[Mingw-notify] [ mingw-Patches-582964 ] Cast OLE error codes to longs

[<no...@so...>]

Patches item #582964, was opened at 2002-07-17 13:55
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=302435&aid=582964&group_id=2435

Category: w32api
Group: None
>Status: Pending
Resolution: None
Priority: 5
Submitted By: Gunnar Degnbol (gdegnbol)
Assigned to: Earnie Boyd (earnie)
Summary: Cast OLE error codes to longs

Initial Comment:
The OLE error codes are longs with high bit set. They
are defined in winerror.h as

#define CO_E_CLASS_CREATE_FAILED 0x80080001L

The high bit makes g++ promote them to unsigned longs.
When they are compared to a variable of type HRESULT
(typedef of long) it then complains about
signed/unsigned comparison (if -Wall).

The patch changes all these defines to 

#define CO_E_CLASS_CREATE_FAILED ((long)(0x80080001L))

Microsoft does the same in their headers.

The patch is made with this command:

perl -e "while (<>) {s/ (0x8.*L)\$/
((long)(\))/;print \sh }" \
 < winerror.h > winerror.h.new



----------------------------------------------------------------------

>Comment By: Earnie Boyd (earnie)
Date: 2002-08-06 08:43

Message:
Logged In: YES 
user_id=15438

I set it to pending so that I could review these old tracker posts to decide what to do with them.  Please leave it set 
as I've set them until further notice.

----------------------------------------------------------------------

Comment By: Gunnar Degnbol (gdegnbol)
Date: 2002-08-06 08:28

Message:
Logged In: YES 
user_id=485110

What is pending here? Should I send a patch doing as Danny
asked? I tried to argue that I'd done the right thing.


----------------------------------------------------------------------

Comment By: Gunnar Degnbol (gdegnbol)
Date: 2002-07-17 19:37

Message:
Logged In: YES 
user_id=485110

Maybe, but a HRESULT is just a long, and will stay so. And
winerror.h didn't depend on any other headers before. 

I only added the casts where it made any difference - all
the other constants are longs as they are.

----------------------------------------------------------------------

Comment By: Danny Smith (dannysmith)
Date: 2002-07-17 18:57

Message:
Logged In: YES 
user_id=11494

Shouldn't they be cast to HRESULT, since that is what  
the OLE procedures return?

eg:

#define E_UNEXPECTED ((HRESULT)0x8000FFFFL)
 ...
etc. 
 for _all_ the OLE error codes (and success codes).

Danny

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=302435&aid=582964&group_id=2435