Please merge nfs-utils 1:1.3.4-2.1 (main) from debian unstable (main)

[Balint Reczey]

There is at least one patch (90-gss-free-lucid-sec-context.patch) which is not present in Debian.

[Balint Reczey (rbalint)]

Debian's version builds fine with libevent 2.1 for which there is an ongoing transition [1] while Ubuntu's version FTBFS.

[1] http://people.canonical.com/~ubuntu-archive/transitions/html/html/libevent-2.1.html

[Balint Reczey (rbalint)]

I have uploaded the updated package to the PPA for testing the libevent 2.1 transition.
https://launchpad.net/~rbalint/+archive/ubuntu/libevent-2.1/+build/13193976

Please grab the source from there:
https://launchpad.net/~rbalint/+archive/ubuntu/libevent-2.1/+files/nfs-utils_1.3.4-2.1ubuntu1.dsc

[Steve Langasek (vorlon)]

Based on the linked Debian discussion (LP: #794112 -> Debian bug #648155), I believe that patch 20-ticket-expired-error.patch is irrelevant for Ubuntu releases beyond 12.04 because this behavior was fixed upstream in the kernel in commit eb96d5c97b0825d542e9c4ba5e0a22b519355166 present in kernel 3.10 and later. Do you agree with this conclusion, and should we drop the patch?

(I'm not sure that this warrants keeping around the non-upstreamed option to gssd for compatibility)

Otherwise this LGTM.

[Balint Reczey (rbalint)]

I agree that the patch became obsolete in Ubuntu unless someone tries running the package on a very old kernel, which may happen in case of some SoC systems. For example this SoC is still shipping but it uses a kernel based on v3.4: http://linux-sunxi.org/Linux_Kernel#Kernel_versions

Dropping the patch can also cause failure on systems passing -e as an option thus we may need to keep the -e option at least as a noop for some time.

Because of the reasons above I kept the patch for the merge and I think it would be reasonable to keep it.

[Steve Langasek (vorlon)]

On Mon, Aug 07, 2017 at 05:12:48AM -0000, Balint Reczey wrote:
> I agree that the patch became obsolete in Ubuntu unless someone tries
> running the package on a very old kernel, which may happen in case of
> some SoC systems. For example this SoC is still shipping but it uses a
> kernel based on v3.4: http://linux-
> sunxi.org/Linux_Kernel#Kernel_versions

However, running Ubuntu on these systems is not supported. It's true that
Linux 3.4 is currently the baseline kernel for Ubuntu userspace because that
was the kernel used in various Android BSPs for Ubuntu Phone enablement, but
that is not a requirement going forward. There's no expectation of
compatibility between Ubuntu userspace and arbitrary third-party SoC
kernels, and even if there were, this is a very niche use case
(kerberos-authenticated NFS clients) and it's been agreed that the bug is in
the kernel. We shouldn't feel obligated to carry a workaround for a bug in
kernels we don't ship.

> Dropping the patch can also cause failure on systems passing -e as an
> option thus we may need to keep the -e option at least as a noop for
> some time.

I don't think this is a requirement at all. This is a non-default option,
and the removal of the option will happen only as part of a release upgrade,
the point at which the behavior of the system with regard to config handling
is not expected to remain consistent. I would say that at most, this should
be a Debian.NEWS entry.

> Because of the reasons above I kept the patch for the merge and I think
> it would be reasonable to keep it.

I'm ok to sponsor this as-is for now, but could you please follow through
with a plan for removal of the patch?

Thanks,
--
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
Ubuntu Developer http://www.debian.org/
<email address hidden> <email address hidden>

[Balint Reczey (rbalint)]

Thank you.

I have opened a bug about dropping the patch: LP: #1709129 .