just
> > > > > > > > want
> > > > > > > > > > to
> > > > > > > > > > > > > > > maintain the performance level of current trunk on the
> > > > > > same
> > > > > > > > > > level,
> > > > > > > > > > > > > > > gradually fixing functional bugs. I don't like to
> > > > > > sacrifice
> > > > > > > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> > > > > > That
> > > > > > > > is, I
> > > > > > > > > > want
> > > > > > > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > "high performance, minor bug -> high performance, no
> > > > > > bugs"
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > rather than
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> > > > > > ->
> > > > > > > > high
> > > > > > > > > > > > > > > performance, no bugs"
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> > > > > > for
> > > > > > > > > > > > > > > performance measurements at any time.
> > > > > > > > > > > > >
> > > > > > > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> > > > > > suffer
> > > > > > > > from
> > > > > > > > > > the
> > > > > > > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > > > > > > >
> > > > > > > > > > > > > > > What do you think?
> > > > > > > > > > > > >
> > > > > > > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > > > > > > most
> > > > > > > > > > > > > > desirable state.  I actually don't mind some minor
> > > > > > performance
> > > > > > > > > > > > > > regression on HEAD between releases provided it is an area
> > > > > > > > being
> > > > > > > > > > > > > > actively worked upon.
> > > > > > > > > > > > >
> > > > > > > > > > > > > +1 especially if it fixes bugs
> > > > > > > > > > > > >
> > > > > > > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> > > > > > ;-)
> > > > > > > > > > > > >
> > > > > > > > > > > > > Me too.
> > > > > > > > > > > > >
> > > > > > > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > > > > > > perfect
> > > > > > > > > > then
> > > > > > > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > > > > > > synch.
> > > > > > > > > >  You
> > > > > > > > > > > > > > could also consider a branch in SVN.
> > > > > > > > > > > > >
> > > > > > > > > > > > > This bothers me too.  Firstly, while it is being developed
> > > > > > in
> > > > > > > > patch
> > > > > > > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> > > > > > at
> > > > > > > > it.
> > > > > > > > > > > > > Secondly, that progress will be slow because of the cost of
> > > > > > > > keeping
> > > > > > > > > > in
> > > > > > > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > > > > > > >
> > > > > > > > > > > > > I can't help thinking we'll make more progress if we apply
> > > > > > the
> > > > > > > > patch
> > > > > > > > > > to
> > > > > > > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> > > > > > new
> > > > > > > > code
> > > > > > > > > > -
> > > > > > > > > > > > > and there are likely issues beyond the performance problems
> > > > > > that
> > > > > > > > > > have
> > > > > > > > > > > > > been the focus so far - and more people will see the benefit
> > > > > > of
> > > > > > > > Tony
> > > > > > > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > > > > > > >
> > > > > > > > > > > > > I certainly don't want all this work to be completed outside
> > > > > > svn
> > > > > > > > > > HEAD
> > > > > > > > > > > > > and committed a week or two before the freeze for next
> > > > > > > > milestone.
> > > > > > > > > > > > >
> > > > > > > > > > > > > Regards,
> > > > > > > > > > > > >  Mark.
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > > --
> > > > > > > > > > > >
> > > > > > > > > > > > Tony Wu
> > > > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > --
> > > > > > > > > > Tony Wu
> > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > --
> > > > > > > > Tony Wu
> > > > > > > > China Software Development Lab, IBM
> > > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > >
> > >
> > > --
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
> >
> > --
> >
> > Tony Wu
> > China Software Development Lab, IBM
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Tony,

Rustem had an input on stability earlier.  My runs show that there's
different types of intermittent hang-ups on SPECjbb2005, presumably
with this commit, since manual reverting of 612718 commit helps to
solve this problem.

More to follow.

Thanks,
Aleksey.

On Jan 30, 2008 11:12 AM, Tony Wu <wu...@gmail.com> wrote:
> Could you please give me more infomation on the stability? I'd like to
> help on this.
>
>
> On 1/30/08, Tony Wu <wu...@gmail.com> wrote:
> > Hi,
> >
> > what do you mean about the great reduce on stability?
> >
> > I noticed 2 JIRAs here, HARMONY-5436 is a minor difference between RI
> > and ICU's MessageFormat, I can work around it in 1 minute.
> >
> > And HARMONY-5435 is an  intermittent crash but I haven't got any crash
> > on my winxp desktop when running specJBB for a whole day.
> >
> > On 1/30/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > Tony,
> > >
> > > Since this patch is known to greatly reduce stability, can we revert
> > > it for a while?
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Jan 28, 2008 7:34 PM, Ilya Berezhniuk <il...@gmail.com> wrote:
> > > > Hi Tony,
> > > >
> > > > I've found yet another issue in EUT caused by r612718.
> > > > ("Invalid name specified: {0}" is returned instead of "Invalid name
> > > > specified: null")
> > > >
> > > > Issue is filed as HARMONY-5436.
> > > >
> > > > Thanks,
> > > > Ilya.
> > > >
> > > > 2008/1/28, Tony Wu <wu...@gmail.com>:
> > > >
> > > > >
> > > > > Ok, investigating.
> > > > >
> > > > > On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > > > > > Hi Tony,
> > > > > >
> > > > > > Please look at another issue connected with the ICU commit.
> > > > > > https://issues.apache.org/jira/browse/HARMONY-5435
> > > > > > It has not been investigated deeply yet.
> > > > > >
> > > > > > --Rustem
> > > > > >
> > > > > >
> > > > > > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > >
> > > > > > > They are all caused by different implementation of DateFormat, which
> > > > > > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > > > > > some real product.  See following output, they have the same meaning
> > > > > > > but don't exactly equal on text.
> > > > > > >
> > > > > > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > > > > > Expected: 'text here 12:17:01 UTC PM and here'
> > > > > > >
> > > > > > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > > > > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > > > > > >
> > > > > > > Result:   'text here 00-04-05 and here'
> > > > > > > Expected: 'text here 05/04/00 and here'
> > > > > > >
> > > > > > > I'll open another thread to ask people's opinions on this difference.
> > > > > > >
> > > > > > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > > > > > Tony,
> > > > > > > >
> > > > > > > > It looks like this commit caused a regression. Could you take a look
> > > > > at
> > > > > > > > HARMONY-5430, please?
> > > > > > > >
> > > > > > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > >
> > > > > > > > > Thanks, Aleksey and all
> > > > > > > > >
> > > > > > > > > patch committed at r612718. Then I'm going to deal with the
> > > > > non-bug
> > > > > > > > > difference. Hopefully many legcy bugs/differences can be fixed
> > > > > this
> > > > > > > > > time.
> > > > > > > > >
> > > > > > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > > > > > Hi, guys!
> > > > > > > > > >
> > > > > > > > > > Sure, correctness is important, but performance is important
> > > > > too. I
> > > > > > > > > > had profiled both versions (clean and patched) and see no
> > > > > > > significant
> > > > > > > > > > difference there: there are more GC happens what I believe
> > > > > connected
> > > > > > > > > > to internal ICU object creation and such. So, there are no
> > > > > obvious
> > > > > > > way
> > > > > > > > > > to maintain the same performance level, and may be we will try
> > > > > again
> > > > > > > > > > to eliminate ICU usage from the hotpath of frequently used
> > > > > workloads
> > > > > > > -
> > > > > > > > > > but previous investigation shows it's not that simple.
> > > > > > > > > >
> > > > > > > > > > Tony, please go ahead with committing this patch, we will deal
> > > > > with
> > > > > > > > > > ICU performance issues a little later.
> > > > > > > > > >
> > > > > > > > > > Thanks,
> > > > > > > > > > Aleksey.
> > > > > > > > > >
> > > > > > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > > Hi Aleksey,
> > > > > > > > > > >
> > > > > > > > > > > Thanks for you help.
> > > > > > > > > > >
> > > > > > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > > > > > mentioned in my previous mail. I suppose the different result
> > > > > on
> > > > > > > your
> > > > > > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> > > > > POV
> > > > > > > is
> > > > > > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > > > > > have
> > > > > > > > > > > to. All of us should be positive on this point. I'd like to
> > > > > > > clarify
> > > > > > > > > > > the factors I'm facing below.
> > > > > > > > > > >
> > > > > > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > > > > > incorrect. It is not reasonable to keep the code as is and
> > > > > refuse
> > > > > > > to
> > > > > > > > > > > correct it just because the bad version has better
> > > > > performance.
> > > > > > > IMHO
> > > > > > > > > > > performance is nothing if there is no correctness.
> > > > > > > > > > >
> > > > > > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > > > > > method
> > > > > > > > > > > calls than we implement it by ourselves, it does harm to
> > > > > > > performance
> > > > > > > > > > > and can not be worked around. But please do not forget we
> > > > > benefit
> > > > > > > from
> > > > > > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > > > > > >
> > > > > > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > > > > > separate
> > > > > > > > > > > in luni and text, I can not guarantee that there is no
> > > > > > > modification
> > > > > > > > > > > during my work, so the synchronization between HEAD and my
> > > > > branch
> > > > > > > is
> > > > > > > > > > > required. Actually only I myself am working on the development
> > > > > > > work
> > > > > > > > > > > (Surely, Aleksey is very helpful on testing), this
> > > > > synchronization
> > > > > > > > > > > might be a nightmare to me. Furthermore, the code on branch
> > > > > will
> > > > > > > not
> > > > > > > > > > > be automatically tested by continuous integration system and
> > > > > BTI,
> > > > > > > I do
> > > > > > > > > > > not want to work without collaboration, that's not an open
> > > > > source
> > > > > > > > > > > style, right? Will you ask a contributor to create a branch
> > > > > and
> > > > > > > play
> > > > > > > > > > > with himself whenever he wants to contribute?
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > > > > > >
> > > > > > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> > > > > t.p.ellison@gmail.com>
> > > > > > > > > wrote:
> > > > > > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > > > > > Yep, Tim, you're right. I believe that new
> > > > > implementation
> > > > > > > fixes
> > > > > > > > > a
> > > > > > > > > > > > > > number of bugs and will try to get it not degrading. I
> > > > > just
> > > > > > > want
> > > > > > > > > to
> > > > > > > > > > > > > > maintain the performance level of current trunk on the
> > > > > same
> > > > > > > > > level,
> > > > > > > > > > > > > > gradually fixing functional bugs. I don't like to
> > > > > sacrifice
> > > > > > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> > > > > That
> > > > > > > is, I
> > > > > > > > > want
> > > > > > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > "high performance, minor bug -> high performance, no
> > > > > bugs"
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > rather than
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> > > > > ->
> > > > > > > high
> > > > > > > > > > > > > > performance, no bugs"
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> > > > > for
> > > > > > > > > > > > > > performance measurements at any time.
> > > > > > > > > > > >
> > > > > > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> > > > > suffer
> > > > > > > from
> > > > > > > > > the
> > > > > > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > > > > > >
> > > > > > > > > > > > > > What do you think?
> > > > > > > > > > > >
> > > > > > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > > > > > most
> > > > > > > > > > > > > desirable state.  I actually don't mind some minor
> > > > > performance
> > > > > > > > > > > > > regression on HEAD between releases provided it is an area
> > > > > > > being
> > > > > > > > > > > > > actively worked upon.
> > > > > > > > > > > >
> > > > > > > > > > > > +1 especially if it fixes bugs
> > > > > > > > > > > >
> > > > > > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> > > > > ;-)
> > > > > > > > > > > >
> > > > > > > > > > > > Me too.
> > > > > > > > > > > >
> > > > > > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > > > > > perfect
> > > > > > > > > then
> > > > > > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > > > > > synch.
> > > > > > > > >  You
> > > > > > > > > > > > > could also consider a branch in SVN.
> > > > > > > > > > > >
> > > > > > > > > > > > This bothers me too.  Firstly, while it is being developed
> > > > > in
> > > > > > > patch
> > > > > > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> > > > > at
> > > > > > > it.
> > > > > > > > > > > > Secondly, that progress will be slow because of the cost of
> > > > > > > keeping
> > > > > > > > > in
> > > > > > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > > > > > >
> > > > > > > > > > > > I can't help thinking we'll make more progress if we apply
> > > > > the
> > > > > > > patch
> > > > > > > > > to
> > > > > > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> > > > > new
> > > > > > > code
> > > > > > > > > -
> > > > > > > > > > > > and there are likely issues beyond the performance problems
> > > > > that
> > > > > > > > > have
> > > > > > > > > > > > been the focus so far - and more people will see the benefit
> > > > > of
> > > > > > > Tony
> > > > > > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > > > > > >
> > > > > > > > > > > > I certainly don't want all this work to be completed outside
> > > > > svn
> > > > > > > > > HEAD
> > > > > > > > > > > > and committed a week or two before the freeze for next
> > > > > > > milestone.
> > > > > > > > > > > >
> > > > > > > > > > > > Regards,
> > > > > > > > > > > >  Mark.
> > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > --
> > > > > > > > > > >
> > > > > > > > > > > Tony Wu
> > > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > --
> > > > > > > > > Tony Wu
> > > > > > > > > China Software Development Lab, IBM
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > > Tony Wu
> > > > > > > China Software Development Lab, IBM
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>
>
> --
>
> Tony Wu
> China Software Development Lab, IBM
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Could you please give me more infomation on the stability? I'd like to
help on this.

On 1/30/08, Tony Wu <wu...@gmail.com> wrote:
> Hi,
>
> what do you mean about the great reduce on stability?
>
> I noticed 2 JIRAs here, HARMONY-5436 is a minor difference between RI
> and ICU's MessageFormat, I can work around it in 1 minute.
>
> And HARMONY-5435 is an  intermittent crash but I haven't got any crash
> on my winxp desktop when running specJBB for a whole day.
>
> On 1/30/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > Tony,
> >
> > Since this patch is known to greatly reduce stability, can we revert
> > it for a while?
> >
> > Thanks,
> > Aleksey.
> >
> > On Jan 28, 2008 7:34 PM, Ilya Berezhniuk <il...@gmail.com> wrote:
> > > Hi Tony,
> > >
> > > I've found yet another issue in EUT caused by r612718.
> > > ("Invalid name specified: {0}" is returned instead of "Invalid name
> > > specified: null")
> > >
> > > Issue is filed as HARMONY-5436.
> > >
> > > Thanks,
> > > Ilya.
> > >
> > > 2008/1/28, Tony Wu <wu...@gmail.com>:
> > >
> > > >
> > > > Ok, investigating.
> > > >
> > > > On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > > > > Hi Tony,
> > > > >
> > > > > Please look at another issue connected with the ICU commit.
> > > > > https://issues.apache.org/jira/browse/HARMONY-5435
> > > > > It has not been investigated deeply yet.
> > > > >
> > > > > --Rustem
> > > > >
> > > > >
> > > > > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > >
> > > > > > They are all caused by different implementation of DateFormat, which
> > > > > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > > > > some real product.  See following output, they have the same meaning
> > > > > > but don't exactly equal on text.
> > > > > >
> > > > > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > > > > Expected: 'text here 12:17:01 UTC PM and here'
> > > > > >
> > > > > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > > > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > > > > >
> > > > > > Result:   'text here 00-04-05 and here'
> > > > > > Expected: 'text here 05/04/00 and here'
> > > > > >
> > > > > > I'll open another thread to ask people's opinions on this difference.
> > > > > >
> > > > > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > > > > Tony,
> > > > > > >
> > > > > > > It looks like this commit caused a regression. Could you take a look
> > > > at
> > > > > > > HARMONY-5430, please?
> > > > > > >
> > > > > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > >
> > > > > > > > Thanks, Aleksey and all
> > > > > > > >
> > > > > > > > patch committed at r612718. Then I'm going to deal with the
> > > > non-bug
> > > > > > > > difference. Hopefully many legcy bugs/differences can be fixed
> > > > this
> > > > > > > > time.
> > > > > > > >
> > > > > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > > > > Hi, guys!
> > > > > > > > >
> > > > > > > > > Sure, correctness is important, but performance is important
> > > > too. I
> > > > > > > > > had profiled both versions (clean and patched) and see no
> > > > > > significant
> > > > > > > > > difference there: there are more GC happens what I believe
> > > > connected
> > > > > > > > > to internal ICU object creation and such. So, there are no
> > > > obvious
> > > > > > way
> > > > > > > > > to maintain the same performance level, and may be we will try
> > > > again
> > > > > > > > > to eliminate ICU usage from the hotpath of frequently used
> > > > workloads
> > > > > > -
> > > > > > > > > but previous investigation shows it's not that simple.
> > > > > > > > >
> > > > > > > > > Tony, please go ahead with committing this patch, we will deal
> > > > with
> > > > > > > > > ICU performance issues a little later.
> > > > > > > > >
> > > > > > > > > Thanks,
> > > > > > > > > Aleksey.
> > > > > > > > >
> > > > > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > Hi Aleksey,
> > > > > > > > > >
> > > > > > > > > > Thanks for you help.
> > > > > > > > > >
> > > > > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > > > > mentioned in my previous mail. I suppose the different result
> > > > on
> > > > > > your
> > > > > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> > > > POV
> > > > > > is
> > > > > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > > > > have
> > > > > > > > > > to. All of us should be positive on this point. I'd like to
> > > > > > clarify
> > > > > > > > > > the factors I'm facing below.
> > > > > > > > > >
> > > > > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > > > > incorrect. It is not reasonable to keep the code as is and
> > > > refuse
> > > > > > to
> > > > > > > > > > correct it just because the bad version has better
> > > > performance.
> > > > > > IMHO
> > > > > > > > > > performance is nothing if there is no correctness.
> > > > > > > > > >
> > > > > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > > > > method
> > > > > > > > > > calls than we implement it by ourselves, it does harm to
> > > > > > performance
> > > > > > > > > > and can not be worked around. But please do not forget we
> > > > benefit
> > > > > > from
> > > > > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > > > > >
> > > > > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > > > > separate
> > > > > > > > > > in luni and text, I can not guarantee that there is no
> > > > > > modification
> > > > > > > > > > during my work, so the synchronization between HEAD and my
> > > > branch
> > > > > > is
> > > > > > > > > > required. Actually only I myself am working on the development
> > > > > > work
> > > > > > > > > > (Surely, Aleksey is very helpful on testing), this
> > > > synchronization
> > > > > > > > > > might be a nightmare to me. Furthermore, the code on branch
> > > > will
> > > > > > not
> > > > > > > > > > be automatically tested by continuous integration system and
> > > > BTI,
> > > > > > I do
> > > > > > > > > > not want to work without collaboration, that's not an open
> > > > source
> > > > > > > > > > style, right? Will you ask a contributor to create a branch
> > > > and
> > > > > > play
> > > > > > > > > > with himself whenever he wants to contribute?
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > > > > >
> > > > > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> > > > t.p.ellison@gmail.com>
> > > > > > > > wrote:
> > > > > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > > > > Yep, Tim, you're right. I believe that new
> > > > implementation
> > > > > > fixes
> > > > > > > > a
> > > > > > > > > > > > > number of bugs and will try to get it not degrading. I
> > > > just
> > > > > > want
> > > > > > > > to
> > > > > > > > > > > > > maintain the performance level of current trunk on the
> > > > same
> > > > > > > > level,
> > > > > > > > > > > > > gradually fixing functional bugs. I don't like to
> > > > sacrifice
> > > > > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> > > > That
> > > > > > is, I
> > > > > > > > want
> > > > > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > > > > >
> > > > > > > > > > > > > "high performance, minor bug -> high performance, no
> > > > bugs"
> > > > > > > > > > > > >
> > > > > > > > > > > > > rather than
> > > > > > > > > > > > >
> > > > > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> > > > ->
> > > > > > high
> > > > > > > > > > > > > performance, no bugs"
> > > > > > > > > > > > >
> > > > > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> > > > for
> > > > > > > > > > > > > performance measurements at any time.
> > > > > > > > > > >
> > > > > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> > > > suffer
> > > > > > from
> > > > > > > > the
> > > > > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > > > > >
> > > > > > > > > > > > > What do you think?
> > > > > > > > > > >
> > > > > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > > > > most
> > > > > > > > > > > > desirable state.  I actually don't mind some minor
> > > > performance
> > > > > > > > > > > > regression on HEAD between releases provided it is an area
> > > > > > being
> > > > > > > > > > > > actively worked upon.
> > > > > > > > > > >
> > > > > > > > > > > +1 especially if it fixes bugs
> > > > > > > > > > >
> > > > > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> > > > ;-)
> > > > > > > > > > >
> > > > > > > > > > > Me too.
> > > > > > > > > > >
> > > > > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > > > > perfect
> > > > > > > > then
> > > > > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > > > > synch.
> > > > > > > >  You
> > > > > > > > > > > > could also consider a branch in SVN.
> > > > > > > > > > >
> > > > > > > > > > > This bothers me too.  Firstly, while it is being developed
> > > > in
> > > > > > patch
> > > > > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> > > > at
> > > > > > it.
> > > > > > > > > > > Secondly, that progress will be slow because of the cost of
> > > > > > keeping
> > > > > > > > in
> > > > > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > > > > >
> > > > > > > > > > > I can't help thinking we'll make more progress if we apply
> > > > the
> > > > > > patch
> > > > > > > > to
> > > > > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> > > > new
> > > > > > code
> > > > > > > > -
> > > > > > > > > > > and there are likely issues beyond the performance problems
> > > > that
> > > > > > > > have
> > > > > > > > > > > been the focus so far - and more people will see the benefit
> > > > of
> > > > > > Tony
> > > > > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > > > > >
> > > > > > > > > > > I certainly don't want all this work to be completed outside
> > > > svn
> > > > > > > > HEAD
> > > > > > > > > > > and committed a week or two before the freeze for next
> > > > > > milestone.
> > > > > > > > > > >
> > > > > > > > > > > Regards,
> > > > > > > > > > >  Mark.
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > --
> > > > > > > > > >
> > > > > > > > > > Tony Wu
> > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > --
> > > > > > > > Tony Wu
> > > > > > > > China Software Development Lab, IBM
> > > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Hi,

what do you mean about the great reduce on stability?

I noticed 2 JIRAs here, HARMONY-5436 is a minor difference between RI
and ICU's MessageFormat, I can work around it in 1 minute.

And HARMONY-5435 is an  intermittent crash but I haven't got any crash
on my winxp desktop when running specJBB for a whole day.

On 1/30/08, Aleksey Shipilev <al...@gmail.com> wrote:
> Tony,
>
> Since this patch is known to greatly reduce stability, can we revert
> it for a while?
>
> Thanks,
> Aleksey.
>
> On Jan 28, 2008 7:34 PM, Ilya Berezhniuk <il...@gmail.com> wrote:
> > Hi Tony,
> >
> > I've found yet another issue in EUT caused by r612718.
> > ("Invalid name specified: {0}" is returned instead of "Invalid name
> > specified: null")
> >
> > Issue is filed as HARMONY-5436.
> >
> > Thanks,
> > Ilya.
> >
> > 2008/1/28, Tony Wu <wu...@gmail.com>:
> >
> > >
> > > Ok, investigating.
> > >
> > > On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > > > Hi Tony,
> > > >
> > > > Please look at another issue connected with the ICU commit.
> > > > https://issues.apache.org/jira/browse/HARMONY-5435
> > > > It has not been investigated deeply yet.
> > > >
> > > > --Rustem
> > > >
> > > >
> > > > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> > > >
> > > > > They are all caused by different implementation of DateFormat, which
> > > > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > > > some real product.  See following output, they have the same meaning
> > > > > but don't exactly equal on text.
> > > > >
> > > > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > > > Expected: 'text here 12:17:01 UTC PM and here'
> > > > >
> > > > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > > > >
> > > > > Result:   'text here 00-04-05 and here'
> > > > > Expected: 'text here 05/04/00 and here'
> > > > >
> > > > > I'll open another thread to ask people's opinions on this difference.
> > > > >
> > > > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > > > Tony,
> > > > > >
> > > > > > It looks like this commit caused a regression. Could you take a look
> > > at
> > > > > > HARMONY-5430, please?
> > > > > >
> > > > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > > >
> > > > > > > Thanks, Aleksey and all
> > > > > > >
> > > > > > > patch committed at r612718. Then I'm going to deal with the
> > > non-bug
> > > > > > > difference. Hopefully many legcy bugs/differences can be fixed
> > > this
> > > > > > > time.
> > > > > > >
> > > > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > > > Hi, guys!
> > > > > > > >
> > > > > > > > Sure, correctness is important, but performance is important
> > > too. I
> > > > > > > > had profiled both versions (clean and patched) and see no
> > > > > significant
> > > > > > > > difference there: there are more GC happens what I believe
> > > connected
> > > > > > > > to internal ICU object creation and such. So, there are no
> > > obvious
> > > > > way
> > > > > > > > to maintain the same performance level, and may be we will try
> > > again
> > > > > > > > to eliminate ICU usage from the hotpath of frequently used
> > > workloads
> > > > > -
> > > > > > > > but previous investigation shows it's not that simple.
> > > > > > > >
> > > > > > > > Tony, please go ahead with committing this patch, we will deal
> > > with
> > > > > > > > ICU performance issues a little later.
> > > > > > > >
> > > > > > > > Thanks,
> > > > > > > > Aleksey.
> > > > > > > >
> > > > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > Hi Aleksey,
> > > > > > > > >
> > > > > > > > > Thanks for you help.
> > > > > > > > >
> > > > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > > > mentioned in my previous mail. I suppose the different result
> > > on
> > > > > your
> > > > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> > > POV
> > > > > is
> > > > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > > > have
> > > > > > > > > to. All of us should be positive on this point. I'd like to
> > > > > clarify
> > > > > > > > > the factors I'm facing below.
> > > > > > > > >
> > > > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > > > incorrect. It is not reasonable to keep the code as is and
> > > refuse
> > > > > to
> > > > > > > > > correct it just because the bad version has better
> > > performance.
> > > > > IMHO
> > > > > > > > > performance is nothing if there is no correctness.
> > > > > > > > >
> > > > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > > > method
> > > > > > > > > calls than we implement it by ourselves, it does harm to
> > > > > performance
> > > > > > > > > and can not be worked around. But please do not forget we
> > > benefit
> > > > > from
> > > > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > > > >
> > > > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > > > separate
> > > > > > > > > in luni and text, I can not guarantee that there is no
> > > > > modification
> > > > > > > > > during my work, so the synchronization between HEAD and my
> > > branch
> > > > > is
> > > > > > > > > required. Actually only I myself am working on the development
> > > > > work
> > > > > > > > > (Surely, Aleksey is very helpful on testing), this
> > > synchronization
> > > > > > > > > might be a nightmare to me. Furthermore, the code on branch
> > > will
> > > > > not
> > > > > > > > > be automatically tested by continuous integration system and
> > > BTI,
> > > > > I do
> > > > > > > > > not want to work without collaboration, that's not an open
> > > source
> > > > > > > > > style, right? Will you ask a contributor to create a branch
> > > and
> > > > > play
> > > > > > > > > with himself whenever he wants to contribute?
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > > > >
> > > > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> > > t.p.ellison@gmail.com>
> > > > > > > wrote:
> > > > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > > > Yep, Tim, you're right. I believe that new
> > > implementation
> > > > > fixes
> > > > > > > a
> > > > > > > > > > > > number of bugs and will try to get it not degrading. I
> > > just
> > > > > want
> > > > > > > to
> > > > > > > > > > > > maintain the performance level of current trunk on the
> > > same
> > > > > > > level,
> > > > > > > > > > > > gradually fixing functional bugs. I don't like to
> > > sacrifice
> > > > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> > > That
> > > > > is, I
> > > > > > > want
> > > > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > > > >
> > > > > > > > > > > > "high performance, minor bug -> high performance, no
> > > bugs"
> > > > > > > > > > > >
> > > > > > > > > > > > rather than
> > > > > > > > > > > >
> > > > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> > > ->
> > > > > high
> > > > > > > > > > > > performance, no bugs"
> > > > > > > > > > > >
> > > > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> > > for
> > > > > > > > > > > > performance measurements at any time.
> > > > > > > > > >
> > > > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> > > suffer
> > > > > from
> > > > > > > the
> > > > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > > > >
> > > > > > > > > > > > What do you think?
> > > > > > > > > >
> > > > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > > > most
> > > > > > > > > > > desirable state.  I actually don't mind some minor
> > > performance
> > > > > > > > > > > regression on HEAD between releases provided it is an area
> > > > > being
> > > > > > > > > > > actively worked upon.
> > > > > > > > > >
> > > > > > > > > > +1 especially if it fixes bugs
> > > > > > > > > >
> > > > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> > > ;-)
> > > > > > > > > >
> > > > > > > > > > Me too.
> > > > > > > > > >
> > > > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > > > perfect
> > > > > > > then
> > > > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > > > synch.
> > > > > > >  You
> > > > > > > > > > > could also consider a branch in SVN.
> > > > > > > > > >
> > > > > > > > > > This bothers me too.  Firstly, while it is being developed
> > > in
> > > > > patch
> > > > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> > > at
> > > > > it.
> > > > > > > > > > Secondly, that progress will be slow because of the cost of
> > > > > keeping
> > > > > > > in
> > > > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > > > >
> > > > > > > > > > I can't help thinking we'll make more progress if we apply
> > > the
> > > > > patch
> > > > > > > to
> > > > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> > > new
> > > > > code
> > > > > > > -
> > > > > > > > > > and there are likely issues beyond the performance problems
> > > that
> > > > > > > have
> > > > > > > > > > been the focus so far - and more people will see the benefit
> > > of
> > > > > Tony
> > > > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > > > >
> > > > > > > > > > I certainly don't want all this work to be completed outside
> > > svn
> > > > > > > HEAD
> > > > > > > > > > and committed a week or two before the freeze for next
> > > > > milestone.
> > > > > > > > > >
> > > > > > > > > > Regards,
> > > > > > > > > >  Mark.
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > --
> > > > > > > > >
> > > > > > > > > Tony Wu
> > > > > > > > > China Software Development Lab, IBM
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > > Tony Wu
> > > > > > > China Software Development Lab, IBM
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> > >
> > > --
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Tony,

Since this patch is known to greatly reduce stability, can we revert
it for a while?

Thanks,
Aleksey.

On Jan 28, 2008 7:34 PM, Ilya Berezhniuk <il...@gmail.com> wrote:
> Hi Tony,
>
> I've found yet another issue in EUT caused by r612718.
> ("Invalid name specified: {0}" is returned instead of "Invalid name
> specified: null")
>
> Issue is filed as HARMONY-5436.
>
> Thanks,
> Ilya.
>
> 2008/1/28, Tony Wu <wu...@gmail.com>:
>
> >
> > Ok, investigating.
> >
> > On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > > Hi Tony,
> > >
> > > Please look at another issue connected with the ICU commit.
> > > https://issues.apache.org/jira/browse/HARMONY-5435
> > > It has not been investigated deeply yet.
> > >
> > > --Rustem
> > >
> > >
> > > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> > >
> > > > They are all caused by different implementation of DateFormat, which
> > > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > > some real product.  See following output, they have the same meaning
> > > > but don't exactly equal on text.
> > > >
> > > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > > Expected: 'text here 12:17:01 UTC PM and here'
> > > >
> > > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > > >
> > > > Result:   'text here 00-04-05 and here'
> > > > Expected: 'text here 05/04/00 and here'
> > > >
> > > > I'll open another thread to ask people's opinions on this difference.
> > > >
> > > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > > Tony,
> > > > >
> > > > > It looks like this commit caused a regression. Could you take a look
> > at
> > > > > HARMONY-5430, please?
> > > > >
> > > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > >
> > > > > > Thanks, Aleksey and all
> > > > > >
> > > > > > patch committed at r612718. Then I'm going to deal with the
> > non-bug
> > > > > > difference. Hopefully many legcy bugs/differences can be fixed
> > this
> > > > > > time.
> > > > > >
> > > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > > Hi, guys!
> > > > > > >
> > > > > > > Sure, correctness is important, but performance is important
> > too. I
> > > > > > > had profiled both versions (clean and patched) and see no
> > > > significant
> > > > > > > difference there: there are more GC happens what I believe
> > connected
> > > > > > > to internal ICU object creation and such. So, there are no
> > obvious
> > > > way
> > > > > > > to maintain the same performance level, and may be we will try
> > again
> > > > > > > to eliminate ICU usage from the hotpath of frequently used
> > workloads
> > > > -
> > > > > > > but previous investigation shows it's not that simple.
> > > > > > >
> > > > > > > Tony, please go ahead with committing this patch, we will deal
> > with
> > > > > > > ICU performance issues a little later.
> > > > > > >
> > > > > > > Thanks,
> > > > > > > Aleksey.
> > > > > > >
> > > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > Hi Aleksey,
> > > > > > > >
> > > > > > > > Thanks for you help.
> > > > > > > >
> > > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > > mentioned in my previous mail. I suppose the different result
> > on
> > > > your
> > > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> > POV
> > > > is
> > > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > > have
> > > > > > > > to. All of us should be positive on this point. I'd like to
> > > > clarify
> > > > > > > > the factors I'm facing below.
> > > > > > > >
> > > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > > incorrect. It is not reasonable to keep the code as is and
> > refuse
> > > > to
> > > > > > > > correct it just because the bad version has better
> > performance.
> > > > IMHO
> > > > > > > > performance is nothing if there is no correctness.
> > > > > > > >
> > > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > > method
> > > > > > > > calls than we implement it by ourselves, it does harm to
> > > > performance
> > > > > > > > and can not be worked around. But please do not forget we
> > benefit
> > > > from
> > > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > > >
> > > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > > separate
> > > > > > > > in luni and text, I can not guarantee that there is no
> > > > modification
> > > > > > > > during my work, so the synchronization between HEAD and my
> > branch
> > > > is
> > > > > > > > required. Actually only I myself am working on the development
> > > > work
> > > > > > > > (Surely, Aleksey is very helpful on testing), this
> > synchronization
> > > > > > > > might be a nightmare to me. Furthermore, the code on branch
> > will
> > > > not
> > > > > > > > be automatically tested by continuous integration system and
> > BTI,
> > > > I do
> > > > > > > > not want to work without collaboration, that's not an open
> > source
> > > > > > > > style, right? Will you ask a contributor to create a branch
> > and
> > > > play
> > > > > > > > with himself whenever he wants to contribute?
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > > >
> > > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> > t.p.ellison@gmail.com>
> > > > > > wrote:
> > > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > > Yep, Tim, you're right. I believe that new
> > implementation
> > > > fixes
> > > > > > a
> > > > > > > > > > > number of bugs and will try to get it not degrading. I
> > just
> > > > want
> > > > > > to
> > > > > > > > > > > maintain the performance level of current trunk on the
> > same
> > > > > > level,
> > > > > > > > > > > gradually fixing functional bugs. I don't like to
> > sacrifice
> > > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> > That
> > > > is, I
> > > > > > want
> > > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > > >
> > > > > > > > > > > "high performance, minor bug -> high performance, no
> > bugs"
> > > > > > > > > > >
> > > > > > > > > > > rather than
> > > > > > > > > > >
> > > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> > ->
> > > > high
> > > > > > > > > > > performance, no bugs"
> > > > > > > > > > >
> > > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> > for
> > > > > > > > > > > performance measurements at any time.
> > > > > > > > >
> > > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> > suffer
> > > > from
> > > > > > the
> > > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > > >
> > > > > > > > > > > What do you think?
> > > > > > > > >
> > > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > > most
> > > > > > > > > > desirable state.  I actually don't mind some minor
> > performance
> > > > > > > > > > regression on HEAD between releases provided it is an area
> > > > being
> > > > > > > > > > actively worked upon.
> > > > > > > > >
> > > > > > > > > +1 especially if it fixes bugs
> > > > > > > > >
> > > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> > ;-)
> > > > > > > > >
> > > > > > > > > Me too.
> > > > > > > > >
> > > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > > perfect
> > > > > > then
> > > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > > synch.
> > > > > >  You
> > > > > > > > > > could also consider a branch in SVN.
> > > > > > > > >
> > > > > > > > > This bothers me too.  Firstly, while it is being developed
> > in
> > > > patch
> > > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> > at
> > > > it.
> > > > > > > > > Secondly, that progress will be slow because of the cost of
> > > > keeping
> > > > > > in
> > > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > > >
> > > > > > > > > I can't help thinking we'll make more progress if we apply
> > the
> > > > patch
> > > > > > to
> > > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> > new
> > > > code
> > > > > > -
> > > > > > > > > and there are likely issues beyond the performance problems
> > that
> > > > > > have
> > > > > > > > > been the focus so far - and more people will see the benefit
> > of
> > > > Tony
> > > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > > >
> > > > > > > > > I certainly don't want all this work to be completed outside
> > svn
> > > > > > HEAD
> > > > > > > > > and committed a week or two before the freeze for next
> > > > milestone.
> > > > > > > > >
> > > > > > > > > Regards,
> > > > > > > > >  Mark.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > --
> > > > > > > >
> > > > > > > > Tony Wu
> > > > > > > > China Software Development Lab, IBM
> > > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>

Re: [classlib]remove the duplicate locale data

Posted by Ilya Berezhniuk <il...@gmail.com>.
Hi Tony,

I've found yet another issue in EUT caused by r612718.
("Invalid name specified: {0}" is returned instead of "Invalid name
specified: null")

Issue is filed as HARMONY-5436.

Thanks,
Ilya.

2008/1/28, Tony Wu <wu...@gmail.com>:
>
> Ok, investigating.
>
> On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> > Hi Tony,
> >
> > Please look at another issue connected with the ICU commit.
> > https://issues.apache.org/jira/browse/HARMONY-5435
> > It has not been investigated deeply yet.
> >
> > --Rustem
> >
> >
> > On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
> >
> > > They are all caused by different implementation of DateFormat, which
> > > I've recorded on wiki. I think they are acceptable unless it breaks
> > > some real product.  See following output, they have the same meaning
> > > but don't exactly equal on text.
> > >
> > > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > > Expected: 'text here 12:17:01 UTC PM and here'
> > >
> > > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> > >
> > > Result:   'text here 00-04-05 and here'
> > > Expected: 'text here 05/04/00 and here'
> > >
> > > I'll open another thread to ask people's opinions on this difference.
> > >
> > > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > > Tony,
> > > >
> > > > It looks like this commit caused a regression. Could you take a look
> at
> > > > HARMONY-5430, please?
> > > >
> > > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > > >
> > > > > Thanks, Aleksey and all
> > > > >
> > > > > patch committed at r612718. Then I'm going to deal with the
> non-bug
> > > > > difference. Hopefully many legcy bugs/differences can be fixed
> this
> > > > > time.
> > > > >
> > > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > > Hi, guys!
> > > > > >
> > > > > > Sure, correctness is important, but performance is important
> too. I
> > > > > > had profiled both versions (clean and patched) and see no
> > > significant
> > > > > > difference there: there are more GC happens what I believe
> connected
> > > > > > to internal ICU object creation and such. So, there are no
> obvious
> > > way
> > > > > > to maintain the same performance level, and may be we will try
> again
> > > > > > to eliminate ICU usage from the hotpath of frequently used
> workloads
> > > -
> > > > > > but previous investigation shows it's not that simple.
> > > > > >
> > > > > > Tony, please go ahead with committing this patch, we will deal
> with
> > > > > > ICU performance issues a little later.
> > > > > >
> > > > > > Thanks,
> > > > > > Aleksey.
> > > > > >
> > > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > Hi Aleksey,
> > > > > > >
> > > > > > > Thanks for you help.
> > > > > > >
> > > > > > > I've proved there is no performance degradation on my machine
> > > > > > > mentioned in my previous mail. I suppose the different result
> on
> > > your
> > > > > > > machine is caused by different options to SpecJBB. Anyway, my
> POV
> > > is
> > > > > > > that we should be willing to pay for the adoption of ICU if we
> > > have
> > > > > > > to. All of us should be positive on this point. I'd like to
> > > clarify
> > > > > > > the factors I'm facing below.
> > > > > > >
> > > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > > incorrect. It is not reasonable to keep the code as is and
> refuse
> > > to
> > > > > > > correct it just because the bad version has better
> performance.
> > > IMHO
> > > > > > > performance is nothing if there is no correctness.
> > > > > > >
> > > > > > > Secondly, we adopt ICU through delegation which involves extra
> > > method
> > > > > > > calls than we implement it by ourselves, it does harm to
> > > performance
> > > > > > > and can not be worked around. But please do not forget we
> benefit
> > > from
> > > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > > >
> > > > > > > Lastly, branching does not make sense to me. My fix is very
> > > separate
> > > > > > > in luni and text, I can not guarantee that there is no
> > > modification
> > > > > > > during my work, so the synchronization between HEAD and my
> branch
> > > is
> > > > > > > required. Actually only I myself am working on the development
> > > work
> > > > > > > (Surely, Aleksey is very helpful on testing), this
> synchronization
> > > > > > > might be a nightmare to me. Furthermore, the code on branch
> will
> > > not
> > > > > > > be automatically tested by continuous integration system and
> BTI,
> > > I do
> > > > > > > not want to work without collaboration, that's not an open
> source
> > > > > > > style, right? Will you ask a contributor to create a branch
> and
> > > play
> > > > > > > with himself whenever he wants to contribute?
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > > >
> > > > > > > > On 14 January 2008 at 16:21, Tim Ellison <
> t.p.ellison@gmail.com>
> > > > > wrote:
> > > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > > Yep, Tim, you're right. I believe that new
> implementation
> > > fixes
> > > > > a
> > > > > > > > > > number of bugs and will try to get it not degrading. I
> just
> > > want
> > > > > to
> > > > > > > > > > maintain the performance level of current trunk on the
> same
> > > > > level,
> > > > > > > > > > gradually fixing functional bugs. I don't like to
> sacrifice
> > > > > > > > > > performance of HEAD revision for non-critical bugfix.
> That
> > > is, I
> > > > > want
> > > > > > > > > > to see HEAD changes like this:
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> high performance, no
> bugs"
> > > > > > > > > >
> > > > > > > > > > rather than
> > > > > > > > > >
> > > > > > > > > > "high performance, minor bug -> low performance, no bugs
> ->
> > > high
> > > > > > > > > > performance, no bugs"
> > > > > > > > > >
> > > > > > > > > > ...because anyone could get the HEAD Harmony revision
> for
> > > > > > > > > > performance measurements at any time.
> > > > > > > >
> > > > > > > > Couldn't someone also get the HEAD Harmony revision and
> suffer
> > > from
> > > > > the
> > > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > > >
> > > > > > > > > > What do you think?
> > > > > > > >
> > > > > > > > > For sure, improving performance and fixing the bugs is the
> > > most
> > > > > > > > > desirable state.  I actually don't mind some minor
> performance
> > > > > > > > > regression on HEAD between releases provided it is an area
> > > being
> > > > > > > > > actively worked upon.
> > > > > > > >
> > > > > > > > +1 especially if it fixes bugs
> > > > > > > >
> > > > > > > > > I'd also like to get to 4.2Mb source code reduction too
> ;-)
> > > > > > > >
> > > > > > > > Me too.
> > > > > > > >
> > > > > > > > > If you and Tony are happy to work on the patch to get it
> > > perfect
> > > > > then
> > > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > > synch.
> > > > >  You
> > > > > > > > > could also consider a branch in SVN.
> > > > > > > >
> > > > > > > > This bothers me too.  Firstly, while it is being developed
> in
> > > patch
> > > > > > > > on JIRA, it is likely only Tony and Aleksey will really look
> at
> > > it.
> > > > > > > > Secondly, that progress will be slow because of the cost of
> > > keeping
> > > > > in
> > > > > > > > sync - this applies to an SVN branch too.
> > > > > > > >
> > > > > > > > I can't help thinking we'll make more progress if we apply
> the
> > > patch
> > > > > to
> > > > > > > > HEAD now.  We'll get wider visibility of problems with the
> new
> > > code
> > > > > -
> > > > > > > > and there are likely issues beyond the performance problems
> that
> > > > > have
> > > > > > > > been the focus so far - and more people will see the benefit
> of
> > > Tony
> > > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > > >
> > > > > > > > I certainly don't want all this work to be completed outside
> svn
> > > > > HEAD
> > > > > > > > and committed a week or two before the freeze for next
> > > milestone.
> > > > > > > >
> > > > > > > > Regards,
> > > > > > > >  Mark.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > >
> > > > > > > Tony Wu
> > > > > > > China Software Development Lab, IBM
> > > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> > >
> > > --
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Ok, investigating.

On 1/28/08, Rustem Rafikov <r....@gmail.com> wrote:
> Hi Tony,
>
> Please look at another issue connected with the ICU commit.
> https://issues.apache.org/jira/browse/HARMONY-5435
> It has not been investigated deeply yet.
>
> --Rustem
>
>
> On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:
>
> > They are all caused by different implementation of DateFormat, which
> > I've recorded on wiki. I think they are acceptable unless it breaks
> > some real product.  See following output, they have the same meaning
> > but don't exactly equal on text.
> >
> > Result:   'text here 12:17:01 PM GMT+00:00 and here'
> > Expected: 'text here 12:17:01 UTC PM and here'
> >
> > Result:   'text here 3:16:03 PM GMT+00:00 and here'
> > Expected: 'text here 3:16:03 o'clock PM UTC and here'
> >
> > Result:   'text here 00-04-05 and here'
> > Expected: 'text here 05/04/00 and here'
> >
> > I'll open another thread to ask people's opinions on this difference.
> >
> > On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > > Tony,
> > >
> > > It looks like this commit caused a regression. Could you take a look at
> > > HARMONY-5430, please?
> > >
> > > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > >
> > > > Thanks, Aleksey and all
> > > >
> > > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > > time.
> > > >
> > > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > > Hi, guys!
> > > > >
> > > > > Sure, correctness is important, but performance is important too. I
> > > > > had profiled both versions (clean and patched) and see no
> > significant
> > > > > difference there: there are more GC happens what I believe connected
> > > > > to internal ICU object creation and such. So, there are no obvious
> > way
> > > > > to maintain the same performance level, and may be we will try again
> > > > > to eliminate ICU usage from the hotpath of frequently used workloads
> > -
> > > > > but previous investigation shows it's not that simple.
> > > > >
> > > > > Tony, please go ahead with committing this patch, we will deal with
> > > > > ICU performance issues a little later.
> > > > >
> > > > > Thanks,
> > > > > Aleksey.
> > > > >
> > > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > Hi Aleksey,
> > > > > >
> > > > > > Thanks for you help.
> > > > > >
> > > > > > I've proved there is no performance degradation on my machine
> > > > > > mentioned in my previous mail. I suppose the different result on
> > your
> > > > > > machine is caused by different options to SpecJBB. Anyway, my POV
> > is
> > > > > > that we should be willing to pay for the adoption of ICU if we
> > have
> > > > > > to. All of us should be positive on this point. I'd like to
> > clarify
> > > > > > the factors I'm facing below.
> > > > > >
> > > > > > Firstly, the original implementation of harmony is faster but
> > > > > > incorrect. It is not reasonable to keep the code as is and refuse
> > to
> > > > > > correct it just because the bad version has better performance.
> > IMHO
> > > > > > performance is nothing if there is no correctness.
> > > > > >
> > > > > > Secondly, we adopt ICU through delegation which involves extra
> > method
> > > > > > calls than we implement it by ourselves, it does harm to
> > performance
> > > > > > and can not be worked around. But please do not forget we benefit
> > from
> > > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > > >
> > > > > > Lastly, branching does not make sense to me. My fix is very
> > separate
> > > > > > in luni and text, I can not guarantee that there is no
> > modification
> > > > > > during my work, so the synchronization between HEAD and my branch
> > is
> > > > > > required. Actually only I myself am working on the development
> > work
> > > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > > might be a nightmare to me. Furthermore, the code on branch will
> > not
> > > > > > be automatically tested by continuous integration system and BTI,
> > I do
> > > > > > not want to work without collaboration, that's not an open source
> > > > > > style, right? Will you ask a contributor to create a branch and
> > play
> > > > > > with himself whenever he wants to contribute?
> > > > > >
> > > > > >
> > > > > >
> > > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > > >
> > > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com>
> > > > wrote:
> > > > > > > > Aleksey Shipilev wrote:
> > > > > > > > > Yep, Tim, you're right. I believe that new implementation
> > fixes
> > > > a
> > > > > > > > > number of bugs and will try to get it not degrading. I just
> > want
> > > > to
> > > > > > > > > maintain the performance level of current trunk on the same
> > > > level,
> > > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > > performance of HEAD revision for non-critical bugfix. That
> > is, I
> > > > want
> > > > > > > > > to see HEAD changes like this:
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > > >
> > > > > > > > > rather than
> > > > > > > > >
> > > > > > > > > "high performance, minor bug -> low performance, no bugs ->
> > high
> > > > > > > > > performance, no bugs"
> > > > > > > > >
> > > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > > performance measurements at any time.
> > > > > > >
> > > > > > > Couldn't someone also get the HEAD Harmony revision and suffer
> > from
> > > > the
> > > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > > >
> > > > > > > > > What do you think?
> > > > > > >
> > > > > > > > For sure, improving performance and fixing the bugs is the
> > most
> > > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > > regression on HEAD between releases provided it is an area
> > being
> > > > > > > > actively worked upon.
> > > > > > >
> > > > > > > +1 especially if it fixes bugs
> > > > > > >
> > > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > > >
> > > > > > > Me too.
> > > > > > >
> > > > > > > > If you and Tony are happy to work on the patch to get it
> > perfect
> > > > then
> > > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> > synch.
> > > >  You
> > > > > > > > could also consider a branch in SVN.
> > > > > > >
> > > > > > > This bothers me too.  Firstly, while it is being developed in
> > patch
> > > > > > > on JIRA, it is likely only Tony and Aleksey will really look at
> > it.
> > > > > > > Secondly, that progress will be slow because of the cost of
> > keeping
> > > > in
> > > > > > > sync - this applies to an SVN branch too.
> > > > > > >
> > > > > > > I can't help thinking we'll make more progress if we apply the
> > patch
> > > > to
> > > > > > > HEAD now.  We'll get wider visibility of problems with the new
> > code
> > > > -
> > > > > > > and there are likely issues beyond the performance problems that
> > > > have
> > > > > > > been the focus so far - and more people will see the benefit of
> > Tony
> > > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > > >
> > > > > > > I certainly don't want all this work to be completed outside svn
> > > > HEAD
> > > > > > > and committed a week or two before the freeze for next
> > milestone.
> > > > > > >
> > > > > > > Regards,
> > > > > > >  Mark.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > >
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Rustem Rafikov <r....@gmail.com>.
Hi Tony,

Please look at another issue connected with the ICU commit.
https://issues.apache.org/jira/browse/HARMONY-5435
It has not been investigated deeply yet.

--Rustem


On Jan 28, 2008 6:49 AM, Tony Wu <wu...@gmail.com> wrote:

> They are all caused by different implementation of DateFormat, which
> I've recorded on wiki. I think they are acceptable unless it breaks
> some real product.  See following output, they have the same meaning
> but don't exactly equal on text.
>
> Result:   'text here 12:17:01 PM GMT+00:00 and here'
> Expected: 'text here 12:17:01 UTC PM and here'
>
> Result:   'text here 3:16:03 PM GMT+00:00 and here'
> Expected: 'text here 3:16:03 o'clock PM UTC and here'
>
> Result:   'text here 00-04-05 and here'
> Expected: 'text here 05/04/00 and here'
>
> I'll open another thread to ask people's opinions on this difference.
>
> On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> > Tony,
> >
> > It looks like this commit caused a regression. Could you take a look at
> > HARMONY-5430, please?
> >
> > On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> >
> > > Thanks, Aleksey and all
> > >
> > > patch committed at r612718. Then I'm going to deal with the non-bug
> > > difference. Hopefully many legcy bugs/differences can be fixed this
> > > time.
> > >
> > > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > > Hi, guys!
> > > >
> > > > Sure, correctness is important, but performance is important too. I
> > > > had profiled both versions (clean and patched) and see no
> significant
> > > > difference there: there are more GC happens what I believe connected
> > > > to internal ICU object creation and such. So, there are no obvious
> way
> > > > to maintain the same performance level, and may be we will try again
> > > > to eliminate ICU usage from the hotpath of frequently used workloads
> -
> > > > but previous investigation shows it's not that simple.
> > > >
> > > > Tony, please go ahead with committing this patch, we will deal with
> > > > ICU performance issues a little later.
> > > >
> > > > Thanks,
> > > > Aleksey.
> > > >
> > > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > Hi Aleksey,
> > > > >
> > > > > Thanks for you help.
> > > > >
> > > > > I've proved there is no performance degradation on my machine
> > > > > mentioned in my previous mail. I suppose the different result on
> your
> > > > > machine is caused by different options to SpecJBB. Anyway, my POV
> is
> > > > > that we should be willing to pay for the adoption of ICU if we
> have
> > > > > to. All of us should be positive on this point. I'd like to
> clarify
> > > > > the factors I'm facing below.
> > > > >
> > > > > Firstly, the original implementation of harmony is faster but
> > > > > incorrect. It is not reasonable to keep the code as is and refuse
> to
> > > > > correct it just because the bad version has better performance.
> IMHO
> > > > > performance is nothing if there is no correctness.
> > > > >
> > > > > Secondly, we adopt ICU through delegation which involves extra
> method
> > > > > calls than we implement it by ourselves, it does harm to
> performance
> > > > > and can not be worked around. But please do not forget we benefit
> from
> > > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > > >
> > > > > Lastly, branching does not make sense to me. My fix is very
> separate
> > > > > in luni and text, I can not guarantee that there is no
> modification
> > > > > during my work, so the synchronization between HEAD and my branch
> is
> > > > > required. Actually only I myself am working on the development
> work
> > > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > > might be a nightmare to me. Furthermore, the code on branch will
> not
> > > > > be automatically tested by continuous integration system and BTI,
> I do
> > > > > not want to work without collaboration, that's not an open source
> > > > > style, right? Will you ask a contributor to create a branch and
> play
> > > > > with himself whenever he wants to contribute?
> > > > >
> > > > >
> > > > >
> > > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > > >
> > > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com>
> > > wrote:
> > > > > > > Aleksey Shipilev wrote:
> > > > > > > > Yep, Tim, you're right. I believe that new implementation
> fixes
> > > a
> > > > > > > > number of bugs and will try to get it not degrading. I just
> want
> > > to
> > > > > > > > maintain the performance level of current trunk on the same
> > > level,
> > > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > > performance of HEAD revision for non-critical bugfix. That
> is, I
> > > want
> > > > > > > > to see HEAD changes like this:
> > > > > > > >
> > > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > > >
> > > > > > > > rather than
> > > > > > > >
> > > > > > > > "high performance, minor bug -> low performance, no bugs ->
> high
> > > > > > > > performance, no bugs"
> > > > > > > >
> > > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > > performance measurements at any time.
> > > > > >
> > > > > > Couldn't someone also get the HEAD Harmony revision and suffer
> from
> > > the
> > > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > > >
> > > > > > > > What do you think?
> > > > > >
> > > > > > > For sure, improving performance and fixing the bugs is the
> most
> > > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > > regression on HEAD between releases provided it is an area
> being
> > > > > > > actively worked upon.
> > > > > >
> > > > > > +1 especially if it fixes bugs
> > > > > >
> > > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > > >
> > > > > > Me too.
> > > > > >
> > > > > > > If you and Tony are happy to work on the patch to get it
> perfect
> > > then
> > > > > > > go ahead.  I hope it is not too troublesome to keep it in
> synch.
> > >  You
> > > > > > > could also consider a branch in SVN.
> > > > > >
> > > > > > This bothers me too.  Firstly, while it is being developed in
> patch
> > > > > > on JIRA, it is likely only Tony and Aleksey will really look at
> it.
> > > > > > Secondly, that progress will be slow because of the cost of
> keeping
> > > in
> > > > > > sync - this applies to an SVN branch too.
> > > > > >
> > > > > > I can't help thinking we'll make more progress if we apply the
> patch
> > > to
> > > > > > HEAD now.  We'll get wider visibility of problems with the new
> code
> > > -
> > > > > > and there are likely issues beyond the performance problems that
> > > have
> > > > > > been the focus so far - and more people will see the benefit of
> Tony
> > > > > > (and Aleksey's) hard work in getting us to this point.
> > > > > >
> > > > > > I certainly don't want all this work to be completed outside svn
> > > HEAD
> > > > > > and committed a week or two before the freeze for next
> milestone.
> > > > > >
> > > > > > Regards,
> > > > > >  Mark.
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > >
> > > > > Tony Wu
> > > > > China Software Development Lab, IBM
> > > > >
> > > >
> > >
> > >
> > > --
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
They are all caused by different implementation of DateFormat, which
I've recorded on wiki. I think they are acceptable unless it breaks
some real product.  See following output, they have the same meaning
but don't exactly equal on text.

Result:   'text here 12:17:01 PM GMT+00:00 and here'
Expected: 'text here 12:17:01 UTC PM and here'

Result:   'text here 3:16:03 PM GMT+00:00 and here'
Expected: 'text here 3:16:03 o'clock PM UTC and here'

Result:   'text here 00-04-05 and here'
Expected: 'text here 05/04/00 and here'

I'll open another thread to ask people's opinions on this difference.

On 1/25/08, Andrey Pavlenko <an...@googlemail.com> wrote:
> Tony,
>
> It looks like this commit caused a regression. Could you take a look at
> HARMONY-5430, please?
>
> On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
>
> > Thanks, Aleksey and all
> >
> > patch committed at r612718. Then I'm going to deal with the non-bug
> > difference. Hopefully many legcy bugs/differences can be fixed this
> > time.
> >
> > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > Hi, guys!
> > >
> > > Sure, correctness is important, but performance is important too. I
> > > had profiled both versions (clean and patched) and see no significant
> > > difference there: there are more GC happens what I believe connected
> > > to internal ICU object creation and such. So, there are no obvious way
> > > to maintain the same performance level, and may be we will try again
> > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > but previous investigation shows it's not that simple.
> > >
> > > Tony, please go ahead with committing this patch, we will deal with
> > > ICU performance issues a little later.
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > Hi Aleksey,
> > > >
> > > > Thanks for you help.
> > > >
> > > > I've proved there is no performance degradation on my machine
> > > > mentioned in my previous mail. I suppose the different result on your
> > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > that we should be willing to pay for the adoption of ICU if we have
> > > > to. All of us should be positive on this point. I'd like to clarify
> > > > the factors I'm facing below.
> > > >
> > > > Firstly, the original implementation of harmony is faster but
> > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > correct it just because the bad version has better performance. IMHO
> > > > performance is nothing if there is no correctness.
> > > >
> > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > calls than we implement it by ourselves, it does harm to performance
> > > > and can not be worked around. But please do not forget we benefit from
> > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > >
> > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > in luni and text, I can not guarantee that there is no modification
> > > > during my work, so the synchronization between HEAD and my branch is
> > > > required. Actually only I myself am working on the development work
> > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > be automatically tested by continuous integration system and BTI, I do
> > > > not want to work without collaboration, that's not an open source
> > > > style, right? Will you ask a contributor to create a branch and play
> > > > with himself whenever he wants to contribute?
> > > >
> > > >
> > > >
> > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > >
> > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com>
> > wrote:
> > > > > > Aleksey Shipilev wrote:
> > > > > > > Yep, Tim, you're right. I believe that new implementation fixes
> > a
> > > > > > > number of bugs and will try to get it not degrading. I just want
> > to
> > > > > > > maintain the performance level of current trunk on the same
> > level,
> > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > performance of HEAD revision for non-critical bugfix. That is, I
> > want
> > > > > > > to see HEAD changes like this:
> > > > > > >
> > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > >
> > > > > > > rather than
> > > > > > >
> > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > performance, no bugs"
> > > > > > >
> > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > performance measurements at any time.
> > > > >
> > > > > Couldn't someone also get the HEAD Harmony revision and suffer from
> > the
> > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > >
> > > > > > > What do you think?
> > > > >
> > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > regression on HEAD between releases provided it is an area being
> > > > > > actively worked upon.
> > > > >
> > > > > +1 especially if it fixes bugs
> > > > >
> > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > >
> > > > > Me too.
> > > > >
> > > > > > If you and Tony are happy to work on the patch to get it perfect
> > then
> > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.
> >  You
> > > > > > could also consider a branch in SVN.
> > > > >
> > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > Secondly, that progress will be slow because of the cost of keeping
> > in
> > > > > sync - this applies to an SVN branch too.
> > > > >
> > > > > I can't help thinking we'll make more progress if we apply the patch
> > to
> > > > > HEAD now.  We'll get wider visibility of problems with the new code
> > -
> > > > > and there are likely issues beyond the performance problems that
> > have
> > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > (and Aleksey's) hard work in getting us to this point.
> > > > >
> > > > > I certainly don't want all this work to be completed outside svn
> > HEAD
> > > > > and committed a week or two before the freeze for next milestone.
> > > > >
> > > > > Regards,
> > > > >  Mark.
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > >
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Andrey Pavlenko <an...@googlemail.com>.
Tony,

It looks like this commit caused a regression. Could you take a look at
HARMONY-5430, please?

On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:

> Thanks, Aleksey and all
>
> patch committed at r612718. Then I'm going to deal with the non-bug
> difference. Hopefully many legcy bugs/differences can be fixed this
> time.
>
> On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > Hi, guys!
> >
> > Sure, correctness is important, but performance is important too. I
> > had profiled both versions (clean and patched) and see no significant
> > difference there: there are more GC happens what I believe connected
> > to internal ICU object creation and such. So, there are no obvious way
> > to maintain the same performance level, and may be we will try again
> > to eliminate ICU usage from the hotpath of frequently used workloads -
> > but previous investigation shows it's not that simple.
> >
> > Tony, please go ahead with committing this patch, we will deal with
> > ICU performance issues a little later.
> >
> > Thanks,
> > Aleksey.
> >
> > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > Hi Aleksey,
> > >
> > > Thanks for you help.
> > >
> > > I've proved there is no performance degradation on my machine
> > > mentioned in my previous mail. I suppose the different result on your
> > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > that we should be willing to pay for the adoption of ICU if we have
> > > to. All of us should be positive on this point. I'd like to clarify
> > > the factors I'm facing below.
> > >
> > > Firstly, the original implementation of harmony is faster but
> > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > correct it just because the bad version has better performance. IMHO
> > > performance is nothing if there is no correctness.
> > >
> > > Secondly, we adopt ICU through delegation which involves extra method
> > > calls than we implement it by ourselves, it does harm to performance
> > > and can not be worked around. But please do not forget we benefit from
> > > ICU in bug fixing, maintenance, smaller code size and so on.
> > >
> > > Lastly, branching does not make sense to me. My fix is very separate
> > > in luni and text, I can not guarantee that there is no modification
> > > during my work, so the synchronization between HEAD and my branch is
> > > required. Actually only I myself am working on the development work
> > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > might be a nightmare to me. Furthermore, the code on branch will not
> > > be automatically tested by continuous integration system and BTI, I do
> > > not want to work without collaboration, that's not an open source
> > > style, right? Will you ask a contributor to create a branch and play
> > > with himself whenever he wants to contribute?
> > >
> > >
> > >
> > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > >
> > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com>
> wrote:
> > > > > Aleksey Shipilev wrote:
> > > > > > Yep, Tim, you're right. I believe that new implementation fixes
> a
> > > > > > number of bugs and will try to get it not degrading. I just want
> to
> > > > > > maintain the performance level of current trunk on the same
> level,
> > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > performance of HEAD revision for non-critical bugfix. That is, I
> want
> > > > > > to see HEAD changes like this:
> > > > > >
> > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > >
> > > > > > rather than
> > > > > >
> > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > performance, no bugs"
> > > > > >
> > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > performance measurements at any time.
> > > >
> > > > Couldn't someone also get the HEAD Harmony revision and suffer from
> the
> > > > known/fixable-with-Tony's-patch bugs at any time?
> > > >
> > > > > > What do you think?
> > > >
> > > > > For sure, improving performance and fixing the bugs is the most
> > > > > desirable state.  I actually don't mind some minor performance
> > > > > regression on HEAD between releases provided it is an area being
> > > > > actively worked upon.
> > > >
> > > > +1 especially if it fixes bugs
> > > >
> > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > >
> > > > Me too.
> > > >
> > > > > If you and Tony are happy to work on the patch to get it perfect
> then
> > > > > go ahead.  I hope it is not too troublesome to keep it in synch.
>  You
> > > > > could also consider a branch in SVN.
> > > >
> > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > Secondly, that progress will be slow because of the cost of keeping
> in
> > > > sync - this applies to an SVN branch too.
> > > >
> > > > I can't help thinking we'll make more progress if we apply the patch
> to
> > > > HEAD now.  We'll get wider visibility of problems with the new code
> -
> > > > and there are likely issues beyond the performance problems that
> have
> > > > been the focus so far - and more people will see the benefit of Tony
> > > > (and Aleksey's) hard work in getting us to this point.
> > > >
> > > > I certainly don't want all this work to be completed outside svn
> HEAD
> > > > and committed a week or two before the freeze for next milestone.
> > > >
> > > > Regards,
> > > >  Mark.
> > > >
> > > >
> > > >
> > >
> > >
> > > --
> > >
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Thanks very much, so we got 4MB src code cutted w/o pay. Cheer!

On 1/23/08, Aleksey Shipilev <al...@gmail.com> wrote:
> Hi Tony,
>
> The good thing is, after the commit I have compared revisions 610727
> (before your commit) and 612866 (after your commit) and haven't
> noticed degradation while running SPECjbb2005 in 2-VM configuration.
> That's definitely good news :)
>
> Thanks,
> Aleksey.
>
> On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> > Thanks, Aleksey and all
> >
> > patch committed at r612718. Then I'm going to deal with the non-bug
> > difference. Hopefully many legcy bugs/differences can be fixed this
> > time.
> >
> >
> > On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > Hi, guys!
> > >
> > > Sure, correctness is important, but performance is important too. I
> > > had profiled both versions (clean and patched) and see no significant
> > > difference there: there are more GC happens what I believe connected
> > > to internal ICU object creation and such. So, there are no obvious way
> > > to maintain the same performance level, and may be we will try again
> > > to eliminate ICU usage from the hotpath of frequently used workloads -
> > > but previous investigation shows it's not that simple.
> > >
> > > Tony, please go ahead with committing this patch, we will deal with
> > > ICU performance issues a little later.
> > >
> > > Thanks,
> > > Aleksey.
> > >
> > > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > Hi Aleksey,
> > > >
> > > > Thanks for you help.
> > > >
> > > > I've proved there is no performance degradation on my machine
> > > > mentioned in my previous mail. I suppose the different result on your
> > > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > > that we should be willing to pay for the adoption of ICU if we have
> > > > to. All of us should be positive on this point. I'd like to clarify
> > > > the factors I'm facing below.
> > > >
> > > > Firstly, the original implementation of harmony is faster but
> > > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > > correct it just because the bad version has better performance. IMHO
> > > > performance is nothing if there is no correctness.
> > > >
> > > > Secondly, we adopt ICU through delegation which involves extra method
> > > > calls than we implement it by ourselves, it does harm to performance
> > > > and can not be worked around. But please do not forget we benefit from
> > > > ICU in bug fixing, maintenance, smaller code size and so on.
> > > >
> > > > Lastly, branching does not make sense to me. My fix is very separate
> > > > in luni and text, I can not guarantee that there is no modification
> > > > during my work, so the synchronization between HEAD and my branch is
> > > > required. Actually only I myself am working on the development work
> > > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > > might be a nightmare to me. Furthermore, the code on branch will not
> > > > be automatically tested by continuous integration system and BTI, I do
> > > > not want to work without collaboration, that's not an open source
> > > > style, right? Will you ask a contributor to create a branch and play
> > > > with himself whenever he wants to contribute?
> > > >
> > > >
> > > >
> > > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > > >
> > > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > > Aleksey Shipilev wrote:
> > > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > > maintain the performance level of current trunk on the same level,
> > > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > > to see HEAD changes like this:
> > > > > > >
> > > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > > >
> > > > > > > rather than
> > > > > > >
> > > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > > performance, no bugs"
> > > > > > >
> > > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > > performance measurements at any time.
> > > > >
> > > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > > known/fixable-with-Tony's-patch bugs at any time?
> > > > >
> > > > > > > What do you think?
> > > > >
> > > > > > For sure, improving performance and fixing the bugs is the most
> > > > > > desirable state.  I actually don't mind some minor performance
> > > > > > regression on HEAD between releases provided it is an area being
> > > > > > actively worked upon.
> > > > >
> > > > > +1 especially if it fixes bugs
> > > > >
> > > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > > >
> > > > > Me too.
> > > > >
> > > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > > could also consider a branch in SVN.
> > > > >
> > > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > > sync - this applies to an SVN branch too.
> > > > >
> > > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > > and there are likely issues beyond the performance problems that have
> > > > > been the focus so far - and more people will see the benefit of Tony
> > > > > (and Aleksey's) hard work in getting us to this point.
> > > > >
> > > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > > and committed a week or two before the freeze for next milestone.
> > > > >
> > > > > Regards,
> > > > >  Mark.
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > >
> > > > Tony Wu
> > > > China Software Development Lab, IBM
> > > >
> > >
> >
> >
> > --
> >
> > Tony Wu
> > China Software Development Lab, IBM
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Hi Tony,

The good thing is, after the commit I have compared revisions 610727
(before your commit) and 612866 (after your commit) and haven't
noticed degradation while running SPECjbb2005 in 2-VM configuration.
That's definitely good news :)

Thanks,
Aleksey.

On Jan 21, 2008 10:17 AM, Tony Wu <wu...@gmail.com> wrote:
> Thanks, Aleksey and all
>
> patch committed at r612718. Then I'm going to deal with the non-bug
> difference. Hopefully many legcy bugs/differences can be fixed this
> time.
>
>
> On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > Hi, guys!
> >
> > Sure, correctness is important, but performance is important too. I
> > had profiled both versions (clean and patched) and see no significant
> > difference there: there are more GC happens what I believe connected
> > to internal ICU object creation and such. So, there are no obvious way
> > to maintain the same performance level, and may be we will try again
> > to eliminate ICU usage from the hotpath of frequently used workloads -
> > but previous investigation shows it's not that simple.
> >
> > Tony, please go ahead with committing this patch, we will deal with
> > ICU performance issues a little later.
> >
> > Thanks,
> > Aleksey.
> >
> > On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > > Hi Aleksey,
> > >
> > > Thanks for you help.
> > >
> > > I've proved there is no performance degradation on my machine
> > > mentioned in my previous mail. I suppose the different result on your
> > > machine is caused by different options to SpecJBB. Anyway, my POV is
> > > that we should be willing to pay for the adoption of ICU if we have
> > > to. All of us should be positive on this point. I'd like to clarify
> > > the factors I'm facing below.
> > >
> > > Firstly, the original implementation of harmony is faster but
> > > incorrect. It is not reasonable to keep the code as is and refuse to
> > > correct it just because the bad version has better performance. IMHO
> > > performance is nothing if there is no correctness.
> > >
> > > Secondly, we adopt ICU through delegation which involves extra method
> > > calls than we implement it by ourselves, it does harm to performance
> > > and can not be worked around. But please do not forget we benefit from
> > > ICU in bug fixing, maintenance, smaller code size and so on.
> > >
> > > Lastly, branching does not make sense to me. My fix is very separate
> > > in luni and text, I can not guarantee that there is no modification
> > > during my work, so the synchronization between HEAD and my branch is
> > > required. Actually only I myself am working on the development work
> > > (Surely, Aleksey is very helpful on testing), this synchronization
> > > might be a nightmare to me. Furthermore, the code on branch will not
> > > be automatically tested by continuous integration system and BTI, I do
> > > not want to work without collaboration, that's not an open source
> > > style, right? Will you ask a contributor to create a branch and play
> > > with himself whenever he wants to contribute?
> > >
> > >
> > >
> > > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > > >
> > > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > > Aleksey Shipilev wrote:
> > > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > > maintain the performance level of current trunk on the same level,
> > > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > > to see HEAD changes like this:
> > > > > >
> > > > > > "high performance, minor bug -> high performance, no bugs"
> > > > > >
> > > > > > rather than
> > > > > >
> > > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > > performance, no bugs"
> > > > > >
> > > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > > performance measurements at any time.
> > > >
> > > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > > known/fixable-with-Tony's-patch bugs at any time?
> > > >
> > > > > > What do you think?
> > > >
> > > > > For sure, improving performance and fixing the bugs is the most
> > > > > desirable state.  I actually don't mind some minor performance
> > > > > regression on HEAD between releases provided it is an area being
> > > > > actively worked upon.
> > > >
> > > > +1 especially if it fixes bugs
> > > >
> > > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > > >
> > > > Me too.
> > > >
> > > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > > could also consider a branch in SVN.
> > > >
> > > > This bothers me too.  Firstly, while it is being developed in patch
> > > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > > Secondly, that progress will be slow because of the cost of keeping in
> > > > sync - this applies to an SVN branch too.
> > > >
> > > > I can't help thinking we'll make more progress if we apply the patch to
> > > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > > and there are likely issues beyond the performance problems that have
> > > > been the focus so far - and more people will see the benefit of Tony
> > > > (and Aleksey's) hard work in getting us to this point.
> > > >
> > > > I certainly don't want all this work to be completed outside svn HEAD
> > > > and committed a week or two before the freeze for next milestone.
> > > >
> > > > Regards,
> > > >  Mark.
> > > >
> > > >
> > > >
> > >
> > >
> > > --
> > >
> > > Tony Wu
> > > China Software Development Lab, IBM
> > >
> >
>
>
> --
>
> Tony Wu
> China Software Development Lab, IBM
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Thanks, Aleksey and all

patch committed at r612718. Then I'm going to deal with the non-bug
difference. Hopefully many legcy bugs/differences can be fixed this
time.

On 1/17/08, Aleksey Shipilev <al...@gmail.com> wrote:
> Hi, guys!
>
> Sure, correctness is important, but performance is important too. I
> had profiled both versions (clean and patched) and see no significant
> difference there: there are more GC happens what I believe connected
> to internal ICU object creation and such. So, there are no obvious way
> to maintain the same performance level, and may be we will try again
> to eliminate ICU usage from the hotpath of frequently used workloads -
> but previous investigation shows it's not that simple.
>
> Tony, please go ahead with committing this patch, we will deal with
> ICU performance issues a little later.
>
> Thanks,
> Aleksey.
>
> On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> > Hi Aleksey,
> >
> > Thanks for you help.
> >
> > I've proved there is no performance degradation on my machine
> > mentioned in my previous mail. I suppose the different result on your
> > machine is caused by different options to SpecJBB. Anyway, my POV is
> > that we should be willing to pay for the adoption of ICU if we have
> > to. All of us should be positive on this point. I'd like to clarify
> > the factors I'm facing below.
> >
> > Firstly, the original implementation of harmony is faster but
> > incorrect. It is not reasonable to keep the code as is and refuse to
> > correct it just because the bad version has better performance. IMHO
> > performance is nothing if there is no correctness.
> >
> > Secondly, we adopt ICU through delegation which involves extra method
> > calls than we implement it by ourselves, it does harm to performance
> > and can not be worked around. But please do not forget we benefit from
> > ICU in bug fixing, maintenance, smaller code size and so on.
> >
> > Lastly, branching does not make sense to me. My fix is very separate
> > in luni and text, I can not guarantee that there is no modification
> > during my work, so the synchronization between HEAD and my branch is
> > required. Actually only I myself am working on the development work
> > (Surely, Aleksey is very helpful on testing), this synchronization
> > might be a nightmare to me. Furthermore, the code on branch will not
> > be automatically tested by continuous integration system and BTI, I do
> > not want to work without collaboration, that's not an open source
> > style, right? Will you ask a contributor to create a branch and play
> > with himself whenever he wants to contribute?
> >
> >
> >
> > On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> > >
> > > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > > Aleksey Shipilev wrote:
> > > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > > number of bugs and will try to get it not degrading. I just want to
> > > > > maintain the performance level of current trunk on the same level,
> > > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > > to see HEAD changes like this:
> > > > >
> > > > > "high performance, minor bug -> high performance, no bugs"
> > > > >
> > > > > rather than
> > > > >
> > > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > > performance, no bugs"
> > > > >
> > > > > ...because anyone could get the HEAD Harmony revision for
> > > > > performance measurements at any time.
> > >
> > > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > > known/fixable-with-Tony's-patch bugs at any time?
> > >
> > > > > What do you think?
> > >
> > > > For sure, improving performance and fixing the bugs is the most
> > > > desirable state.  I actually don't mind some minor performance
> > > > regression on HEAD between releases provided it is an area being
> > > > actively worked upon.
> > >
> > > +1 especially if it fixes bugs
> > >
> > > > I'd also like to get to 4.2Mb source code reduction too ;-)
> > >
> > > Me too.
> > >
> > > > If you and Tony are happy to work on the patch to get it perfect then
> > > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > > could also consider a branch in SVN.
> > >
> > > This bothers me too.  Firstly, while it is being developed in patch
> > > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > > Secondly, that progress will be slow because of the cost of keeping in
> > > sync - this applies to an SVN branch too.
> > >
> > > I can't help thinking we'll make more progress if we apply the patch to
> > > HEAD now.  We'll get wider visibility of problems with the new code -
> > > and there are likely issues beyond the performance problems that have
> > > been the focus so far - and more people will see the benefit of Tony
> > > (and Aleksey's) hard work in getting us to this point.
> > >
> > > I certainly don't want all this work to be completed outside svn HEAD
> > > and committed a week or two before the freeze for next milestone.
> > >
> > > Regards,
> > >  Mark.
> > >
> > >
> > >
> >
> >
> > --
> >
> > Tony Wu
> > China Software Development Lab, IBM
> >
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Hi, guys!

Sure, correctness is important, but performance is important too. I
had profiled both versions (clean and patched) and see no significant
difference there: there are more GC happens what I believe connected
to internal ICU object creation and such. So, there are no obvious way
to maintain the same performance level, and may be we will try again
to eliminate ICU usage from the hotpath of frequently used workloads -
but previous investigation shows it's not that simple.

Tony, please go ahead with committing this patch, we will deal with
ICU performance issues a little later.

Thanks,
Aleksey.

On Jan 15, 2008 1:52 PM, Tony Wu <wu...@gmail.com> wrote:
> Hi Aleksey,
>
> Thanks for you help.
>
> I've proved there is no performance degradation on my machine
> mentioned in my previous mail. I suppose the different result on your
> machine is caused by different options to SpecJBB. Anyway, my POV is
> that we should be willing to pay for the adoption of ICU if we have
> to. All of us should be positive on this point. I'd like to clarify
> the factors I'm facing below.
>
> Firstly, the original implementation of harmony is faster but
> incorrect. It is not reasonable to keep the code as is and refuse to
> correct it just because the bad version has better performance. IMHO
> performance is nothing if there is no correctness.
>
> Secondly, we adopt ICU through delegation which involves extra method
> calls than we implement it by ourselves, it does harm to performance
> and can not be worked around. But please do not forget we benefit from
> ICU in bug fixing, maintenance, smaller code size and so on.
>
> Lastly, branching does not make sense to me. My fix is very separate
> in luni and text, I can not guarantee that there is no modification
> during my work, so the synchronization between HEAD and my branch is
> required. Actually only I myself am working on the development work
> (Surely, Aleksey is very helpful on testing), this synchronization
> might be a nightmare to me. Furthermore, the code on branch will not
> be automatically tested by continuous integration system and BTI, I do
> not want to work without collaboration, that's not an open source
> style, right? Will you ask a contributor to create a branch and play
> with himself whenever he wants to contribute?
>
>
>
> On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
> >
> > On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > > Aleksey Shipilev wrote:
> > > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > > number of bugs and will try to get it not degrading. I just want to
> > > > maintain the performance level of current trunk on the same level,
> > > > gradually fixing functional bugs. I don't like to sacrifice
> > > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > > to see HEAD changes like this:
> > > >
> > > > "high performance, minor bug -> high performance, no bugs"
> > > >
> > > > rather than
> > > >
> > > > "high performance, minor bug -> low performance, no bugs -> high
> > > > performance, no bugs"
> > > >
> > > > ...because anyone could get the HEAD Harmony revision for
> > > > performance measurements at any time.
> >
> > Couldn't someone also get the HEAD Harmony revision and suffer from the
> > known/fixable-with-Tony's-patch bugs at any time?
> >
> > > > What do you think?
> >
> > > For sure, improving performance and fixing the bugs is the most
> > > desirable state.  I actually don't mind some minor performance
> > > regression on HEAD between releases provided it is an area being
> > > actively worked upon.
> >
> > +1 especially if it fixes bugs
> >
> > > I'd also like to get to 4.2Mb source code reduction too ;-)
> >
> > Me too.
> >
> > > If you and Tony are happy to work on the patch to get it perfect then
> > > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > > could also consider a branch in SVN.
> >
> > This bothers me too.  Firstly, while it is being developed in patch
> > on JIRA, it is likely only Tony and Aleksey will really look at it.
> > Secondly, that progress will be slow because of the cost of keeping in
> > sync - this applies to an SVN branch too.
> >
> > I can't help thinking we'll make more progress if we apply the patch to
> > HEAD now.  We'll get wider visibility of problems with the new code -
> > and there are likely issues beyond the performance problems that have
> > been the focus so far - and more people will see the benefit of Tony
> > (and Aleksey's) hard work in getting us to this point.
> >
> > I certainly don't want all this work to be completed outside svn HEAD
> > and committed a week or two before the freeze for next milestone.
> >
> > Regards,
> >  Mark.
> >
> >
> >
>
>
> --
>
> Tony Wu
> China Software Development Lab, IBM
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Hi Aleksey,

Thanks for you help.

I've proved there is no performance degradation on my machine
mentioned in my previous mail. I suppose the different result on your
machine is caused by different options to SpecJBB. Anyway, my POV is
that we should be willing to pay for the adoption of ICU if we have
to. All of us should be positive on this point. I'd like to clarify
the factors I'm facing below.

Firstly, the original implementation of harmony is faster but
incorrect. It is not reasonable to keep the code as is and refuse to
correct it just because the bad version has better performance. IMHO
performance is nothing if there is no correctness.

Secondly, we adopt ICU through delegation which involves extra method
calls than we implement it by ourselves, it does harm to performance
and can not be worked around. But please do not forget we benefit from
ICU in bug fixing, maintenance, smaller code size and so on.

Lastly, branching does not make sense to me. My fix is very separate
in luni and text, I can not guarantee that there is no modification
during my work, so the synchronization between HEAD and my branch is
required. Actually only I myself am working on the development work
(Surely, Aleksey is very helpful on testing), this synchronization
might be a nightmare to me. Furthermore, the code on branch will not
be automatically tested by continuous integration system and BTI, I do
not want to work without collaboration, that's not an open source
style, right? Will you ask a contributor to create a branch and play
with himself whenever he wants to contribute?


On 1/15/08, Mark Hindess <ma...@googlemail.com> wrote:
>
> On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> > Aleksey Shipilev wrote:
> > > Yep, Tim, you're right. I believe that new implementation fixes a
> > > number of bugs and will try to get it not degrading. I just want to
> > > maintain the performance level of current trunk on the same level,
> > > gradually fixing functional bugs. I don't like to sacrifice
> > > performance of HEAD revision for non-critical bugfix. That is, I want
> > > to see HEAD changes like this:
> > >
> > > "high performance, minor bug -> high performance, no bugs"
> > >
> > > rather than
> > >
> > > "high performance, minor bug -> low performance, no bugs -> high
> > > performance, no bugs"
> > >
> > > ...because anyone could get the HEAD Harmony revision for
> > > performance measurements at any time.
>
> Couldn't someone also get the HEAD Harmony revision and suffer from the
> known/fixable-with-Tony's-patch bugs at any time?
>
> > > What do you think?
>
> > For sure, improving performance and fixing the bugs is the most
> > desirable state.  I actually don't mind some minor performance
> > regression on HEAD between releases provided it is an area being
> > actively worked upon.
>
> +1 especially if it fixes bugs
>
> > I'd also like to get to 4.2Mb source code reduction too ;-)
>
> Me too.
>
> > If you and Tony are happy to work on the patch to get it perfect then
> > go ahead.  I hope it is not too troublesome to keep it in synch.  You
> > could also consider a branch in SVN.
>
> This bothers me too.  Firstly, while it is being developed in patch
> on JIRA, it is likely only Tony and Aleksey will really look at it.
> Secondly, that progress will be slow because of the cost of keeping in
> sync - this applies to an SVN branch too.
>
> I can't help thinking we'll make more progress if we apply the patch to
> HEAD now.  We'll get wider visibility of problems with the new code -
> and there are likely issues beyond the performance problems that have
> been the focus so far - and more people will see the benefit of Tony
> (and Aleksey's) hard work in getting us to this point.
>
> I certainly don't want all this work to be completed outside svn HEAD
> and committed a week or two before the freeze for next milestone.
>
> Regards,
>  Mark.
>
>
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Mark Hindess <ma...@googlemail.com>.
On 14 January 2008 at 16:21, Tim Ellison <t....@gmail.com> wrote:
> Aleksey Shipilev wrote:
> > Yep, Tim, you're right. I believe that new implementation fixes a
> > number of bugs and will try to get it not degrading. I just want to
> > maintain the performance level of current trunk on the same level,
> > gradually fixing functional bugs. I don't like to sacrifice
> > performance of HEAD revision for non-critical bugfix. That is, I want
> > to see HEAD changes like this:
> > 
> > "high performance, minor bug -> high performance, no bugs"
> > 
> > rather than
> > 
> > "high performance, minor bug -> low performance, no bugs -> high
> > performance, no bugs"
> > 
> > ...because anyone could get the HEAD Harmony revision for
> > performance measurements at any time.

Couldn't someone also get the HEAD Harmony revision and suffer from the
known/fixable-with-Tony's-patch bugs at any time?

> > What do you think?

> For sure, improving performance and fixing the bugs is the most
> desirable state.  I actually don't mind some minor performance
> regression on HEAD between releases provided it is an area being
> actively worked upon.

+1 especially if it fixes bugs

> I'd also like to get to 4.2Mb source code reduction too ;-)

Me too.

> If you and Tony are happy to work on the patch to get it perfect then
> go ahead.  I hope it is not too troublesome to keep it in synch.  You
> could also consider a branch in SVN.

This bothers me too.  Firstly, while it is being developed in patch
on JIRA, it is likely only Tony and Aleksey will really look at it.
Secondly, that progress will be slow because of the cost of keeping in
sync - this applies to an SVN branch too.

I can't help thinking we'll make more progress if we apply the patch to
HEAD now.  We'll get wider visibility of problems with the new code -
and there are likely issues beyond the performance problems that have
been the focus so far - and more people will see the benefit of Tony
(and Aleksey's) hard work in getting us to this point.

I certainly don't want all this work to be completed outside svn HEAD
and committed a week or two before the freeze for next milestone.

Regards,
 Mark.



Re: [classlib]remove the duplicate locale data

Posted by Tim Ellison <t....@gmail.com>.
Aleksey Shipilev wrote:
> Yep, Tim, you're right. I believe that new implementation fixes a
> number of bugs and will try to get it not degrading. I just want to
> maintain the performance level of current trunk on the same level,
> gradually fixing functional bugs. I don't like to sacrifice
> performance of HEAD revision for non-critical bugfix. That is, I want
> to see HEAD changes like this:
> 
> "high performance, minor bug -> high performance, no bugs"
> 
> rather than
> 
> "high performance, minor bug -> low performance, no bugs -> high
> performance, no bugs"
> 
> ...because anyone could get the HEAD Harmony revision for performance
> measurements at any time.
> 
> What do you think?

For sure, improving performance and fixing the bugs is the most 
desirable state.  I actually don't mind some minor performance 
regression on HEAD between releases provided it is an area being 
actively worked upon.  I'd also like to get to 4.2Mb source code 
reduction too ;-)

If you and Tony are happy to work on the patch to get it perfect then go 
ahead.  I hope it is not too troublesome to keep it in synch.  You could 
also consider a branch in SVN.

Regards,
Tim

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Yep, Tim, you're right. I believe that new implementation fixes a
number of bugs and will try to get it not degrading. I just want to
maintain the performance level of current trunk on the same level,
gradually fixing functional bugs. I don't like to sacrifice
performance of HEAD revision for non-critical bugfix. That is, I want
to see HEAD changes like this:

"high performance, minor bug -> high performance, no bugs"

rather than

"high performance, minor bug -> low performance, no bugs -> high
performance, no bugs"

...because anyone could get the HEAD Harmony revision for performance
measurements at any time.

What do you think?

Thanks,
Aleksey,
ESSD, Intel.


On Jan 14, 2008 6:43 PM, Tim Ellison <t....@gmail.com> wrote:
> While I agree that performance is important, it is not the only
> consideration.  If this patch fixes a number of bugs then it should be
> considered on that merit, and we should look into retrieving the
> performance numbers on a more capable implementation.

Re: [classlib]remove the duplicate locale data

Posted by Tim Ellison <t....@gmail.com>.
Aleksey Shipilev wrote:
> I still see 2% *degradation* on SPECjbb2005. That's better than
> previous 7% degradation, but please postpone this patch anyway. We
> might look into the issue.

While I agree that performance is important, it is not the only 
consideration.  If this patch fixes a number of bugs then it should be 
considered on that merit, and we should look into retrieving the 
performance numbers on a more capable implementation.

To put it another way, if we don't have to comply with the correctness 
principle, then we can make much greater progress on the performance 
goals :-)

Regards,
Tim


Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Tony,

I still see 2% *degradation* on SPECjbb2005. That's better than
previous 7% degradation, but please postpone this patch anyway. We
might look into the issue.

Thanks,
Aleksey,
ESSD, Intel.

On Jan 14, 2008 2:30 PM, Aleksey Shipilev <al...@gmail.com> wrote:
> Sure, Tony. I will check it ASAP.
>
>
> On Jan 14, 2008 2:22 PM, Tony Wu <wu...@gmail.com> wrote:
> > Oh yes, thank you Aleksay. I hadn't commit this patch yet.
> >
> > My thinkpad got something wrong these days, so I just get back to work
> > today. I double checked the patch on HARMONY-5061, seems it is
> > outdated because of the test structure change in luni. I've provided a
> > new one named "allinone0114.diff". would you please verify it? thanks
> > a lot!
> >
> >
> > On 1/10/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > > Hi, Tony!
> > >
> > > Sorry for late reply, I should notice the thread earlier. Have you
> > > committed the patch already? If not, please point latest patch in JIRA
> > > so I can confirm it has no degradations on DRLVM.
> > >
> > > Thanks,
> > > Aleksey,
> > > ESSD, Intel
> > >
> > > On Jan 4, 2008 5:23 AM, Tony Wu <wu...@gmail.com> wrote:
> > > > I assume this patch is ok finally if no more reponse :)
> > >
> >
> >
> > --
> >
> > Tony Wu
> > China Software Development Lab, IBM
> >
>

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Sure, Tony. I will check it ASAP.

On Jan 14, 2008 2:22 PM, Tony Wu <wu...@gmail.com> wrote:
> Oh yes, thank you Aleksay. I hadn't commit this patch yet.
>
> My thinkpad got something wrong these days, so I just get back to work
> today. I double checked the patch on HARMONY-5061, seems it is
> outdated because of the test structure change in luni. I've provided a
> new one named "allinone0114.diff". would you please verify it? thanks
> a lot!
>
>
> On 1/10/08, Aleksey Shipilev <al...@gmail.com> wrote:
> > Hi, Tony!
> >
> > Sorry for late reply, I should notice the thread earlier. Have you
> > committed the patch already? If not, please point latest patch in JIRA
> > so I can confirm it has no degradations on DRLVM.
> >
> > Thanks,
> > Aleksey,
> > ESSD, Intel
> >
> > On Jan 4, 2008 5:23 AM, Tony Wu <wu...@gmail.com> wrote:
> > > I assume this patch is ok finally if no more reponse :)
> >
>
>
> --
>
> Tony Wu
> China Software Development Lab, IBM
>

Re: [classlib]remove the duplicate locale data

Posted by Tony Wu <wu...@gmail.com>.
Oh yes, thank you Aleksay. I hadn't commit this patch yet.

My thinkpad got something wrong these days, so I just get back to work
today. I double checked the patch on HARMONY-5061, seems it is
outdated because of the test structure change in luni. I've provided a
new one named "allinone0114.diff". would you please verify it? thanks
a lot!

On 1/10/08, Aleksey Shipilev <al...@gmail.com> wrote:
> Hi, Tony!
>
> Sorry for late reply, I should notice the thread earlier. Have you
> committed the patch already? If not, please point latest patch in JIRA
> so I can confirm it has no degradations on DRLVM.
>
> Thanks,
> Aleksey,
> ESSD, Intel
>
> On Jan 4, 2008 5:23 AM, Tony Wu <wu...@gmail.com> wrote:
> > I assume this patch is ok finally if no more reponse :)
>


-- 
Tony Wu
China Software Development Lab, IBM

Re: [classlib]remove the duplicate locale data

Posted by Aleksey Shipilev <al...@gmail.com>.
Hi, Tony!

Sorry for late reply, I should notice the thread earlier. Have you
committed the patch already? If not, please point latest patch in JIRA
so I can confirm it has no degradations on DRLVM.

Thanks,
Aleksey,
ESSD, Intel

On Jan 4, 2008 5:23 AM, Tony Wu <wu...@gmail.com> wrote:
> I assume this patch is ok finally if no more reponse :)

Re: [classlib]remove the duplicate locale data

Posted by Sean Qiu <se...@gmail.com>.
Great news.

2008/1/4, Tony Wu <wu...@gmail.com>:
>
> Hi,
>
> I assume this patch is ok finally if no more reponse :)
>
> On 12/26/07, Tony Wu <wu...@gmail.com> wrote:
> > On 12/22/07, Alexey Petrenko <al...@gmail.com> wrote:
> > > Tony, can we fix all the issues with ICU in the branch before
> > > committing these changes to trunk?
> >
> > Thanks for your suggestion, I think I've got almost done on my
> > machine.  I'll consider about branching when I face a big task again
> > :)
> >
> > >
> > > SY, Alexey
> > >
> > > 2007/12/21, Aleksey Shipilev <al...@gmail.com>:
> > > > As I've concerned, SPECjbb2005 degradation is not gone: it is still
> 7%.
> > > > I think we shouldn't commit this patch in main trunk.
> > > >
> > > > Thanks,
> > > > Aleksey.
> > > >
> > > > On Dec 21, 2007 5:50 PM, Aleksey Shipilev <
> aleksey.shipilev@gmail.com> wrote:
> > > > > Tony,
> > > > >
> > > > > I think it's better to check new patch performance before
> committing.
> > > > > I'm going to check it in next several hours.
> > > > >
> > > > > Thanks,
> > > > > Aleksey,
> > > > > ESSD, Intel.
> > > > >
> > > > >
> > > > > On Dec 21, 2007 12:01 PM, Tony Wu <wu...@gmail.com> wrote:
> > > > > > Hi all,
> > > > > > I'd like to recommit the patch if no one objects.
> > > > > >
> > > > > > My new patch on 5061 fixes
> > > > > > 1. the performance issue addressed in 5129
> > > > > > 2. the return type gap on 5085
> > > > > > 3. the testcases in text module, now they can pass on both RI
> and Harmony
> > > > > >
> > > > > > Still this patch willl exclude following tests in luni
> > > > > > +tests/api/java/util/CurrencyTest.java
> > > > > > +tests/api/java/util/FormatterTest.java
> > > > > > +tests/api/java/util/GregorianCalendarTest.java
> > > > > > +tests/api/java/util/LocaleTest.java
> > > > > > +tests/api/java/util/ScannerTest.java
> > > > > >
> > > > > > In total there are 1 error and 13 failures, all of which are
> caused by
> > > > > > differences between RI and ICU. I'll record them as non-bug
> difference
> > > > > > later.
> > > > > >
> > > > > >
> > > > > > On 11/12/07, Vladimir Strigun <vs...@gmail.com> wrote:
> > > > > > > Thanks Mark,
> > > > > > >
> > > > > > > All dacapo benches passed after your commit.
> > > > > > >
> > > > > > > Thanks.
> > > > > > > Vladimir.
> > > > > > >
> > > > > > > On 11/12/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > I noticed Mark has committed your patch, Thank you and Mark
> :)
> > > > > > > >
> > > > > > > > On 11/10/07, Vladimir Strigun <vs...@gmail.com> wrote:
> > > > > > > > > On 11/9/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > Hi Vladimir
> > > > > > > > > > really busy this weekend. I've recorded it as JIRA 5100,
> high
> > > > > > > > > > appreciate if you could provide a simple testcase.
> > > > > > > > >
> > > > > > > > > Hi Tony,
> > > > > > > > >
> > > > > > > > > It seems we've faced with first isues in new ICU version.
> ICU returns
> > > > > > > > > null instead of correct string for time zone. Could you
> please check
> > > > > > > > > the patch attached to JIRA and apply it to fix the
> regression?
> > > > > > > > >
> > > > > > > > > Thanks.
> > > > > > > > > Vladimir.
> > > > > > > > >
> > > > > > > > > > On 11/7/07, Vladimir Strigun <vs...@gmail.com> wrote:
> > > > > > > > > > > Hi Tony,
> > > > > > > > > > >
> > > > > > > > > > > It seems your commit r592434 introduce regression for
> Dacapo.jython
> > > > > > > > > > > (the bench passed on r592433). Here is the execution
> log:
> > > > > > > > > > > ===== DaCapo jython starting =====
> > > > > > > > > > >
> -------------------------------------------------------------------------------
> > > > > > > > > > > PYBENCH 2.0
> > > > > > > > > > >
> -------------------------------------------------------------------------------
> > > > > > > > > > > * using Python 2.2a1
> > > > > > > > > > > * Python version doesn't support garbage collection
> > > > > > > > > > > * system check interval set to maximum: 2147483647
> > > > > > > > > > > * using timer: time.time
> > > > > > > > > > >
> > > > > > > > > > > Traceback (innermost last):
> > > > > > > > > > >  File ".\scratch\jython\pybench\pybench.py", line 946,
> in ?
> > > > > > > > > > >  File
> "Z:\UBS\Storage\.\scratch\jython\pybench\CommandLine.py", line
> > > > > > > > > > > 346, in __init__
> > > > > > > > > > >  File ".\scratch\jython\pybench\pybench.py", line 903,
> in main
> > > > > > > > > > >  File ".\scratch\jython\pybench\pybench.py", line 425,
> in __init__
> > > > > > > > > > > java.lang.IllegalArgumentException: Cannot create
> PyString from null!
> > > > > > > > > > >        at org.python.core.PyString.<init>(Unknown
> Source)
> > > > > > > > > > >        at org.python.modules.time.classDictInit(Unknown
> Source)
> > > > > > > > > > >        at java.lang.reflect.VMReflection.invokeMethod(
> VMReflection.java)
> > > > > > > > > > >        at java.lang.reflect.Method.invoke(Method.java
> :317)
> > > > > > > > > > >        at org.python.core.PyJavaClass.initialize(Unknown
> Source)
> > > > > > > > > > >        at
> org.python.core.PyJavaClass.lookupGivingClass(Unknown Source)
> > > > > > > > > > >        at org.python.core.PyClass.lookup(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyJavaClass.__findattr__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyObject.__getattr__(Unknown
> Source)
> > > > > > > > > > >        at org.python.pycode._pyx1.__init__$14
> (.\scratch\jython\pybench\pybench.py:425)
> > > > > > > > > > >        at org.python.pycode._pyx1.call_function
> (.\scratch\jython\pybench\pybench.py)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyFunction.__call__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyInstance.__init__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyClass.__call__(Unknown
> Source)
> > > > > > > > > > >        at org.python.pycode._pyx1.main$25
> (.\scratch\jython\pybench\pybench.py:903)
> > > > > > > > > > >        at org.python.pycode._pyx1.call_function
> (.\scratch\jython\pybench\pybench.py)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyFunction.__call__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyMethod.__call__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyObject.__call__(Unknown
> Source)
> > > > > > > > > > >        at
> CommandLine$py.__init__$15(Z:\UBS\Storage\.\scratch\jython\pybench\CommandLine.py:346)
> > > > > > > > > > >        at
> CommandLine$py.call_function(Z:\UBS\Storage\.\scratch\jython\pybench\CommandLine.py)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyFunction.__call__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyInstance.__init__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyClass.__call__(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyObject.__call__(Unknown
> Source)
> > > > > > > > > > >        at org.python.pycode._pyx1.f$0
> (.\scratch\jython\pybench\pybench.py:946)
> > > > > > > > > > >        at org.python.pycode._pyx1.call_function
> (.\scratch\jython\pybench\pybench.py)
> > > > > > > > > > >        at org.python.core.PyTableCode.call(Unknown
> Source)
> > > > > > > > > > >        at org.python.core.PyCode.call(Unknown Source)
> > > > > > > > > > >        at org.python.core.Py.runCode(Unknown Source)
> > > > > > > > > > >        at org.python.core.__builtin__.execfile_flags(Unknown
> Source)
> > > > > > > > > > >        at org.python.util.PythonInterpreter.execfile(Unknown
> Source)
> > > > > > > > > > >        at org.python.util.jython.main(Unknown Source)
> > > > > > > > > > >        at dacapo.jython.JythonHarness.iterate(
> JythonHarness.java:38)
> > > > > > > > > > >        at dacapo.Benchmark.run(Benchmark.java:126)
> > > > > > > > > > >        at dacapo.TestHarness.runBenchmark(
> TestHarness.java:302)
> > > > > > > > > > >        at dacapo.TestHarness.main(TestHarness.java
> :242)
> > > > > > > > > > >        at Harness.main(Harness.java:5)
> > > > > > > > > > >        at java.lang.reflect.VMReflection.invokeMethod(
> VMReflection.java)
> > > > > > > > > > >        at java.lang.reflect.Method.invoke(Method.java
> :317)
> > > > > > > > > > >        at org.apache.harmony.vm.JarRunner.main(
> JarRunner.java:80)
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > java.lang.IllegalArgumentException:
> > > > > > > > > > > java.lang.IllegalArgumentException: Cannot create
> PyString from null!
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > Thanks.
> > > > > > > > > > > Vladimir.
> > > > > > > > > > >
> > > > > > > > > > > On 11/6/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > > > I've committed the patch at r592434.
> > > > > > > > > > > > the excluded classes are,
> > > > > > > > > > > >
> > > > > > > > > > > > tests.api.java.util.CurrencyTest
> > > > > > > > > > > > tests.api.java.util.FormatterTest
> > > > > > > > > > > > tests.api.java.util.GregorianCalendarTest
> > > > > > > > > > > > tests.api.java.util.LocaleTest
> > > > > > > > > > > > tests.api.java.util.ScannerTest
> > > > > > > > > > > >
> > > > > > > > > > > >
> org.apache.harmony.text.tests.java.text.DecimalFormatSymbolsTest
> > > > > > > > > > > >
> org.apache.harmony.text.tests.java.text.NumberFormatTest
> > > > > > > > > > > >
> org.apache.harmony.text.tests.java.text.SimpleDateFormatTest
> > > > > > > > > > > >
> > > > > > > > > > > > pls kindly let me know if you find any problem,
> thanks.
> > > > > > > > > > > >
> > > > > > > > > > > > On 11/6/07, Tony Wu <wu...@gmail.com> wrote:
> > > > > > > > > > > > > On 11/6/07, Tim Ellison <t....@gmail.com>
> wrote:
> > > > > > > > > > > > > > Tony Wu wrote:
> > > > > > > > > > > > > > > I've raised a JIRA[1] for migrating the
> dependencies of locale related
> > > > > > > > > > > > > > > data to icu4j and remove the data in
> > > > > > > > > > > > > > >
> luni/src/main/java/org/apache/harmony/luni/internal/locale/. Currently
> > > > > > > > > > > > > > > I have delegated all the locale related
> classes to corresponding
> > > > > > > > > > > > > > > classes in icu4j and successfully removed the
> dependencies of harmony
> > > > > > > > > > > > > > > resource bundles.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Cool -- this is good because it removes the code
> duplication we have at
> > > > > > > > > > > > > > present, it shifts the maintenance of the locale
> data into the right
> > > > > > > > > > > > > > place (IMHO) that being the ICU project, and
> gives us the opportunity to
> > > > > > > > > > > > > > use the ICU tools to customize/refresh/etc the
> data for Harmony each
> > > > > > > > > > > > > > time the USA decide to introduce another
> daylight savings change ;-)
> > > > > > > > > > > > >
> > > > > > > > > > > > > exactly.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > > As expected, there are around 20 failures of
> harmony test. Many of
> > > > > > > > > > > > > > > them are caused by data difference. I've
> listed all of them on harmony
> > > > > > > > > > > > > > > wiki[2] and will raise them to icu soon.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Ack.  I believe that (pretty much) everyone gets
> the data from the CLDR,
> > > > > > > > > > > > > > so I would expect any differences to be tests
> that have out of date
> > > > > > > > > > > > > > assumptions about the data.
> > > > > > > > > > > > >
> > > > > > > > > > > > > Yes, I'll check it based on CLDR.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > > By applying this patch, we can get 4.2 mega
> bytes decrease in harmony
> > > > > > > > > > > > > > > source code, including the svn data.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > ooh, say that again!  Removing 4.2Mb of
> redundant source code is excellent.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > > My proposal is to apply the patch on 5061
> first and exclude these
> > > > > > > > > > > > > > > tests which fail on different locale data.
> Then I'll take follow up
> > > > > > > > > > > > > > > action to contact icu team and get solution to
> move them out.
> > > > > > > > > > > > > > > Otherwise I'm afraid my fix will be outdated
> for waiting until icu
> > > > > > > > > > > > > > > team get ready to fix the data. Do you have
> any objection?
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Later, I'll take care of the performance issue
> of this delegation.
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Thanks in advance for your comments
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Sounds like a good plan.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Regards,
> > > > > > > > > > > > > > Tim
> > > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > >
> > > > > > > > > > > > > --
> > > > > > > > > > > > > Tony Wu
> > > > > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > >
> > > > > > > > > > > > --
> > > > > > > > > > > > Tony Wu
> > > > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > --
> > > > > > > > > > Tony Wu
> > > > > > > > > > China Software Development Lab, IBM
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > --
> > > > > > > > Tony Wu
> > > > > > > > China Software Development Lab, IBM
> > > > > > > >
> > > > > > >
> > > > > >
> > > > > >
> > > > > > --
> > > > > >
> > > > > > Tony Wu
> > > > > > China Software Development Lab, IBM
> > > > > >
> > > > >
> > > >
> > >
> >
> >
> > --
> > Tony Wu
> > China Software Development Lab, IBM
> >
>
>
> --
> Tony Wu
> China Software Development Lab, IBM
>



-- 
Sean Qiu
http://xiaoxia.turendui.com