Re: [sv-cc] Handle lifetimes for one-time callbacks

Jim,

That fits what I thought, primarily because it always returns a handle 
and the rest of the rules are implications.  The last rule suggests that 
both remove_cb and a triggered one time callback would not be in the 
list of callbacks one can iterate over.

I would advocate the same future plan I mentioned, align with VHPI.  It 
cleans a couple things beyond this clarification.

Regards, John

Jim Vellenga wrote:
> I asked an ex-VPI developer at Cadence who once served on the 1364 PLI 
> Task Force about handle lifetimes for one-time callbacks such as 
> cbReadWriteSynch.  What he said (and presumably what NC and Verilog-XL 
> have implemented) is
>
> -- The routine vpi_register_cb() always returns a handle to the 
> callback.  If the application does not intend to use the handle again 
> later, it should call vpi_free_object() to release the handle.
>
> -- When the one-time callback triggers, the callback is removed, but 
> the handle, if not already released, persists.  The application should 
> call vpi_free_object(), if necessary, to release the handle.
>
> -- If the application calls vpi_remove_cb() before the callback has 
> triggered, vpi_remove_cb() both removes the callback and releases the 
> handle.
>
> -- It is an error for the application to call vpi_remove_cb() with a 
> handle to a one-time callback that has already triggered.
>
> Regards, 
> Jim Vellenga
>
> --------------------------------------------------------- 
> James H. Vellenga                            978-262-6381 
> Software Architect                     (FAX) 978-262-6636 
> Cadence Design Systems, Inc.         vellenga@cadence.com 
> 270 Billerica Rd
> Chelmsford, MA 01824-4179
> "We all work with partial information." 
> ---------------------------------------------------------- 
>
>   

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.