Fix off by 1 when checking state_t heights causing sync failure #759
Conversation
| std::vector<std::pair<cryptonote::blobdata, cryptonote::block>> blocks; | ||
| for (uint64_t i = 0; m_state.height < current_height; i++) | ||
| { | ||
| if (i > 0 && i % 10 == 0) | ||
| LOG_PRINT_L0("... scanning height " << m_state.height); | ||
| MGINFO("... scanning height " << m_state.height); |
There was a problem hiding this comment.
This (and the Recalculataing message above) was a useful bit of log info to have at L0 for rescans, even with not running with any debugging, to let the user know a potentially long operation is actually doing something: particularly because while it happens everything else is locked out. In theory we ought to hit this much less, but even just 10k blocks could take a minute.
There was a problem hiding this comment.
MGINFO always shows up at the default log level? My main concern here is that L0 prints as a warning. I added your suggestion that we store state 100 blocks back from the tip of the chain (in other PR) so keeping this as L0 would emit the message as a warning on startup every time which might concern operators.
There was a problem hiding this comment.
Oh, okay, I thought MGINFO was suppressed at default. Sounds good then.
Found when trying to sync mainnet and we fell onto a 3 block long alt-chain and try to reorg back to the canonical chain. The service node winner was reported incorrect because the service node list reverted 1 too many states.
Upon reconnecting with this fix, the canonical chain was rebroadcasted, detached hooks were called on service node list and correctly popped back to the right state and was able to sync back to the canonical chain.
@jagerman