Re: [nbpp-devel] Concurrency in daemontest handleConnection() as a multithreader()

[Benjamin G. <ben...@bt...>]

On Wed, May 16, 2001 at 07:53:33PM +0100, Daniel Barron wrote:
> With a MultiThreader() daemontest, is there a concurrency issue with the
> handleConnection() code?  I.e. does the for() i integer have a race
> condition?  Or is there a seperate copy of the TestDaemon class object
> running in each thread and so is thread-safe?

You needn't worry about `i'.  There's no race condition, because `i' is a
local variable; i.e. a new `i' is created and destroyed (on the calling
thread's stack) within the scope of each handleConnection() method call.
Thread safety is only an issue with class members and static variables.

> Also I thought that sleep() caused the entire process to sleep and you needed
> to call a pthread sleep?

I think you're right; that should be a call to reliable_usleep(), an
internal nb++ function that just sleeps the current thread (using
select()).  I'll fix it.

> I hope I'm using the correct jargon for these questions!  I'm new to the
> concepts of multi-threading.

It takes some getting used to.  You might want to read something like
_Programming with POSIX Threads_ by David R. Butenhof (the Addison-Wesley
series tends to be quite good).

-- 
Benjamin Geer
http://www.btinternet.com/~benjamin.geer