Re: Problem with Win32Ada... Please help!

[dewar@merv.cs.nyu.edu (Robert Dewar)]

megamix asks

<<I am trying to install Win32ADA.
Each time I compile 'win32-winnt.ads', I have this problem:
 
C:\Win32ada\src>gcc -c win32-winnt.ads
win32-winnt.ads:1599:05: warning: no more representation items for type 
"LARGE_INTEGER" defined at line 1404
win32-winnt.ads:1599:05: warning: no more representation items for type 
"anonymous1_t" defined at line 1393
win32-winnt.ads:1599:05: warning: no more representation items for type 
"LUID_AND_ATTRIBUTES" defined at line 1592
win32-winnt.ads:1616:05: warning: no more representation items for type 
"SID_IDENTIFIER_AUTHORITY" defined at line 1608
win32-winnt.ads:2642:05: representation item appears too late
win32-winnt.ads:2645:05: representation item appears too late
win32-winnt.ads:2658:05: representation item appears too late
win32-winnt.ads:2659:05: representation item appears too late
 
I'm using GNAT 3.09 for Windows 95.
Can you help me ?
Thanks...>>



First, the answer to what is happening is that you are using a version of the
Win32 bindings that contains obvious illegalities, and GNAT is giving very
clear diagnostics (you have representation items after the freeze points,
and GNAT is identifying both the freeze points and the rep items.

The fix is easy, move the rep items earlier so that they are legal.

It would be nice if the posted versions of this package could be fixed, but
this package is copyright Microsoft (megamix: I hope you have an SDK license,
or you are violating this copyright, see headers on the source files).

Unfortunately, no one seems to be maintaining the Win32 bindings, and because
of the copyright issue, it is not straightforward for (e.g. ACT) to post fixed
versions.

In future public releases of GNAT, we will at least include necessary patches
to fix the public version if it still has not been fixed.

The idea behind the Intermetrics contract to produce these bindings was
that we would have freely distributable versions of the bindings, and indeed
I think it would be perfectly possible to generate such a distributable
version, but the approach used by Intermetrics of doing direct translation
of copyrighted headers ended up (unintentionally on their part -- it was
certainly a reasonable approach technically) in this copyright entanglement.

Note that these illegalities are not somehow GNAT dependent. Any Ada 95
compiler that accepts this version of the bindings has an error (as the
old version of GNAT did, which is why these errors crept in!)

We (ACT and other Ada vendors) are trying to figure out how to resolve
this somewhat unsatisfactory situation, and we may be able to at least
succeed in getting this particular problem fixed!

Robert Dewar
Ada Core Technologies