[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: FreeBSD 9.1 (was: Clang)

2013/6/23 Steven Chamberlain <steven@pyro.eu.org>:
> On 23/06/13 13:00, Robert Millan wrote:
>> You mean the kernel only? In my experience this is not as easy as it
>> looks. A new kernel often drags in the rest of FreeBSD userland due to
>> ABI changes.
> Yes I meant the kernel only;  I was optimistically thinking they
> wouldn't break ABI between minor versions but maybe that is fatally
> wrong to assume.

Maybe they do, maybe they don't. Their policy doesn't prevent it, and
it tends to happen. Often major subsystem changes are backported from
HEAD into 9-STABLE ("MFC'ed").

> I minimally tested running wheezy/sid system including
> ZFS on a 9.1 kernel some months ago.

Well it's difficult to tell the effect. There are lots of
non-essential things that could break. Now that you mention it, ZFS
admin tools is one of them.

CAM, PF, Jails, and in general anything that depends on freebsd-libs
is a good target for inspection.

> I think I could accept the risk.  I only found one or maybe two things
> that seemed viable to backport to 9.0, whereas many more people have
> driver issues that only an upgrade to 9.1 could fix.

So we proceed with 9.1 upgrade then?

> This is very much a personal opinion but maybe clang-3.2 could be easier
> to keep up with for the jessie cycle.  This puts me slightly in favour
> of trying clang-3.2 but unfortunately it is a preference not based on
> technical merit.

Technical or not, it has technical consequences. I think you make a valid point.

However if we do this, I would not mix Clang transition with 9.1
upgrade. Having two separate transitions makes it easier to isolate
problems if/when they occur.

Robert Millan

Reply to: