In TestIndexWriterDelete#testUpdatesOnDiskFull especially between line 538 and 553 we could get a random exception from the MockDirectoryWrapper which makes the test fail since we are not catching / expecting those exceptions. I can make this fail on trunk even in 1000 runs but on realtime it fails quickly after I merged this morning. I think we should just disable the random exception for this part and reenable after we are done, see patch below! - Thoughts?
Index: lucene/src/test/org/apache/lucene/index/TestIndexWriterDelete.java =================================================================== --- lucene/src/test/org/apache/lucene/index/TestIndexWriterDelete.java (revision 1091721) +++ lucene/src/test/org/apache/lucene/index/TestIndexWriterDelete.java (working copy) @@ -536,7 +536,9 @@ fail(testName + " hit IOException after disk space was freed up"); } } - + // prevent throwing a random exception here!! + final double randomIOExceptionRate = dir.getRandomIOExceptionRate(); + dir.setRandomIOExceptionRate(0.0); if (!success) { // Must force the close else the writer can have // open files which cause exc in MockRAMDir.close @@ -549,6 +551,7 @@ _TestUtil.checkIndex(dir); TestIndexWriter.assertNoUnreferencedFiles(dir, "after writer.close"); } + dir.setRandomIOExceptionRate(randomIOExceptionRate); // Finally, verify index is not corrupt, and, if // we succeeded, we see all docs changed, and if --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org