On 22/02/12 19:13, Stephen Allen wrote:
Hmm, now I'm seeing some inconsistent test failures on my Windows
machine. The tests work sometimes (such as when they are run in
isolation), but they occasionally fail when all tests are run ("mvn
test"). I've attached the test failures.
All tests work fine on my Linux machine.
I'm hazarding a guess that this may be related to the JDK bug with
memory mapped files not being released in Windows?
-Stephen
Stephen, thanks for the details.
Th tests "should" be windows-safe (independent files are created,
instead of reusing files - you should see lots of directories in the
test area). It wasn't an occasional - it was every time!
How occasional are these errors? 1 in 2? 1 in 10?
Both the errors are occurring in the test set up, not a test itself!
Things that occur to me:
1/ Antivirus software checking kicking in
2/ Two sets of tests running (and very bad luck!)
3/ err ...
At this point, I'm inclined to continue with the release.
Andy
Details:
TS_Block:
AbstractTestBlockMgr.before:
Previous tests in the run used this "before()" code OK!
This is what suggested AV to me.
The failure is in TestBlockMgrDirect so it should be a direct mode test,
not a mmap files issue.
FileOps.delete(filename) ;
* BlockAccess file = new BlockAccessDirect(filename, BlkSize) ;
Caused by: java.io.FileNotFoundException: target\tdb-testing\block-mgr
(Access is denied)
TS_Store:
TestGraphTDB.beforeClass:
(DatasetPrefixesTDB.java:175) =>
String prefix = t.get(1).getLiteralLexicalForm() ;
so t is null, or t.get(1) is null, both of which suggest something
messing around with the file (prefix table) behind the tests back.
