This is the mail archive of the cygwin-developers@sourceware.cygnus.com mailing list for the Cygwin project. See the Cygwin home page for more information.
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Patch, Version 4: Problem solved


On Tue, 4 May 1999, DJ Delorie wrote:

> The alignment of A changes from 4 bytes (sizeof(int)) to 8 bytes
> (sizeof(double)), so the alignment of B (which contains A) must also
> change from 4 to 8.  The size of B is increased to a multiple of the
> alignment so that each element of an array of such objects is properly
> aligned.  IMHO, this is correct and expected behavior.

Exactly.

> If MS wants them "less aligned" you'll have to use gcc's alignment
> directives to reduce the alignment of that double to 4.

MS and GNU do agree on this for a change! The bug is elsewhere since as 
far as I can tell, MS and GNU are doing exactly the same thing (I did try 
out Anders' code posted to egcs-bugs, and get the same result, as expected
in this particular case, from MSVC 5.0 and EGCS-1.1.2).

Anders, are you sure the code for the particular structure is in the right
packing mode (I believe it should pack(4) for LUID_AND_ATTRIBUTES)?

I get the same:
  
  B: 12
  .a: 8, 0
  .i: 4, 8

and,

  B: 16
  .a: 8, 0
  .i: 4, 8


I also checked the code using `#pragma pack(4)', and get the same result
from both compilers. Here I get:

  B: 12
  .a: 8, 0
  .i: 4, 8

and,

  B: 12		<<< NOTE it's back to 12 for both compilers as expected.
  .a: 8, 0
  .i: 4, 8

Regards,
Mumit