Add an option to keep the traceID unique during the whole execution

[astamarr]

Hi,

We are using the tracing feature to quickly see all errors during an user session. It worked well, but since the may SDK update, the TraceID is regenerated everytime a scened changes or the user switch focus. It means that we can't easily see all errors during execution at the same time, and it makes debugging for us harder.

We use no transactions, spans or anything, we just want to see quickly all errors that happened during the same execution to have a quick and global idea of what went wrong and when.

(as documentation states : The Sentry SDK regenerates the trace ID whenever the game gains focus or the active scene changes. This creates smaller, more focused trace groups rather than one continuous trace from startup to shutdown, helping you identify truly related errors.)

Would it be possible to have an option at SDK init, that force the same trace ID so we can have one continuous trace from startup to shutdown?

[bitsandfoxes]

Hey @astamarr, happy to talk about this.
I'm curious, why does it have to be the full length of the game? Errors that occur minutes, maybe hours apart are not likely to be related. How does your debugging workflow look like? Maybe there is something already in Sentry that would support what you're looking for.

[astamarr]

Because in our case, they do get related from time to time!
We make turn based, "slow paced" 4X games with lots of systems working in parallel. One of these systems can face an error, then players do a bunch of other things (text readings, slow battles, alt+tab to read an encyclopedia...), and eventually get stucks minutes (or tens of minutes) later when they want to get to the next turn, when all of these systems synchronise. Sometimes, an IA failed order that happens 10 turns before eventually stuck the whole game later on.

I do realise that it's not regular software debugging workflow, but it was super effective for us to find "weird" cause and effects :D

We've added a custom tag 'execution ID' so we can see every error during a same game session, but it lacks the nice UI and chronological order that traces provides.

[bitsandfoxes]

I'll raise this internally, but here we are meeting where the concept of transactions and spans from the web-world meet with the needs of gaming and those don't map all too well.
Because even if were to surface session related events, even those do get refreshed i.e. when tabbing out of the game and back into it.