-
Notifications
You must be signed in to change notification settings - Fork 384
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Remove dependencies on k8s.io/apimachinery #225
Conversation
Instead, copy the error.NewAggregate implementation inside. This allows use of this package in OpenShift, which uses an older version of k8s.io packages. Signed-off-by: Miloslav Trmač <[email protected]>
Instead, inline net.SetTransportDefaults and copy net.NewProxierWithNoProxyCIDR. This allows use of this package in OpenShift, which uses an older version of k8s.io packages. Signed-off-by: Miloslav Trmač <[email protected]>
… broken again, by kubernetes/kubernetes@335ef74 . |
Sorry to distract.
I was working on a vendoring patch to help solve this problem in a few
spots (github.com/Microsoft has a few gems too if you guys have been
watching the things I tagged you on).
I think (open)containers needs to settle on a global vendoring solution.
This "rsync into dir" does not scale for the reasons we're seeing now.
A well defined, centralized dep list would go a long way towards stability
in the projects (due to matched deps) and also make it easier to isolate
problems.
However, I lack the political clout to carry out such an action. If this is
a desired thing I can put together the patches across
openshift/opencontainers/containers to sync this all up to standardized
versions and maybe settle on `hack/vendor.sh` or godep to keep the
repositories small since they only import parts of repos. I talked to alex
on the vndr project and he told me vndr cannot do subsets of repositories;
making it so that we have to import all of docker/docker for
docker/pkg/term.
Anyways, I can open an issue/set of issues if you think this is a good idea.
…On Mon, Jan 23, 2017 at 10:16 AM, Miloslav Trmač ***@***.***> wrote:
… broken again, by ***@***.***
<kubernetes/kubernetes@335ef74>
.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#225 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AABJ66bEhwFqISn3565i-cG-VAhnkJFRks5rVO5tgaJpZM4LrXEc>
.
|
While I would certainly like not having to care, the trouble is that we don’t want to freeze the versions forever; if this did succeed and we had 20 projects using a centralized dependency list, it would in turn be politically very difficult to change the list in the future. The various upstreams do add quite a few bug fixes, not only breaking changes, and we do want to get those bug fixes with little hassle :) Sharing a dependency list between
That may be misleading: with |
Hmm. @LK4D4 told me otherwise, perhaps he's willing to comment?
…On Mon, Jan 23, 2017 at 10:38 AM, Miloslav Trmač ***@***.***> wrote:
A well defined, centralized dep list would go a long way towards stability
in the projects (due to matched deps) and also make it easier to isolate
problems.
While I would certainly like not having to care, the trouble is that we
don’t want to freeze the versions *forever*; if this did succeed and we
had 20 projects using a centralized dependency list, it would in turn be
politically very difficult to change the list in the future. The various
upstreams do add quite a few bug fixes, not only breaking changes, and we
do want to get those bug fixes with little hassle :)
Sharing a dependency list between skopeo and containers/image does sound
very attractive, but with that we would probably have not encountered the
recent breaking changes for a good few months, if not years—which would
*not* make the porting work smaller, it would only aggregate it in time
to a single week (and probably a time when it would be much more urgent
than now, when we are haunted only by our own CI).
I talked to alex on the vndr project and he told me vndr cannot do subsets
of repositories; making it so that we have to import all of docker/docker
for docker/pkg/term.
That may be misleading: with vndr, one has to record an *entire*
repository in vendor.conf, and perhaps all of that repository is pulled
into a temporary directory, but the resulting vendor/github.com/docker
directory is pruned to only contain the actually used subpackages.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#225 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AABJ6whoyLlex9kv2dwtaSm3BjwejaIZks5rVPPDgaJpZM4LrXEc>
.
|
Either way, the vendoring tool doesn't concern me much; the centralized
list does.
I agree the list would be hard to change; that is a huge problem to solve.
I'll think about this and start another ticket when I have actual plans. :)
…On Mon, Jan 23, 2017 at 10:44 AM, Erik Hollensbe ***@***.***> wrote:
Hmm. @LK4D4 told me otherwise, perhaps he's willing to comment?
On Mon, Jan 23, 2017 at 10:38 AM, Miloslav Trmač ***@***.***
> wrote:
> A well defined, centralized dep list would go a long way towards
> stability in the projects (due to matched deps) and also make it easier to
> isolate problems.
>
> While I would certainly like not having to care, the trouble is that we
> don’t want to freeze the versions *forever*; if this did succeed and we
> had 20 projects using a centralized dependency list, it would in turn be
> politically very difficult to change the list in the future. The various
> upstreams do add quite a few bug fixes, not only breaking changes, and we
> do want to get those bug fixes with little hassle :)
>
> Sharing a dependency list between skopeo and containers/image does sound
> very attractive, but with that we would probably have not encountered the
> recent breaking changes for a good few months, if not years—which would
> *not* make the porting work smaller, it would only aggregate it in time
> to a single week (and probably a time when it would be much more urgent
> than now, when we are haunted only by our own CI).
>
> I talked to alex on the vndr project and he told me vndr cannot do
> subsets of repositories; making it so that we have to import all of
> docker/docker for docker/pkg/term.
>
> That may be misleading: with vndr, one has to record an *entire*
> repository in vendor.conf, and perhaps all of that repository is pulled
> into a temporary directory, but the resulting vendor/github.com/docker
> directory is pruned to only contain the actually used subpackages.
>
> —
> You are receiving this because you commented.
> Reply to this email directly, view it on GitHub
> <#225 (comment)>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/AABJ6whoyLlex9kv2dwtaSm3BjwejaIZks5rVPPDgaJpZM4LrXEc>
> .
>
|
@erikh everything unused is pruned, but what is used has the same version within project, so all deps which starts from i.e. |
gotcha; misunderstood. thanks alex.
vndr should work fine, then! :)
…On Mon, Jan 23, 2017 at 10:49 AM, Alexander Morozov < ***@***.***> wrote:
@erikh <https://github.com/erikh> everything unused is pruned, but what
is used has the same version within project, so all deps which starts from
i.e. github.com/coreos/etcd will have same version.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#225 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AABJ6yvLPBglosI0d_VTJ2GfC6a9-55Tks5rVPYvgaJpZM4LrXEc>
.
|
…o/client-go Because this is an entirely new package, depending on any version of it should hopefully not be problematic for downstreams. Signed-off-by: Miloslav Trmač <[email protected]>
Updated to also handle the |
this has been included in #223 |
Per #207 (comment) , depending on the latest versions of
k8s.io
makes this code very difficult to use inopenshift/origin
. So, drop dependencies onk8s.io/apimachinery/pkg/util/{error,net}
by inlining or copying the necessary parts.(This leaves
k8s.io/pkg/util/homedir
, which seems not to cause issues right now.)