harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Oliver Deakin <oliver.dea...@googlemail.com>
Subject Re: [classlib][luni] Arrays.sort() incompatibility with RI
Date Wed, 16 May 2007 10:18:38 GMT
Agreed. Should sort() be expected to behave like the RI for invalid 
comparators? IMHO the answer is no. I think in this case a bug should be 
raised with Eclipse to get the test fixed.

Regards,
Oliver

Aleksey Ignatenko wrote:
> The only fix for all related issues is to have the same implementation of
> such functionality as RI, as you understand it is unreachable goal unless
> you use RI classlib. So therefore I would say the right solution is to 
> fix
> tests.
>
> Aleksey.
>
>
> On 5/16/07, Sergey Kuksenko <sergey.kuksenko@gmail.com> wrote:
>>
>> As we discussed the source of the errror is incorrect comparator.
>> However, I can suggest a patch to Arrays which makes Arrays.sort()
>> slightly
>> close to RI.
>> With the patch sorting algorithm is not changed, but 
>> ArraysSortTest.javais
>> passed as on RI.
>> The patch is attached to the issue.
>>
>> On 5/16/07, Aleksey Ignatenko <aleksey.ignatenko@gmail.com> wrote:
>> >
>> > From H-3339 comment:
>> >
>> > As the bug is in the tests then it is to be filed on Eclipse and this
>> JIRA
>> > closed with comment about Eclipse bug.
>> > Please, see https://bugs.eclipse.org/bugs/show_bug.cgi?id=176157 as
>> > example
>> > of Harmony related bug.
>> >
>> > BR,
>> > Aleksey Ignatenko.
>> >
>> > On 5/15/07, Sergey Kuksenko <sergey.kuksenko@gmail.com> wrote:
>> > >
>> > > Hi All,
>> > >
>> > > As you wrote in the JIRA the reason of failure is incorrect 
>> Comparator
>> > > which
>> > > is used in the test.
>> > > The Comparator violates the follwing rule from specification:
>> > > "The implementor must ensure that sgn(compare(x, y)) ==
>> > > -sgn(compare(y, x))for all
>> > > x and y."
>> > > So I thing that It would be more correct to fix a comparator from
>> test.
>> > >
>> > > Some words about bug2bug compatibility. The fact that RI has 
>> different
>> > > result on this comparator means nothing, because in case of 
>> different
>> > > incorect comparator RI may give a different unpredictable 
>> results. So
>> > > should
>> > > we check *all possible set of incorrect comparators* and move 
>> exactly
>> in
>> > > same results as done by RI?
>> > > I think that it will mean that we should have completely the same
>> > > implementation of sorting.
>> > >
>> > >
>> > >
>> > > On 5/15/07, Eugene Ostrovsky <eugene.s.ostrovsky@gmail.com> wrote:
>> > > >
>> > > > Hi all.
>> > > >
>> > > > Some of eclipse tests pass on RI but fail on harmony. See
>> > > > *HARMONY-3339<https://issues.apache.org/jira/browse/HARMONY-3339>
>> > > > * .
>> > > > The tests are incorrect. They pass on RI due to difference in
>> > > > implementation
>> > > > of Arrays.sort() method.
>> > > >
>> > > > Shall we alter our Array.sort() implementation to be consistent 
>> with
>> > RI
>> > > > ant
>> > > > to make these buggy tests pass?
>> > > >
>> > > > Thanks,
>> > > > Eugene.
>> > > >
>> > >
>> > >
>> > >
>> > > --
>> > > Best regards,
>> > > ---
>> > > Sergey Kuksenko.
>> > > Intel Enterprise Solutions Software Division.
>> > >
>> >
>>
>>
>>
>> -- 
>> Best regards,
>> ---
>> Sergey Kuksenko.
>> Intel Enterprise Solutions Software Division.
>>
>

-- 
Oliver Deakin
Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598. 
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU


Mime
View raw message