test failures when built & run on linux...(from thread re derby 13)

[myrnap <my...@Golux.Com>]

Daniel John Debrunner wrote:

>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>Shreyas Kaushik (JIRA) wrote:
>  
>
>>     [
>>    
>>
>http://issues.apache.org/jira/browse/DERBY-13?page=comments#action_57179 ]
>  
>
>>I also ran the tests of derbylang and am attaching the diffs of them
>>    
>>
>as well. Some of the tests fail there and I don't know the exact reason
>for the failures.( I think they are not failures )
>
>I actually think you are seeing unrelated failures. It seems that if
>Derby is built and tested on linix or unix then about six tests fail, it
>looks like a difference in the number of characters in a long line. I'm
>not sure if it's the build process, ij or the test diff program that has
>the problem. If you build on Windows and test on linux or unix the all
>the tests pass.
>
>Dan.
>
>  
>
Hi, Can you tell me which tests fail? Maybe it should be at least 
documented.

Thx,
Myrna

Re: test failures when built & run on linux...(from thread re derby 13)

[Daniel John Debrunner <dj...@debrunners.com>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Shreyas Kaushik wrote:

> Sure, here are the tests that failed,
>
> derbylang/derbylang.fail:lang/desc_index.sql
> derbylang/derbylang.fail:lang/distinctElimination.sql
> derbylang/derbylang.fail:lang/inbetween.sql
> derbylang/derbylang.fail:lang/predicatesIntoViews.sql
> derbylang/derbylang.fail:lang/subqueryFlattening.sql
> derbylang/derbylang.fail:lang/triggerGeneral.sql


I have an idea of why some of these are failing, those that have long
VARCHAR columns with output truncated by ij. They contain diffs with an
& that is the truncation indicator. e.g. predicatesIntoViews.

The output contains runtime statistics output which in turn contains the
text of the SQL statement. The String from getString() of this column
output is truncated by ij as it execeeds the specified maxDisplayWidth.

If the text of the SQL statement is multi-line, or maybe just multi-line
with an initial comment (I haven't looked that deeply) then if Derby is
built on Windows that text contains '\r\n' for the new line(s). If Derby
is built on Unix/Linux the new line will be '\n'. Thus the length of the
runtime statistics String is different and truncating results in a
different last line. Note the new-lines in the rest of the
runtime-statistics output are consistently '\n'.
The actual bug for this could be in reading the script file in ij, how
ij parses SQL statements and passes them onto the engine, or how the
engine obtains the text of the SQL statement for runtime stats.

A quick workaround is to increase the max display width on those tests
to avoid the problem. I will try that and see how many of these tests
will pass that way.

Dan.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFB5X1YIv0S4qsbfuQRAqhBAJ9VBIbmWv8M205VvgrqM4M+CCYOZQCfS2XX
zFpCzfBtSkFxpQA4YXgwZCU=
=jaAT
-----END PGP SIGNATURE-----

Re: test failures when built & run on linux...(from thread re derby 13)

[Shreyas Kaushik <Sh...@Sun.COM>]

Sure, here are the tests that failed,

derbylang/derbylang.fail:lang/desc_index.sql
derbylang/derbylang.fail:lang/distinctElimination.sql
derbylang/derbylang.fail:lang/inbetween.sql
derbylang/derbylang.fail:lang/predicatesIntoViews.sql
derbylang/derbylang.fail:lang/subqueryFlattening.sql
derbylang/derbylang.fail:lang/triggerGeneral.sql

I have atatched the full report of the test run in JIRA for Derby-13.

thanks
Shreyas

myrnap wrote:

> Daniel John Debrunner wrote:
>
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Shreyas Kaushik (JIRA) wrote:
>>  
>>
>>>     [
>>>   
>>
>> http://issues.apache.org/jira/browse/DERBY-13?page=comments#action_57179 
>> ]
>>  
>>
>>> I also ran the tests of derbylang and am attaching the diffs of them
>>>   
>>
>> as well. Some of the tests fail there and I don't know the exact reason
>> for the failures.( I think they are not failures )
>>
>> I actually think you are seeing unrelated failures. It seems that if
>> Derby is built and tested on linix or unix then about six tests fail, it
>> looks like a difference in the number of characters in a long line. I'm
>> not sure if it's the build process, ij or the test diff program that has
>> the problem. If you build on Windows and test on linux or unix the all
>> the tests pass.
>>
>> Dan.
>>
>>  
>>
> Hi, Can you tell me which tests fail? Maybe it should be at least 
> documented.
>
> Thx,
> Myrna