svn commit: r309658 - head/sys/vm

[Warner Losh]

On Wed, Dec 7, 2016 at 8:36 PM, Gleb Smirnoff <glebius at freebsd.org> wrote:
> On Thu, Dec 08, 2016 at 01:07:06PM +1100, Bruce Evans wrote:
> B> >> M> +#ifndef BURN_BRIDGES
> B> >> M> +/*
> B> >> M> + * Provide compatibility sysctls for the benefit of old utilities which exit
> B> >> M> + * with an error if they cannot be found.
> B> >> M> + */
> B> >> M> +SYSCTL_UINT(_vm_stats_vm, OID_AUTO, v_cache_count, CTLFLAG_RD,
> B> >> M> +    (u_int *)NULL, 0, "Dummy for compatibility");
> B> >> M> +SYSCTL_UINT(_vm_stats_vm, OID_AUTO, v_tcached, CTLFLAG_RD,
> B> >> M> +    (u_int *)NULL, 0, "Dummy for compatibility");
> B> >> M> +#endif
> B> >>
> B> >> IMHO, there should be some garbage collecting timeout for them. I'd suggest to
> B> >> delete them from head after stable/12 branch is forked.
> B> >
> B> > Other places define #if BURN_BRUDGES < 1200000 to make this automatic.
> B>
> B> The correct ifdef is something like '#if COMPAT_FREEBSD12' to make this
> B> non-automatic.  Support for this option should of course be removed some
> B> time after the support for COMPAT_43 option.  That is only 25-30 years old.
>
> By the way, can anyone explain me why do we still have COMPAT_43 option?
>
> I also bet $100 against $10 that it is broken in head in many aspects,
> since there is no regression suite for it and since nobody uses it.

I'd say that we don't know what is or isn't used if it's working. We
only know it isn't working if it fails in a spectacular way for a
popular program (cf sbrk and emacs).

I'd also say those are about the odds I'd give for all the rest.

Warner