Re: [Boost-users] Re: boost::thread::join causes assertion (pthreads)

From: Ben Hutchings (ben.hutchings_at_[hidden])
Date: 2005-03-21 16:29:32

Thorsten Froehlich wrote:
> Ben Hutchings wrote:
> 
>> I'm not sure whether you mean that the thread has exited or that it 
>> has also been joined.
> 
> The thread ended normally as the thread's function returned.  There does 
> not appear to be any way to detect this in boost::thread.

Right.

>>> EINVAL, thus triggering the assertion res == 0.  Is this intentional 
>>> or a bug?
>>
>> Which part of this behaviour are you questioning?
> 
> That when I try to join a thread that is no longer running, which cannot 
> be reliably determined by the given thread class interface (well, it 
> cannot at all in fact), then getting an assertion upon try to join that 
> thread is a bug.

If the thread is supposed to be joinable, then of course this is a bug, yes.

>> The reasons why pthread_join may fail all represent bugs rather than 
>> unpredictable errors, hence the assertion. 
> 
> 
> Sorry, this is incorrect, see below.

I think you misunderstand me.  I don't mean that a failure of 
pthread_join necessarily indicates a bug in your program, but that it 
indicates a bug at some level.  The specification doesn't provide for 
any error conditions like lack of resources.

>  > Joining a thread makes the
> 
>> thread object an empty shell which has no further use.    Detaching a 
>> thread also makes it non-joinable and makes the thread object almost 
>> useless.  If your program attempts to join a non-joinable thread, 
>> that's the bug.  Otherwise, it is conceivable that you have found a 
>> bug in MacOS X or in Boost.Threads, but I'd be surprised.
> 
> 
> I am perfectly aware what happens in the code.  My code is certainly 
> correct and doing nothing invalid at all - the problem is with boost 
> thread behavior or interface (well, lack of it) to prevent the 
> assertion, which would only be possible by checking if a thread 
> terminated already, and checking for this in a reliable manner.  Boost 
> threads does not provide that facility. Thus, an assertion is incorrect 
> at that point in the boost thread join code.

If a thread is joinable then pthread_join must succeed whether it is 
called before or after the thread exits.  Similarly for 
boost::thread::join.  Since you apparently don't understand the 
semantics of joining, perhaps you have in fact made an error.  Why don't 
you post an example that shows the behaviour you're reporting?

Ben.