Hi Frederik,

> > The biggest potential problem is that the thread-safe locking
> > mechanism that we use does not extend into the C++ wrappers.
> > Addressing this is a fairly high-priority task for us. Pending funding,
> we'd like to get it into the library as soon as possible.
> 
> This sounds a bit odd to me. There is usually no need to thread-safe a thin
> "wrapper"-library separately if the underlying library is already thread-
> safe. Are there some specific problems that I should be aware of, like e.g.
> shared state between C++ objects or deadlocks?


The problem is that the wrappers are not always thin. Some of the wrapper code 
makes multiple HDF5 calls and state could change if there were a context switch 
during the C++ wrapper calls.

Dana

_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org

Reply via email to