[#LUCENE-1457] There are a few binary search implmentations in lucene that suffer from a now well known overflow bug

Lucene - Java

There are a few binary search implmentations in lucene that suffer from a now well known overflow bug

Created: 17/Nov/08 02:17 AM Updated: 25/Sep/09 04:23 PM

Return to search

Component/s:

None

Affects Version/s:

None

Fix Version/s:

2.4.1, 2.9

Time Tracking:

Not Specified

Lucene Fields:	New
Resolution Date:	19/Feb/09 09:59 AM

Description

« Hide

http://googleresearch.blogspot.com/2006/06/extra-extra-read-all-about-it-nearly.html

The places I see it are:

MultiSearcher.subSearcher(int)
TermInfosReader.getIndexOffset(Term)
MultiSegmentReader.readerIndex(int, int[], int)
MergeDocIDRemapper.remap(int)

I havn't taken much time to consider how likely any of these are to overflow. The values being averaged would have to be very large. That would rule out possible problems for at least a couple of these, but how about something like the MergeDocIDRemapper? Is there a document number that could be reached that has a chance of triggering this bug? If not we can close this and have a record of looking into it.

All

Comments

Work Log

Change History

Subversion Commits

Sort Order:

[ Permlink | « Hide ]

Michael McCandless added a comment - 03/Dec/08 11:47 PM

Committed revision 723149.

I fixed these cases, plus one more in FieldCache. Thanks Mark!

[ Permlink | « Hide ]

Michael McCandless added a comment - 19/Feb/09 01:37 AM

Reopening for backport to 2.4.1.

[ Permlink | « Hide ]

Michael McCandless added a comment - 19/Feb/09 09:59 AM

Committed revision 745798 on 2.4 branch.