Re: safe use of pthread_xxx_init functions

[Larry Dwyer <Larry.Dwyer@xxxxxx>]

At 11:48 AM 2/15/2007, Wojtek Lerch wrote:
> Larry Dwyer wrote:
> >   It would be wrong for the implementation to reject the request 
> > to initialize the object because of a misguided desire on the part 
> > of the system implementors to protect the lazy/stupid/careless 
> > application developers from themselves.
> No it wouldn't:
>
> "The pthread_mutex_init() function may fail if:
>
> [EBUSY]
>     The implementation has detected an attempt to reinitialize the 
> object referenced by mutex, a previously initialized, but not yet 
> destroyed, mutex."
The operant condition is "may fail".

Mutex's are special cases anyway.  Since mutex objects are shared by 
multiple threads, and the locking of a mutex is done by many threads, 
there are special conditions imposed on the implementation that may 
be difficult to meet  unless the implementation is allowed to do 
something that consumes resources and/or maintains state beyond what 
the application does, such as allocating memory.  If the destroy was 
not called, the implementation-allocated-memory would become a memory leak.

It is OK for the application to abandon any memory it allocates.  It 
is not OK for the implementation to abandon hidden memory that it 
allocates on behalf of the application.

                Cheers,
                Larry

                Cheers,
                Larry