[LUCENE-1143] DocumentsWriter.abort fails to clear docStoreOffset

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.3
• Fix Version/s: 2.3
• Lucene Fields: New
• Assignee: Michael McCandless
• Reporter: Michael McCandless
• Created: 20/Jan/08 16:30
• Updated: 28/Aug/22 11:45
• Resolved: 20/Jan/08 17:05

Description

I hit this in working on LUCENE-1044.

If you disk full event during flush, then DocumentsWriter will abort
(clear all buffered docs). Then, if you then add another doc or two,
and then close your writer, and this time succeed in flushing (say
because it's only a couple buffered docs so the resulting segment is
smaller), you can flush a corrupt segment (that incorrectly has a
non-zero docStoreOffset).

I modified the TestConcurrentMergeScheduler test to show this bug.
I'll attach a patch shortly.