assurances/3_007 panics on Fib(9) without checkpoint but the new service and bless changes are in the post-state

[jaymansfield]

So in Assurances/3_006 there are two services that are accumulated. Both end up panic according to assurances.txt.

When a panic occurs, the collapse function discards changes by reverting to Yu so I don't think the new service 2227924372 should be in post state.

[sourabhniyogi]

fixed -- no new services are created in assurances/3_006 (Fib(9))

[jaymansfield]

@sourabhniyogi assurances/2_011 has a similar issue now with the collapse function.

Service 2953942612 ends in panic but includes an assign host call and then a checkpoint host call after.

The post-state does not have any record of the assign host call but the accumulate state should have been persisted to Yu at the checkpoint. Only changes after the checkpoint should be discarded.

My phi value:
phi=[AuthQueue{authQueue=[0x6a0d4a19d199505713fc65f531038e73f1d885645632c8ae503c4f0c4d5e19a7, 0x6a0d4a19d199505713fc65f531038e73f1d885645632c8ae503c4f0c4d5e19a7, 0x6a0d4a19d199505713fc65f531038e73f1d885645632c8ae503c4f0c4d5e19a7, 0x6a0d4a19d199505713fc65f531038e73f1d885645632c8ae503c4f0c4d5e19a7, .......

All other assurances vectors pass for me with the latest release.

[sourabhniyogi]

solved here -- Thank you for a thorough check!

[jaymansfield]

@sourabhniyogi Thought we had it this time, but in the final vector (3_006) we have the issue again.

Service 2953942612 ends in panic without a checkpoint, but the new service and bless changes were not reverted and are included in post state.

UPDATE: After the vectors were re-generated the issue is now in 3_007 (and not 3_006).

[sourabhniyogi]

Ok, unless we have another issue from 0.6.2.9 (from another team), we will get this perfected in 0.6.3.0, probably need 3-4 days to get there.

[qiweiii]

I can confirm we (boka) also encountered this issue in state transition 2011 and 3007

[sourabhniyogi]

Updated dataset and tagged 0.6.2.10.

Notes:

1. @jaymansfield @qiweiii This fixes assurances/3_007 in that
   (a) the NEW service (2227924372) is NOT in the post-state and
   (b) the BLESS host result (m=0 a=1 v=1 g=map[0:911911]) is NOT in the post-state

2. @qiweiii Even though both assurances/2_011 (Fib(6)) and assurances/3_007 (Fib(9)) have a panic, in assurances/2_011 there is a CHECKPOINT, so almost all the state (everything except this) makes it through. In contrast, 3_007 does not have a checkpoint, so nothing makes it through.

[jaymansfield]

Looks good! All data sets pass for JavaJAM now.