ref: Make span types more robust

[mydea]

With more strict TS settings, these could lead to problems:

1. startChild() should return a span interface, to match the interface implementation.
2. All properties that are optional in span/transactions need to also accept | undefined to be really correct - otherwise when we extend this and set undefined explicitly it complains there.

Fixes #10654

[github-actions]

size-limit report 📦

Path	Size
@sentry/browser (incl. Tracing, Replay, Feedback) - Webpack (gzipped)	77.82 KB (0%)
@sentry/browser (incl. Tracing, Replay) - Webpack (gzipped)	69.06 KB (0%)
@sentry/browser (incl. Tracing, Replay with Canvas) - Webpack (gzipped)	72.98 KB (0%)
@sentry/browser (incl. Tracing, Replay) - Webpack with treeshaking flags (gzipped)	62.67 KB (0%)
@sentry/browser (incl. Tracing) - Webpack (gzipped)	33.25 KB (0%)
@sentry/browser (incl. browserTracingIntegration) - Webpack (gzipped)	33.16 KB (0%)
@sentry/browser (incl. Feedback) - Webpack (gzipped)	31.16 KB (0%)
@sentry/browser (incl. sendFeedback) - Webpack (gzipped)	31.16 KB (0%)
@sentry/browser - Webpack (gzipped)	22.43 KB (0%)
@sentry/browser (incl. Tracing, Replay, Feedback) - ES6 CDN Bundle (gzipped)	76.11 KB (0%)
@sentry/browser (incl. Tracing, Replay) - ES6 CDN Bundle (gzipped)	67.64 KB (0%)
@sentry/browser (incl. Tracing) - ES6 CDN Bundle (gzipped)	33.6 KB (0%)
@sentry/browser - ES6 CDN Bundle (gzipped)	24.71 KB (0%)
@sentry/browser (incl. Tracing, Replay) - ES6 CDN Bundle (minified & uncompressed)	212.88 KB (0%)
@sentry/browser (incl. Tracing) - ES6 CDN Bundle (minified & uncompressed)	101.5 KB (0%)
@sentry/browser - ES6 CDN Bundle (minified & uncompressed)	74.01 KB (0%)
@sentry/browser (incl. Tracing) - ES5 CDN Bundle (gzipped)	36.83 KB (0%)
@sentry/react (incl. Tracing, Replay) - Webpack (gzipped)	69.43 KB (0%)
@sentry/react - Webpack (gzipped)	22.45 KB (0%)
@sentry/nextjs Client (incl. Tracing, Replay) - Webpack (gzipped)	87.2 KB (0%)
@sentry/nextjs Client - Webpack (gzipped)	50.37 KB (0%)
@sentry-internal/feedback - Webpack (gzipped)	17.12 KB (0%)

[timfish]

Oh wow, TS2416 is new to me 😂

If this fixes the issue, it's might be worth adding a test of some kind to ensure we don't break exactOptionalPropertyTypes anywhere else in the future!

[mydea]

TS2416

Yeah, it's a bit obscure and hard to reason about IMHO 😬 Let's see if that fixes it first, maybe we can opt-into this stricter setting in our TS E2E test 🤔

[mydea]

OK, I actually tried this on the reproduction, and turns out the updateName thing was a bit misleading - it was only complaining because the transaction type was not "correct" because we had mismatches between types in interface and in the classes. Now, I made all optional properties in span/transaction accept xxx?: string | undefined, to ensure this is actually correct. The repro built for me then!

[ansf]

exactOptionalPropertyTypes makes

 sampled?: boolean;

and

sampled: boolean | undefined

incompatible.

This is more in-line with the actual javascript runtime (e.g.):

const x = { sampled: undefined }
assert("sampled" in x)

[Lms24]

Hmm I guess this falls into the "if this is what we have to do then let's do it" category 😅