[Mondrian] State of the build

Sherman Wood swood at jaspersoft.com
Thu Jul 19 19:11:12 EDT 2007


The Javadoc problem is fixed.

Sherman Wood
JasperSoft

----- Original Message -----
From: "Julian Hyde" <julianhyde at gmail.com>
To: mondrian at pentaho.org
Sent: Friday, July 20, 2007 4:24:41 AM (GMT+1000) Australia/Sydney
Subject: [Mondrian] State of the build

Given that mondrian-2.4 coming up in the next 2 or 3 weeks, I thought I'd
give you an update on the state of the mondrian build and regress.

It would be great if the build ran clean every night. But given the
extensive test suite - 1500 unit tests, run on 5 database platforms, 3
versions of the JDK, multiple combinations of configuration parameters (over
20 runs per night, some of which are randomly generated each night) - I
think we're doing pretty well. Here's a list of the problems in last night's
run.

Full build output is attached.

1. Richard, your most recent change fixed the build on JDK 1.4 - thanks.

2. There's an intermittent exception in executeHR. It's been occurring for
several months now. No one is investigating this problem, as far as I know.
We can presume that it's caused by some combination of parent-child
hierarchies, XMLA, aggregates, and non-determinacy in the cell value cache.

Running test with JDK=jdk1.4 database=mysql props={ mondrian.test.VUsers=1
mondrian.rolap.aggregates.Use=true mondrian.rolap.aggregates.Read=true}
     [java] 1)
executeHR(mondrian.xmla.test.XmlaTest)junit.framework.ComparisonFailure:
expected:<...5537...> but was:<...6691...>
     [java] Tests run: 1445,  Failures: 1,  Errors: 0

3. Thiyagu, I think these exceptions relating to the DESCENDANTS function
are the result of the GROUPING SETS changes. Are you looking into these?

Running test with JDK=jdk1.6 database=oracle props={
mondrian.rolap.aggregates.Use=true mondrian.rolap.aggregates.Read=true
mondrian.rolap.aggregates.ChooseByVolume=true
mondrian.rolap.star.disableCaching=true mondrian.native.filter.enable=false
mondrian.native.nonempty.enable=false
mondrian.xmla.drillthroughTotalCount.enable=false
mondrian.olap.fun.crossjoin.optimizer.size=0
mondrian.rolap.groupingsets.enable=true}
     [java] 1)
testDescendantsMLSelfBeforeAfter(mondrian.olap.fun.FunctionTest)mondrian.ola
p.MondrianException: Mondrian Error:Internal error: Error while executing
query [select {Descendants([Time].[1997], [Time].[Quarter],
SELF_BEFORE_AFTER)} ON COLUMNS
     [java] 2)
testDescendantsMLBeforeAfter(mondrian.olap.fun.FunctionTest)mondrian.olap.Mo
ndrianException: Mondrian Error:Internal error: Error while executing query
[select {Descendants([Time].[1997], [Time].[Quarter], BEFORE_AND_AFTER)} ON
COLUMNS
     [java] 3)
testDescendantsMNY(mondrian.olap.fun.FunctionTest)mondrian.olap.MondrianExce
ption: Mondrian Error:Internal error: Error while executing query [select
{Descendants([Time].[1997], 1.0, BEFORE_AND_AFTER)} ON COLUMNS
     [java] 4)
testDescendantsSBA(mondrian.olap.fun.FunctionTest)mondrian.olap.MondrianExce
ption: Mondrian Error:Internal error: Error while executing query [select
{Descendants([Time].[1997], 1.0, SELF_BEFORE_AFTER)} ON COLUMNS
     [java] Tests run: 1352,  Failures: 0,  Errors: 4


4. Intermittent failure in testFlush (cache flushing)

Running test with JDK=jdk1.4 database=mysql props={
mondrian.test.ExpDependencies=10 mondrian.rolap.evaluate.MaxEvalDepth=6
mondrian.native.crossjoin.enable=false
mondrian.xmla.drillthroughTotalCount.enable=false
mondrian.rolap.LargeDimensionThreshold=1000
mondrian.rolap.SparseSegmentValueThreshold=10}
     [java] 1)
testFlush(mondrian.rolap.CacheControlTest)junit.framework.ComparisonFailure:
Expected:
     [java] Tests run: 1352,  Failures: 1,  Errors: 0

5. This SQL generation test is probably poorly written. With this
combination of properties, we would not expect that SQL to occur. I will
take a look.

     [java] 1)
testNativeSetCaching(mondrian.rolap.VirtualCubeTest)junit.framework.Assertio
nFailedError: expected query [select "store"."store_country" from "store" as
"store", "sales_fact_1997" as "sales_fact_1997" where
"sales_fact_1997"."store_id" = "store"."store_id" group by
"store"."store_country" order by 1 ASC] did not occur
     [java] Tests run: 1352,  Failures: 1,  Errors: 0

6. Broken javadoc. Sherman, it's yours.

/home/jhyde/open2/mondrian/src/main/mondrian/gui/I18n.java:264: warning -
@param argument "cID" is not a parameter name.

Julian




More information about the Mondrian mailing list