[concurrency-interest] Java Concurrency torture tests: need your help

Aleksey Shipilev aleksey.shipilev at oracle.com
Fri Oct 19 04:32:15 EDT 2012


Hi guys,

Following up on May discussion on volatile bug [1], I had started the
project to collect the suite of concurrency tests which are to expose
concurrency bugs [2], hopefully much more thoroughly than JCK and other
targeted tests do.

It had been going for a while, and so far it tests only the basic things
about primitives, arrays, making the infamous singleton tests, etc. etc.
My plan is hook up this suite for OpenJDK nightly testing, so don't
forget make a pull requests for your changes ;)

Here's the call for action! I would appreciate your contributions,
including but not limited to:

  a) reviewing the test correctness. Doug had been lightly supervising
this before, but "given enough eyeballs all bugs are shallow". Please
reference the exact github revision when referencing the test.

  b) running the suite on interesting hardware. Even though the tests
are scarce, I would be happy to see the results from J9/PPC and various
ARM implementations. There are couple of negative tests against legal
races (i.e. UnsafeDCLSingletonTest), it will be comforting to see they
fail expectedly on non-x86 hardware. If you do run the tests, please zip
up the results/ folder and post it somewhere.

  c) brainstorming the coverage area: what areas are missing (there are
lots!), and what should be covered in the first place.

  d) proposing basic testcases and/or pointing out known regression
tests assorted over the Internet which should be included into the suite.

  e) raising the general suggestions about usability, reliability,
methodology, etc.

Will appreciate your help!

Thanks,
Aleksey.

[1] http://cs.oswego.edu/pipermail/concurrency-interest/2012-May/009440.html
[2] https://github.com/shipilev/java-concurrency-torture


More information about the Concurrency-interest mailing list