On Monday 16 February 2009 14:47:52 Wolfgang Denk wrote:
> In message <200902161429.10360.vap...@gentoo.org> you wrote:
> > > > at any rate, is there a define that can be keyed off of ? 
> > > > CONFIG_HAS_FDT or some such junk ?  then the behavior can be changed
> > > > to like: #ifndef CONFIG_HAS_FDT
> > > >         /* ABI compat junk */
> > > >         uchar bi_padding_was_enetaddr[6];
> > > > #endif
> > >
> > > Why not just leave the bd_t alone, until such a time as we can get rid
> > > of it altogether?
> > >
> > > Note that just because we support booting with device trees doesn't
> > > mean we don't also support booting legacy kernels.
> >
> > the name has to go.  u-boot no longer uses that field and leaving it in
> > will simply lead to code misusing it creeping it back into the tree. 
> > whether you choose to force the warts on everyone is up to you ... i'm
> > not a ppc consumer so i could care less.
>
> You are not permitted  to  change  this  interface.  It  must  remain
> stable.  There  are  systems  out  there  running  really  old kernel
> versions, and every now and then they may even need to update  U-Boot
> to adapt it for recent hardware changes, while they cannot change the
> kernel.  It  is  mandatory  that even old 2.4 kernel versions must be
> bootable with recent versions of U-Boot.
>
> As mentioned before, bd_info is a binary interfwace between U-Boot and
> the Linux kernel, and it must not be changed.

i'm not talking about changing the binary aspect of it.  i'm talking about 
changing the API.  two completely different things.
-mike

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot

Reply via email to