[classlib] Testing multithreaded behaviour

[Mark Hindess <ma...@googlemail.com>]

Those clever guys at google wrote so code to help testing multithread
behaviour:

  http://code.google.com/p/thread-weaver/

Anyone tried this?  I wonder if we could use this to avoid arbitrary
sleeps that slow down our tests and potentially make them unstable?

Regards,
 Mark.

Re: [classlib] Testing multithreaded behaviour

[Sean Qiu <se...@gmail.com>]

Sound interesting, our unit test on hand didn't cover this area.

On the other hand, maybe we can write our code in a thread safe way.
Then all exist tests can be reused in concurrent environment.

Is anyone familiar with this?

Best Regards
Sean, Xiao Xia Qiu



2009/11/13 Mark Hindess <ma...@googlemail.com>

>
> Those clever guys at google wrote so code to help testing multithread
> behaviour:
>
>  http://code.google.com/p/thread-weaver/
>
> Anyone tried this?  I wonder if we could use this to avoid arbitrary
> sleeps that slow down our tests and potentially make them unstable?
>
> Regards,
>  Mark.
>
>
>

Re: [classlib] Testing multithreaded behaviour

[Jesse Wilson <je...@google.com>]

On Fri, Nov 13, 2009 at 1:48 AM, Mark Hindess
<ma...@googlemail.com>wrote:

> Anyone tried this?  I wonder if we could use this to avoid arbitrary
> sleeps that slow down our tests and potentially make them unstable?
>

That's what Weaver is intended for! And it also lets you test concurrency
problems that are usually too complex to bother, which I think is more
important than the flakiness. And just having Weaver around makes us more
likely to spend time thinking through our tricky concurrency problems.

I recommend Weaver. (although I'm quite biased, scripts were my idea/fault)