BJ>> Ah....  That would explain it.....  Yes, sizeof() may
 BJ>> not work in all environments now.....  :(

 BS>> Well I just don't understand it, because it does
 BS>> normally work under Linux, maybe it's the arrays..

 BJ> The problem is that sizeof() returns 
 BJ> the number of bytes in
 BJ> memory.  By default, structures are not packed.  If the structure was
 BJ> packed, then sizeof() should have worked for reading and
 BJ> writing files.

 BS> Oh! I thought so, but I wasn't scure about it.

I thought of an additional item.....  Even a packed structure might round
up the sizeof() to be a multiple of 4 or 8 bytes....  That
would mean you would still need to specify the size via another means for
reading and/or writing the packed structure....  I haven't tested this....

 BJ> Wes has a define made to allow packing the structures
 BJ> under GCC but not on other compilers where the packing isn't needed.

 BS> It's needed for use with fx. OS/2 or Win32, BTW one of 
 BS> my links told that the user.bbs from a Max/2 can't be 
 BS> readed by Max/Linux

Ok....  I haven't tried that yet.  I will have either a conversion routine
or that fixed before my BBS transitions from OS/2 to Linux.....

 BJ> Any structure that has more than one item (which it should or you
 BJ> wouldn't have used a structure), and has elements that are not an
 BJ> exact multiple of 4 or 8 bytes may get located on 4 (or 8) byte
 BJ> boundaries, resulting in unused bytes in the 
 BJ> structures memory immage.

 BS> Ah!

 BJ> So, sizeof() is needed for memory copies, but fail when
 BJ> reading and/or writing binary files in these cases.....

 BS> I see :-)

 BS> Thanks for the explanation!

You're welcome.

Take care.....

Bob Jones, 1:343/41



--- Maximus/2 3.01