Commit Graph
2 Commits
Author SHA1 Message Date
Myasnikov Daniil cc942fed68 test(tests): run unit suites with Bats and retain POSIX coverage
The shell translator rewrites test bodies and loses Bats reporting.
The unit lane needs complete TAP results and JUnit reports without
changing how live-cluster suites run.

Sourced production helpers still need a dash pass, and unit test bodies
need nounset without exporting shell options into subprocesses.

Assisted-by: LLM
Signed-off-by: Myasnikov Daniil <myasnikovdaniil2001@gmail.com>
2026-10-07 11:04:00 +03:00
Marian Koreniuk 9934014908 fix(capi): ship the core provider manifests the repository actually holds
#2946 added a startupProbe to capi-controller-manager so it survives the
certificate provisioning delay. It edited
capi-providers-core/files/core-components.yaml, which is the copy a reviewer
reads. The chart mounts files/components.gz, and that blob was never
regenerated, so the probe never reached a cluster.

The blob is regenerated from the YAML it is supposed to carry:

  gzip -9 -n -c files/core-components.yaml > files/components.gz

The decompressed blob now matches the YAML byte for byte, and the only change
it carries is the six lines of #2946.

This reaches new installs only. capi-operator re-applies a provider when the
hash of its spec and its config Secret changes; the ConfigMap holding the
components is in neither, so a cluster that already runs capi-providers-core
keeps its manager without the probe until the provider's version moves.
Whether existing clusters should get the probe, and how, is left open here on
purpose: moving spec.version and the ConfigMap name together would force the
re-apply, but that is a version decision for the package, not a side effect of
regenerating a blob.

.helmignore is corrected in the same change: `files/.*-components.yaml` is a
glob, so it matched only names beginning with a dot, and `helm package`
shipped the readable YAML alongside the blob that duplicates it.

hack/capi-core-components_test.bats holds the two files to the same bytes from
now on, so the next edit of one without the other fails in CI.

Signed-off-by: Marian Koreniuk <moriarti@cp.if.ua>
Assisted-by: LLM
2026-09-25 16:26:13 +02:00