Experimental Physics and
Industrial Control System

Marty Kraimer <[email protected]> · Fri, 30 Aug 1996 07:42:41 -0500

Mark M. Ito wrote:
> 
> Dear EPICS-guru-types,
> 
> I am having trouble building databases on a Sun Ultra running solaris
> 2.5. db2database bombs when I introduce one of our non-standard
> records. This procedure has worked for us on a HP. On the Sun I get a
> segmentation fault in strcmp in extensions/src/mdct/klib/field.h while
> calling value_not_default(). The strcmp is between value() and
> this->get_default_value() and the latter comes back null. What is
> kinda interesting is that if I take my default.dctsdr from the HP and
> ftp it over to the Sun, db2database is as happy as a clam. I am told
> that I should suspect data structure alignment differences between HP
> and Sun. Full-word boundaries? Double-word boundaries? Arch Deluxe?
> Whopper w/ cheese?
> 
> Anyone care to comment?
> 
> --
> Mark M. Ito, Thomas Jefferson National Accelerator Facility
> 12000 Jefferson Ave., Mail Stop 12H, Newport News, VA 23606
> Email: [email protected], Office: (757)249-7175, Pager: (757)680-7175

Since base release 3.13 uses all ascii database description and
instance files, alignment problems should not occur (assuming that
all code is compiled by compatible compilers).
Since db2database and default.dctsdr both go away for 3.13
it is worth pursuing this problem. Note that by relinking gdct
to use the 3.13 libDb it works with 3.13 databases. Details are in
the WWW release notes.

Marty Kraimer

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System