Re: Bug with initDatabase()

[<greene%denali.UUCP@cub.kscorp.com>]

> > 
> > I've encountered a bug while porting EPICS to the Hytec CAMAC controller.
> > The problem is in initDatabase() which is called by iocInit().  The
> > interesting part is how this function ever worked.
> > 
> > The problem is in the first call to ellFirst().  If there are no
> > records defined for a specific type, then the field precLoc->preclist
> > is null.  However, the call to ellFirst() does not check for this
> > condition and will return a garbage pointer.  This works fine with
> > the NI VXI cpu030 card because the data at 0x0 is also 0x0 which causes
> > the for-loop to continue with the next record type.  However, with
> > the Hytec card, 0x0 is the start of EPROM which contains non-zero
> > data.
> > 
> > Obviously I can easily fix this by checking for a null preclist value
> > but I'm curious how this works on other controllers.  I'm also concerned
> > with other areas in the EPICS code which might pass a null ELLLIST value
> > to ellFirst().
> 
> I just looked at the code. A few ststements above the call to ellFirst is
> the statement
> 
>     if(!precLoc->preclist) continue
> 
> Thus if no records exist the call to ellFirst will never be made.
> 
> Marty Kraimer
> 

You must have version 3.12.  We haven't upgraded yet to that version.

This problem seems fairly pervasive - I've had to add a check in several
places so far, including initialProcess() and buildScanLists().  I get
similar problems running some of the IOC test routines, such as dbl(),
but I haven't yet confirmed that it is related.  My guess is that
even 3.12 hasn't fixed all the places where this occurs.


-- 
--------------------------------------------------------------------------------
				Greg Greene
				KineticSystems Corp.
				ph:    303-220-7455 x13
				fax:   303-220-7458
				email: greene@kscorp.com
--------------------------------------------------------------------------------