At 12:59 PM 7/28/00 +0100, Mike Nordell wrote:

>I've now managed to build gds32.dll and the NT server .exe!
>OK, it's not working very well, upon connect it never responds and it chews
>a couple of houndred megs of memory befor I have to kill it, but at least it
>starts! :-)

Congratulations.

>Now for the interesting questions:
>I'm not really familiar with IB, and Ann explained that gpre is used to
>generate "largely system independent" code. As I understand it, but I
>definitely could be wrong due to ignorance in this case, the (only?) reason
>gpre needs access to a (a specific?) already generated DB, at least when
>building IB itself, is due to the fact that it needs to know the on-disk
>format of the database for the specific platform.

Gpre needs access to the table, column, and procedure definitions for
the database. There's no reason why the database has to be on the
same platform - gpre doesn't care about the ODS. I said "largely" because
when we started, C was not completely transportable and there are still
alignment issues. I believe we solved the alignment problems by ordering
elements of a structure to eliminate the problem. So, it's likely that
a C program generated by gpre will compile and run on any platform ...
except Cray.

>If this is the case, wouldn't it be easier to put this on-disk-format into
>platform specific directories/header/libraries to drop the dependency on an
>already running DB when building IB?

We've got a project active to let gpre read the metadata from flat files.

Ann