textlogger: allow caller to override stack unwinding

[pohly]

What this PR does / why we need it:

One use case is to use Ginkgo for stack unwinding. Then ginkgo.GingkgoHelper also works for log output, not just for E2E failures.

Special notes for your reviewer:

This is going to be used in extended support for test contexts. A preview of what's to come is currently in https://github.com/kubernetes/kubernetes/pull/122481/commits

Release note:

textlogger: stack unwinding can be implemented outside of klog

[pohly]

The apidiff errors is something that we can ignore:

 FAIL: k8s.io/klog/v2/textlogger contains incompatible changes:
- Config: old is comparable, new is not
- configOptions: old is comparable, new is not

We never said that the struct with no exported members is comparable and I can't think of a reason why someone would compare them anyway.

[dashpole]

/assign @mengjiao-liu
/triage accepted

[pohly]

@harshanarayana : can you perhaps take a look?

[harshanarayana]

@pohly ack

[pohly]

@harshanarayana : gentle ping 😄

I have some changes pending which depend on this.

[harshanarayana]

Do you think it is a good idea to have an if not nil check for the l.config.co.unwind ? I don't think anyone uses &Config{} to create the config. But just in case if they do and forget to call the Backtrace handler and set the unwind this might run into an issue. or do you think this is an overkill to add this check ?

[harshanarayana]

@pohly Just a minor comment. I am good otherwise.

[pohly]

My initial reaction was that someone might indeed be doing that because we can't prevent it. But then I remembered that it's explicitly not allowed:

klog/textlogger/options.go

Lines 33 to 34 in 2d08296

	// Must be constructed with NewConfig.
	type Config struct {

If we were to add a nil check now for members, then we would enter a slippery slope of supporting something that wasn't meant to be supported and make the code more complex. Besides, Config.output also doesn't have a nil check, so someone who isn't following the documentation is already crashing.

No nil check?

[harshanarayana]

That is a valid point. I think we can avoid the nil check then

[harshanarayana]

/lgtm

/cc @dims

[dims]

/approve
/lgtm
/skip

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: dims, harshanarayana, pohly

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [dims,pohly]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[pohly]

@dims : you might have to merge manually?