[LUCENE-1075] Possible thread hazard in IndexWriter.close(false)

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.3
• Fix Version/s: 2.3
• Component/s: core/index
• Lucene Fields: New
• Assignee: Michael McCandless
• Reporter: Michael McCandless
• Created: 03/Dec/07 15:34
• Updated: 28/Aug/22 11:43
• Resolved: 04/Dec/07 22:19

Description

Spinoff from this thread:

http://www.gossamer-threads.com/lists/lucene/java-dev/55391

On reviewing the code I found one case where an aborted merge (from
calling close(false)) could write to files that a newly opened
IndexWriter would also try to write to.

I strengthened an existing test case in TestConcurrentMergeScheduler
to tickle this case, and also modified MockRAMDirectory to throw an
IOException if ever a file besides segments.gen is overwritten.

However, strangely, I can't get an unhandled exception to occur during
the test and I'm not sure why. Still I think this is a good defensive
check so we should commit it.