cvs commit: src/sys/dev/fxp if_fxpreg.h

[Stefan Farfeleder]

On Sun, Apr 06, 2003 at 02:16:07PM +1000, Bruce Evans wrote:
> On Sat, 5 Apr 2003, Maxime Henrion wrote:
> 
> > mux         2003/04/05 15:46:58 PST
> >
> >   FreeBSD src repository
> >
> >   Modified files:
> >     sys/dev/fxp          if_fxpreg.h
> >   Log:
> >   ...
> >   - Change some u_int to u_int8_t which make more sense here since
> >     we're really defining bytes.  That produces the same code due to
> >     how bitfields work.
> 
> This gives undefined behaviour and thus produces random code if it is
> compiled by a C compiler (unless Bool_t happens to be u_int8_t).  From
> n869.txt:
> 
>        [#8]  A  bit-field  shall have a type that is a qualified or
>        unqualified version of _Bool, signed int, or  unsigned  int.
> 
> I fixed this bug in many places, including in rev.1.13 of if_fxpreg.h.
> but it keeps getting reintroduced :-(.
> 
> Bit-fields of other integer types are an unportable gcc extension.
> They affect the struct layout in unportable apparently-undocumented
> ways.  IIRC, they don't affect internal padding but they do affect the
> size and alignment the struct -- a struct that has only uint8_t
> bit-fields in it has only the size and alignment requirements of
> uint8_t, while a struct with only u_int bit-fields in it has the size
> and alignment requirements of u_int.  This may be controlled to some
> extent using other unportable gcc extensions.

FYI, the final standard says

4  A bit-field shall have a type that is a qualified or unqualified version of _Bool, signed
   int, unsigned int, or some other implementation-defined type.

and moved it from the Semantics to the Constraints section, so a C
compiler not supporting u_int8_t has to issue at least a diagnostic
before producing random code :)

Regards,
Stefan Farfeleder