Simplify sync by cloning in a temporary directory

[mscherer]

See #72 and #73

[mscherer]

So I have not tested yet the code, just pushed to run CI, but there isn't one.

[jberkus]

Hey, this is not the desired approach.

The reason we use pull() against a semi-persistent META directory is that clone() can be very expensive if the target repo is large. For example, the Kubernetes community repo (which includes elections) is 125MB. We do not want to be cloning the whole thing every time a webhook fires.

So yes we want to check for all errors, but we do not want to always clone instead of pulling.

[mscherer]

Wouldn't a shallow clone alleviate that issue ?

[jberkus]

125MB is the shallow clone.

[mscherer]

Ok, so given the amount of commits not related to elections on the community kubernetes repo, a filter on the hook would also help, but that would just complexify the code on the hook to simplify elsewhere. Guess we just need the change for check_call then ?

[jberkus]

Yeah, both pull and clone should check for all error conditions