lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Robert Muir <rcm...@gmail.com>
Subject Re: DirectSolrSpellCheckerTest failure
Date Sun, 28 Nov 2010 13:48:30 GMT
Thanks Shai!

I put up a patch on https://issues.apache.org/jira/browse/SOLR-2257 if
you are able to test.

The problem is the spell checking comparators don't have a tertiary
comparison (tie-break).
So the results are very undefined, depends on which order you add
documents to your index (for the index-based case), the order of your
text file (the plain text case), etc etc.

Somehow we need to setup hudson to use alternative runtimes (IBM, BEA,
Harmony, etc) to detect problems like this!

On Sun, Nov 28, 2010 at 7:30 AM, Shai Erera <serera@gmail.com> wrote:
> OS: Windows 7 Pro
>
> IBM's:
> java version "1.6.0"
> Java(TM) SE Runtime Environment (build pwa6460sr8fp1-20100624_01(SR8 FP1))
> IBM J9 VM (build 2.4, JRE 1.6.0 IBM J9 2.4 Windows 7 amd64-64
> jvmwa6460sr8ifx-20100609_59383 (JIT enabled, AOT enabled)
> J9VM - 20100609_059383
> JIT  - r9_20100401_15339ifx2
> GC   - 20100308_AA)
> JCL  - 20100624_01
>
> Oracle's:
> java version "1.6.0_21"
> Java(TM) SE Runtime Environment (build 1.6.0_21-b07)
> Java HotSpot(TM) Client VM (build 17.0-b17, mixed mode, sharing)
>
> Shai
>
> On Sun, Nov 28, 2010 at 2:16 PM, Robert Muir <rcmuir@gmail.com> wrote:
>>
>> Shai, now that I think the test reproducibility issue is resolved, i
>> need to pressure this test to reproduce your problem.
>> Can you tell me the IBM JDK version / OS version you used (I will use
>> IBM JDK, just in case its specific to that one)?
>>
>> On Sun, Nov 28, 2010 at 7:02 AM, Robert Muir <rcmuir@gmail.com> wrote:
>> > On Sun, Nov 28, 2010 at 6:55 AM, Shai Erera <serera@gmail.com> wrote:
>> >> Hmm .. I previously ran the test w/ IBM's JDK (6) and after seeing your
>> >> mail, I tried w/ Oracle's (6) -- to my surprise the latter succeeds.
>> >>
>> >> What happens in the test that the JDK version can affect?
>> >>
>> >
>> > After testing, It seems to be an issue only with solr tests.
>> >
>> > When solr indexes, it seems the order of the fields being indexed is
>> > not well-defined. Perhaps they are using some iterator with undefined
>> > order to iterate over the fields.
>> >
>> > I didnt mess with this, instead i changed RandomCodecProvider to be
>> > insensitive to the order of fields being added.
>> > In other words, say you have "fieldA" and "fieldB".
>> > If RandomCodecProvider wants to assign "Pulsing" to fieldA, and
>> > "Standard" to fieldB, this will always happen, regardless of whether
>> > you add fieldA before fieldB, or vice versa.
>> >
>> > Now i get the same codec mapping for solr tests with the same seed,
>> > across different jvms...
>> >
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
>> For additional commands, e-mail: dev-help@lucene.apache.org
>>
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message