svn commit: r257696 - in head: libexec/rbootd share/man/man9 sys/compat/svr4 sys/net sys/sys

[Peter Wemm]

On 11/6/13, 9:00 AM, Adrian Chadd wrote:
> I think the important thing here is that there _are_ organisations
> that rely on some reasonable attempt at supporting historical APIs
> where needed.
>
> This IMHO should've explicitly gone into a compat macro for people who
> want support of this older stuff.
>
> My suggestion for a saner way to handle this deprecation schedule:
>
> * do the announce - I'd have to go looking for that, but we should be
> placing these somewhere obvious (like a wiki page that lists
> deprecated APIs in order, with the date/release they're going to be
> deprecated);
> * deprecate the userland use of the ioctl values first so they use the
> newer API;
> * deprecate the kernel API after the announced amount of time, hiding
> things behind COMPAT_xxx as appropriate.

That's how it was before - behind COMPAT_43 etc and he removed it. 
COMPAT_43 now does less than it did before.

There were a bunch of ioctl's that had been renamed ages ago with an O 
prefix that were exposed to userland.  While they weren't part of any 
API they should probably have been #ifdef COMPAT_43 in the includes to 
avoid accidental use.  However, things like ktrace/kdump/strace wanted 
access to them for decoding so it would have made the change even messier.

-Peter