transfer_for_ejected_service-1 wrong statistics

[vekexasia]

Accumulation statistics should not include service 2 (even if all zeros) because service 2 is not in R* at all in GP0.7.1

[jaymansfield]

This one is also not listed in the accumulate README (@davxy).

[davxy]

@vekexasia why service 2 is not in R*?

Let's focus on work_for_ejected_service-3:

1. Service 2 is in the queue and depends on a work package (d54f42578...) that will be accumulated in this block.
2. When d54f42578 is accumulated, it unlocks service 2.
3. Consequently, the work of service 2 becomes part of R* (as defined in §12.12 and related expressions).

---

Edit:

It should be noted that, at the time of execution, the code cannot be loaded since the entire account structure has been ejected. Statistics are still updated according to ${R^{*}}_{..n}$ regardless.

We're currently treating this "missing account" structure the same way we would handle "missing code" (GP ref).
However, this situation currently constitutes undefined behavior, as the GP idoesn't specify what to do here - it assumes the service account should still exist at this point (GP ref).
But in this case, we have e_d[s] = ∅. I plan to propose an amendment to the GP to address this with the behavior just described

[vekexasia]

Hello @davxy sorry I only found out after that there were multiple new tests with similar names. I am talking about transfer_for_ejected_service-1.

Service id 2 exists but is not in the queue like in -3. Instead in this service i can see a transfer + eject. when executing ∆+, the inner call to ∆* yields t* which contains service 2.

But service 2 is not included in R* because:

• given reports do not contain any digest/result of service id =2
• R! is based on R above
• the ready Queue does not contain sid =2. So neither should RQ

[jaymansfield]

Just asked the same question here #102 (comment)

[davxy]

@vekexasia @jaymansfield probably you are right. I'll have a look

[HanaYukii]

+1, we ran into the same issue with service ID 2 .

[davxy]

Should be fixed