Fix DEP0152 / deprecated PerformanceEntry.kind usage

[sirreal]

Fix #435

Node 16 deprecated PerformanceEntry.kind, moving it to PerformanceEntry.detail.kind.

Use entry.detail.kind if available. Fall back to deprecated entry.kind.

See https://nodejs.org/docs/latest-v16.x/api/deprecations.html#deprecations_dep0152_extension_performanceentry_properties

[analytik]

While I'm grateful for this, maybe we should also fix the deprecated format and an ignored parameter, see the node.js bug comment.

If think if (entry.detail.kind) passes, we should assume a newer Node.js version and also instead of { entryTypes: ['gc'], buffered: false } use { type: 'gc' } - it will likely be more future-proof.

[sirreal]

Sure, buffered: false appears to be completely redundant as it was the default value and has now been removed. I'll push a change for that.

Regarding type/entryTypes, I understand these are synonymous, where the second is compatible with older versions:

obs.observe({ type: 'gc' });
obs.observe({ entryTypes: ['gc'] });

{ type: 'gc' } may be the preferred version for observing a single type, but { entryTypes: ['gc'] } remains a valid way of observing one or more types according to the v16 documentation.

I don't believe checking event.detail for this will work. We would need to know whether to use type or entryTypes before we start observing, and we only see event after we've started observing. We may have to resort to inspecting process.version. I'd leave this as-is.

It may be worth asking for clarification of nodejs/node#38412 (comment) around type/entryTypes. Would you like me to do that?

[sirreal]

I dug a bit into the specs because nodejs/node#38412 (comment) mentioned aligning with them.

Both MDN and the w3c spec describe type and entryTypes. The difference seems to set the observer type to "single" or "multiple", respectively. It's unclear to me what the implications are or if there are downsides to a "multiple" type observer observing a single type.

The spec also mentions buffered, so perhaps I'm not looking in the right place.

[zbjornson]

Excellent, thank you. Sorry for the slow review.

I pushed to update the changelog and get CI to run.

[gruselhaus]

I'm running node v16.6.1 along with prom-client v13.1.0 and still get this deprecation warning.

Is anyone able to reproduce this?

[zbjornson]

@gruselhaus sorry, I just published v13.2.0 with the fix.