nolar / kopf / 29 / 9
87%
main: 89%

LAST BUILD BRANCH: settion_timeout
DEFAULT BRANCH: main
Ran
Files 56
Run time 2min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 84.662%. First build
CHANGE_MINIKUBE_NONE_USER=true KUBECONFIG=$HOME/.kube/config MINIKUBE_WANTUPDATENOTIFICATION=false MINIKUBE_WANTREPORTERRORPROMPT=false MINIKUBE_IN_STYLE=true MINIKUBE_HOME=$HOME MINIKUBE_VERSION=latest KUBERNETES_VERSION=latest CLIENT=yes

push

travis-ci

nolar 
Remember the actually invoked sun-handlers and purge their state

Previously, the sub-handlers' state was not purged, as the purge operation
works on specific keys only, and the sub-handlers' key are not known to the
moment when the purge happens (for example, because that main handler was not
invoked during this handling cycle, as it has finished in the previous one).

Scanning annotations for all the keys that start with the main handler's key
is: (1) feels wrong, since it is a write operation for one item, we should not
read/scan there; (2) would not work for long sub-handler ids due to hashing of
the suffixes to fit the annotation name into 63 chars (incl. the prefix).

Purging the sub-handlers' states in the `kopf.execute()` is also wrong: if the
main handler has 2+ calls of `kopf.execute()`, or fails after the call, the
main handler will be retried, while the sub-handlers information will be purged,
and so they will be re-executed — this violates the "succeed only once" promise
for handlers (regular and sub).

Therefore, the only way to properly purge the sub-handlers' state is to remember
the actual sub-handlers' ids in the parent handlers, mainly in the root handler
(intermediate parents remember the ids just for information), and explicitly
purge those states when the root handler's state is purged.

Note that a set of sub-handlers is not guaranteed to be fixed on multiple calls
of the main handler: the sub-handlers can depend on the object state, number of
retries, external environment circusmstances, etc — until the root handler fully
succeeds or permanently fails. This is one additional reason to keep the state
of the sub-handlers until the end of the handling cycle, and purge only when
the root handler is purged (together with all other root handlers).

1036 of 1297 branches covered (79.88%)

Branch coverage included in aggregate %.

3330 of 3860 relevant lines covered (86.27%)

0.86 hits per line

Source Files on job 29.9 (CHANGE_MINIKUBE_NONE_USER=true KUBECONFIG=$HOME/.kube/config MINIKUBE_WANTUPDATENOTIFICATION=false MINIKUBE_WANTREPORTERRORPROMPT=false MINIKUBE_IN_STYLE=true MINIKUBE_HOME=$HOME MINIKUBE_VERSION=latest KUBERNETES_VERSION=latest CLIENT=yes)