Re: R3.15.0 deadline approaching

[Michael Davidsaver <mdavidsaver@bnl.gov>]

On 5/30/2012 11:29 AM, Andrew Johnson wrote:
> ...
> > Ralph's server side plugins work.  The remaining work is to figure out
> > how best to run the unit tests on RTEMS and vxWorks.
> >
> > https://code.launchpad.net/~epics-core/epics-base/server-side-plugins
> I added the necessary test harness files to the src/ioc/db/test directory on
> the 3.15 branch yesterday.  We will probably have to make some changes to the
> tests to ensure that they can all be run within the same process image though
> (you can only run iocInit once); I haven't looked at the tests so this might
> need some creativity...

I imagine this will be tricky.  There is only one test which runs 
iocInit.  However several tests load and free a database.

> > And if I have time
> >
> > https://code.launchpad.net/~epics-core/epics-base/callback-error
> >
> > Although I think I will make some changes to this as well.  I'm having
> > second thoughts about adding a field to the CALLBACK structure which
> > existing code may not initialize correctly.
> If the user's source code needs to be changed you should ensure that the
> compiler aborts with an error if they haven't made that change.  This is how
> we handled similar API changes between 3.13 and 3.14.

Yes, I consider this a requirement as well.  However, I don't think 
there is a way to do this without effectively breaking all usage of 
CALLBACK.  The issue is that users are allowed to embedded these 
structures directly and there has been no official way to initialize 
them.  I can't think of a nice way of enforcing the use of memset() or 
similar.


Michael