Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
populate-owners: Also slurp OWNERS_ALIASES
Some slurped repositories use OWNERS_ALIASES [1], so we need to pull
these in to resolve entries in their OWNERS. OWNERS_ALIASES can only
exist in the repo root [2]. That means we may need to namespace the
upstream aliases to avoid collisions when assembing the composite
OWNERS_ALIASES here.
Defining aliases in upstream repos will be useful for situations where
a particular repo team (e.g. the installer team) should have control
over other directories within this repositories. For example,
$ cat cluster/test-deploy/aws/OWNERS
approvers:
- installer-approvers
reviewers:
- installer-reviewers
is simpler and more maintainable with an auto-maintained
OWNERS_ALIASES than the copy/paste we used in 49f60b7
(cluster/test-deploy/aws/OWNERS: Delegate to openshift/installer,
2018-08-27, openshift#1290).
With script control over OWNERS_ALIASES, *this* repo can't define its
own aliases. But that doesn't seem like a large limitation, because
this repo is unlikely to need aliases that are not defined in *any* of
the upstream repositories. And if we need to grow a location for
local alias configuration, we can always add that in the future.
I'm using 'git clone...' over SSH to access the files, because that
provides access to private repositories. Using:
https://raw.githubusercontent.com/{organization}/{repository}/HEAD/{path}
only works if you have an auth token or the repository is public.
Using clone for two files is a bit heavy. Ideally we could pull just
the two files we're interested. Git's archive command supports that
use-case, but GitHub doesn't seem to have enabled the server side of
that transaction:
$ git archive --format=tar --remote ssh://[email protected]/openshift/installer.git HEAD OWNERS OWNERS_ALIASES
Invalid command: 'git-upload-archive '/openshift/installer.git''
You appear to be using ssh to clone a git:// URL.
Make sure your core.gitProxy config option and the
GIT_PROXY_COMMAND environment variable are NOT set.
fatal: The remote end hung up unexpectedly
$ git --version
git version 1.8.3.1
Until they enable that, I'm fine consuming more of their bandwidth
than is strictly necessary ;).
Using reflect in assertEqual gives us a nice, compact implementation.
But the output can be a bit hard to read. With an introduced error:
$ go test .
...
--- FAIL: TestOrgRepos (0.00s)
main_test.go:15: unexpected result: [0xc42006e1e0 0xc42006e240] != [0xc42006e2a0]
--- FAIL: TestGetOwners (2.30s)
main_test.go:15: unexpected result: &{Directories:[] Organization:openshift Repository:installer Owners:0xc42006e420 Aliases:<nil> Commit:aaa47f6a54f0fa0449d36de01ccb9f56729b608a} != &{Directories:[] Organiza
...
I'd mocked up an assertEqual based on comparing json.MarshalIndent
strings which produced:
--- FAIL: TestOrgRepos (0.00s)
main_test.go:30: unexpected result: [
{
"directories": [
"/tmp/populate-owners-389376240/a/b"
],
"organization": "a",
"repository": "b"
},
{
"directories": [
"/tmp/populate-owners-389376240/c/d"
],
"organization": "c",
"repository": "d"
}
] != [
{
"directories": [
"/tmp/populate-owners-389376240/a/b"
],
"organization": "a",
"repository": "b"
}
]
--- FAIL: TestGetOwners (2.38s)
main_test.go:30: unexpected result: {
"organization": "openshift",
"repository": "installer",
"owners": {
"approvers": [
"aaronlevy",
"abhinavdahiya",
"crawford",
"smarterclayton",
"wking",
"yifan-gu"
],
"reviewers": [
"vikramsk"
]
},
"commit": "aaa47f6a54f0fa0449d36de01ccb9f56729b608a"
} != {
"organization": "openshift",
"repository": "installer",
"owners": {
"approvers": [
"aaronlevy",
"abhinavdahiya",
"crawford",
"smarterclayton",
"wking",
"yifan-gu"
],
"reviewers": [
"vikramsk"
]
},
"commit": "2587b3ed493c18747f2b37e1ab6daebdd277631a"
}
and of course, there are a number of third-party libraries that do an
even better job (e.g. [3]). But I don't think this functionality is
worth an external dependency, and Steve doesn't think it's worth
maintaining much local code for comparison [4], so we're going with
the more compact reflect approach despite the less-useful output. And
the tests should all be passing anyway, right? ;)
The GIT_*_DATE tick business is based on [5]. I haven't bothered
running a new commit after each test.setup change, because we don't
need a different commit hash. In production, extractOwners will be
run on a fresh clone, so we will always have a different commit hash
if the content changes.
The stub populate-owners.sh wraps the 'go run' at Steve's request [6].
We no longer need anything Bash-specific, so I'm using a POSIX shebang
in the wrapper. The 'exec' allows us to replace the shell command
with Go without executing a new process [7] (although Go itself will
proceed to launch a few processes).
The O_TRUNC is an attempt to avoid occasional corruption like:
diff --git a/ci-operator/jobs/openshift/azure-misc/OWNERS b/ci-operator/jobs/openshift/azure-misc/OWNERS
index 132c684..bf0dec5 100644
--- a/ci-operator/jobs/openshift/azure-misc/OWNERS
+++ b/ci-operator/jobs/openshift/azure-misc/OWNERS
@@ -3,10 +3,11 @@
# See the OWNERS docs: https://git.k8s.io/community/contributors/guide/owners.md
approvers:
+- pweil-
- jim-minter
- kargakis
- kwoodson
-- mjudeikis
+is
- pweil-
reviewers:
- jim-minter
although the exact source of that corruption is not clear to me.
[1]: https://github.com/openshift/cluster-api-provider-aws/blob/44e974de04f496f9552ecd37b73cad01b6d69f4d/OWNERS_ALIASES
[2]: https://github.com/kubernetes/community/blame/0741bdfbb56dcd4829754560f79a2f3da32cb34f/contributors/guide/owners.md#L54
[3]: https://godoc.org/github.com/stretchr/testify/assert#Equal
[4]: openshift#1285 (comment)
[5]: https://github.com/git/git/blob/v2.18.0/t/test-lib-functions.sh#L128-L138
[6]: openshift#1285 (comment)
[7]: http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#exec- Loading branch information
