[concurrency-interest] Concurrent sequential access to a mutable field without the volatile modifier

[David M. Lloyd]

On 01/07/2011 11:14 AM, Marco Villalobos wrote:
> Sébastien,
>
> It is very difficult to prove that thread unsafe code or bad
> concurrency will fail.

I've always wished there was a testing JVM emulator which was 
deliberately designed to exacerbate threading problems - it would, for 
example, never publish data between threads without a proper volatile or 
synch access, or do so randomly based on configuration.  Also it would 
be configurable to do stuff like deliberately starve certain threads, or 
inject context switches or allow the user to manually set up other 
different concurrency scenarios.

Such a thing would be devilishly hard to create, but it would be 
incredibly useful for testing concurrent code and algorithms.

-- 
- DML