[Lsb-infrastructure] reconciling devchk-generated size/offset data

[Wichmann, Mats D]

With devchk builds fairly clean now, I've got reams of
data that could be applied to the database, but I'm
hesitating, because I want to make sure we get things
set up cleanly.  Some earlier data may have already
made things a little messy.

Here's an example:

32-bit platforms are seeing the size of "struct _CompositeClassPart"
and "CompositeClassPart" as 20 while expecting 40; the
proposed fix is to update the ArchType value for 
Type=100647,100648 and Arch=1 to 20.  But that of
course will mess up the value for 64-bit architectures,
which are currently happy.

What we want, of course, is to have arch-specific entries,
which since I see this one I'd do; but in this case I
assume we also want the generic entry set to a size of 0,
or perhaps not marked as included.

Do we have a cleanup tool for this?


Secondly, there's the case of devchk, when it gets started
doing arch-specific entries, doing them for all kinds of
values, even when there's not a distinction.  In some
cases I think those should set the generic entry, in some
it should not - I think a structure can't mix generic and
arch-specific offsets/sizes, for example, that would 
probably be too hard for the scripts.  Any comments on
this, and making sure that the state is consistent?