[LUCENE-6242] SparseFixedBitDocIdSet.ramBytesUsed() reports wrong size if alignment of JVM is not 8

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 5.1
• Lucene Fields: New
• Assignee: Unassigned
• Reporter: Uwe Schindler
• Created: 12/Feb/15 13:16
• Updated: 28/Aug/22 14:25
• Resolved: 13/Feb/15 15:16

Description

There seems to be a bug in SparseFixedBitDocIdSet's ramBytesUsed. To me this is a bit crazy implemented, so I have not yet found the issue. To me it looks like some of the summing up breaks if the alignment of the JVM is not 8:

   [junit4] Suite: org.apache.lucene.util.TestSparseFixedBitDocIdSet
   [junit4]   2> NOTE: reproduce with: ant test  -Dtestcase=TestSparseFixedBitDocIdSet -Dtests.method=testRamBytesUsed -Dtests.seed=
C1F3B881CB1C5E8A -Dtests.locale=es_BO -Dtests.timezone=America/Detroit -Dtests.asserts=true -Dtests.file.encoding=ISO-8859-1
   [junit4] FAILURE 0.06s J1 | TestSparseFixedBitDocIdSet.testRamBytesUsed <<<
   [junit4]    > Throwable #1: java.lang.AssertionError: expected:<264> but was:<256>
   [junit4]    >        at __randomizedtesting.SeedInfo.seed([C1F3B881CB1C5E8A:3350AAC1016341DC]:0)
   [junit4]    >        at org.apache.lucene.util.BaseDocIdSetTestCase.testRamBytesUsed(BaseDocIdSetTestCase.java:104)
   [junit4]    >        at java.lang.Thread.run(Thread.java:745)

To reproduce this failure, run inside core (with 64 bits JVM):

ant test -Dtestcase=TestSparseFixedBitDocIdSet -Dtests.method=testRamBytesUsed -Dargs="-XX:ObjectAlignmentInBytes=16" -Dtests.seed=C1F3B881CB1C5E8A

The default works:

ant test -Dtestcase=TestSparseFixedBitDocIdSet -Dtests.method=testRamBytesUsed -Dargs="-XX:ObjectAlignmentInBytes=8" -Dtests.seed=C1F3B881CB1C5E8A

I think we should randomly also specify the ObjectAlignmentInBytes for test runs on Policeman Jenkins. Any Power of 2 is fine.