From: pavenis AT lanet DOT lv
To: Waldemar Schultz <schultzma AT mathematik DOT tu-muenchen DOT de>, djgpp AT delorie DOT com
Date: Wed, 18 Apr 2001 15:14:59 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: MINGW vs DJGPP
Message-ID: <3ADDAF73.24905.59B576@localhost>
In-reply-to: <3ADD7C03.ACF8938C@ma.tum.de>
X-mailer: Pegasus Mail for Win32 (v3.12c)
Reply-To: djgpp AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

On 18 Apr 2001, at 13:35, Waldemar Schultz wrote:

> Eli Zaretskii schrieb:
> > 
> > On Wed, 18 Apr 2001, Waldemar Schultz wrote:
> > 
> > > BTW I couldn't resist to try the `-fpack-struct' switch Eli mentioned:
> > > now MINGW reports the size of the struct being _97_ vs _88_ without the
> > > switch.
> > 
> > That sounds like a bug in MinGW.  What version of GCC is that?  Is the
> > version of GCC you use for DJGPP compilation different?
> 
> MINGW>gcc -v
> Reading specs from
> E:\MINGW\BIN\..\lib\gcc-lib\i386-mingw32msvc\2.95.2\specs
> gcc version 2.95.2 19991024 (release)
> 
> DJGPP>gcc -v
> Reading specs from e:/djgpp/lib/gcc-lib/djgpp/2.952/specs
> gcc version 2.95.2 19991024 (release)
> 

-fpack-struct seems to be broken with gcc-2.95.X for DJGPP.

I tested following short test program:

#include <iostream>

struct foo { char foo_1;
             short foo_2;
             long foo_3;
             double foo_4; }    ;

foo x;

#define _offset_(a) (((char *) &x.a) - ((char *) &x.foo_1))

int main (void) { std::cout << sizeof(foo) << "    (" 
                  << _offset_(foo_1) << " "
                  << _offset_(foo_2) << " "
                  << _offset_(foo_3) << " "
                  << _offset_(foo_4)
                  << ")\n"; }

Here is output I got:

gcc-2.95.3 for DJGPP:    16    (0 2 4 8)
gcc-2.95.2 DJGPP to MINGW cross-compiler:   15    (0 1 3 7)
gcc-3.0 20010314 (prerelease) for DJGPP:   15    (0 1 3 7)

I didn't test other versions.

My suggestion:   put 

#pragma pack()

before definition of structure

and 

#pragma pack(1)

after it. I'm using it already for a long time without problems (except 
early prereleases of gcc-2.95 where #pragma pack() was buggy).
This is better way than -fpack-struct as it's only packs structures one
wants to pack.

Andris