77f
Subj : Re: bit field bug?
To   : borland.public.cpp.borlandcpp
From : "Jeff" <js15432@flightexplorer.com>
Date : Mon Nov 17 2003 12:24 pm

Thanks for all of your feedback... byte alignment is set to 1 byte.  That is
not the problem... I am aware of the pitfalls of bitfields... and the
compiler dependent way they are used... but this is more a bug than a
compiler dependent usage:

Some more details:

A structure that should be 15 bytes long is buily by BC 5.01 to be 16 bytes
long... a single pad byte is being inserted after on of my bitfields:

BYTE iOne : 6;
BYTE iTwo : 2;
pad byte is being put in here

Interestingly enough, if I insert a dummy field so I can check the offsets
of each bitfield:

BYTE iOne : 6;
BYTE iTwo : 2;
char pcTemp[];
..
..

the extra byte goes away and now the structure is the 15 bytes it is
supposed to be.

f I remember correctly a while back, I read something about a situation
where Borland C++ inserted blank padding bytes between bitfield elements for
no reason... I just need to find that document or figure out the ryme or
reason so I can get the alignment right...

Thanks,
Jeff


"Andrue Cope" <not.a.valid@email.address.sorry> wrote in message
news:VA.00001ff5.00be2c15@email.address.sorry...
> Jeff,
>
> > The app uses bitfields in data structures
> >
>
> Arrrrrggghh!
>
> Bitfields should be avoided as if they were the spawn of the Devil
> ]:-)
>
> The C++ standard has never bothered to specify how the bits are to
> stored. It is implementation specific. I can tell you now that
> different Borland compilers store them differently so trying to
> compile code from another vendor's compiler is really asking for
> trouble.
>
> I would:
>
>   1.Get rid of the bitfields.
>   2.Give the developer who used them a bit of a kicking or at least a
> talking to.
>
> --
> Andrue Cope
> [Bicester, UK]
>

.

0