You might get some ideas by looking at how the Boehm GC does it Enrico
> On 7 Dec 2018, at 14:15, Rick McGuire <object.r...@gmail.com> wrote: > > Ignoring for a moment the elephant in the corner of the Apple deprecated (and > non-functional) sem_* functions, the *ix versions of the semaphore functions > are really sort of broken. > > For the event semaphores, one of the functions is a reset, which uses > sem_init() to perform the reset. This probably doesn't work (or at least, is > not guaranteed to work) because the documentation states that using > sem_init() on an existing semaphore handle is undefined. > > The interpreter code uses the SysSemaphore class which implements all of the > desired behavior, but since it uses pthread, it only works for the unnamed > case. It would not be a huge effort to dual path named vs. unnamed, but the > sem_init() problems still exist for the named case. > > There are similar problems with the Mutex functions. These are implemented > using the same semaphore functions so a lot of the same issues exist. Also, > I'm not convinced that what is implemented here is a true mutex function. > Again, the classes implemented in SysSemaphore can be used here, but only for > the unnamed cases. I have not been able to find any equivalent to windows > named mutex semaphores in the linux world. > > In summary, fixing the unnamed versions of these would be a fairly simple > thing, fixing the named versions, I have no solutions. > > Rick > > > _______________________________________________ > Oorexx-devel mailing list > Oorexx-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/oorexx-devel _______________________________________________ Oorexx-devel mailing list Oorexx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/oorexx-devel
