* Move Proof and Difficulty types to pow module. Difficulty scaling calculation.
* Diffculty scaling and size shift for 2nd PoW
* Backport e7eb26ee
* Test compilation fixes
* Scaling only need to go one way, as @tromp pointed out
Currently on startup we wait for 30 secs if we don't have at least 4 peers with more work. If a node was quickly restarted it is already fully synced so thereare no peers with more work at all. This pr assumes that if we've done some work already and we have enough peers but still 0 with more work we are fully synced and good to go.
* improve: HeaderSync optimization (#1372)
* remove get_locator() optimization, which should be an independent pr for security review
* refactoring: move 'headers_streaming_body()' from Message to Protocol
* move 2 headers utils functions out of Protocol, and remove 'pub'
* support reading variable size of BlockHeader, from Cuckoo30 to Cuckoo36
* fix: use global::min_sizeshift() instead of hardcoded 30, because Cuckoo10 will be used for AutomatedTesting chain
* fix: should use global::proofsize() instead of hardcoded 42 when calculate serialized_size_of_header
* replace another 42 with global::proofsize()
* refactoring get_locator(), to ensure the gap between neighbours >= 2^n
* security enhancement for get_locator
* in case of sync head and the header head are different, clear history locators
Header sync used to run when complete (right count of blocks
received) or every 10 sec regardless. Sometimes 10 secs is not
enough, especially when a big batch of blocks bodies is arriving
at the same time.
This tweaks the time-based heuristic just a bit to only ask for
headers when stalled, meaning that the header head has not
*progressed* for 10 seconds.
* Fix bad header sync timeout logic
* Improve stability of fast sync test
Adds regular pings so that the 2 test servers know each others'
heights (necessary to find most work peer). Stop mining on the
first server once the 2nd kicks in. Avoids chasing a fast moving
target on Cuckoo10.
* fix: in case txhashset validation fail, 'TxHashsetDownload' should retry (#1293)
And add 2 more fixes:
1. handle sending request failure of 'send_txhashset_request';
2. use a 10-mins timeout to handle noresponse failure, in case a peer ignore our txhashset request, or in case a peer fail to send a file for some reason.
* improve: time crate flagged as deprecated, switch to actively maintained chrono crate
* improve: complete the switching from deprecated time crate to chrono crate, for all the tests part
* improve: complete switching to chrono crate, for the left tests in 'chain' and 'wallet'
* fix misbehaving simulnet fastsync test
cleanup redundant cutoff vs bitmap params is rewind and check_compact
* make sure we do not verify full kernel history on a writeable txhashset extension
rework simulnet simulate_fast_sync test to be more robust
* fixup store tests
* sleep for a bit longer to give nodes time to update
their sync_state correctly
* tweak timing of simulate_block_propagation
* A new sync status enum encapsulated in a state struct allows tracking of where sync is at. Leveraging it in the TUI to provide more helpful messages.
* Percentage progression for most sync steps
* Add kernel commitments sum and kernel and output MMR sizes to block header
* Sum a block without including previous sums, cleanup. Blocks are now summed and validated based on their own totals and not the totals since genesis. This allows to get rid of BlockSum and simplified the setting of a new block's roots, kernel sum and MMR sizes. Fixes#116
* Additional kernel MMR validation to check all prior header roots successively
* Wallet tests fix
Small refactoring of one folder, if it makes sense I could extend the scope.
* Remove some cloning (real and just verbosity in the code)
* Naming conventions like to/into*
* Some Clippy's suggestions
I found that we don't use field init shorthand syntax, so I didn't touch this part, was it discussed before?
Previous code used to always get full blocks from archival peers.
Fast sync nodes shouldn't need to, other fast sync nodes have all
the history they need. This only uses archival nodes when blocks
are requested at a height that's older than the horizon.
We test if we're in fast sync or not by checking if our head is
more recent than horizon. However, by the time we get the data and
validate, more blocks have passed, and we'd be after horizon again.
So we request the txhashset for a more recent block than the one
at horizon.
The current value was 20, leaving only 20 min of slack, which isn't
enough for debug builds and isn't future-proof. Change to use
`0.9*horizon`, leaving about 2.5 hours.
* wip BlockMarker struct, get rid of PMMRMetadata
* use rewind to init the txhashet correctly on startup, we do not need to track pos via metadata (we have block markers), we do not need to open the txhashset with specific pos (we have rewind)
* better logging on init
* keep rewinding and validating on init, to find a good block
* use validate_roots on chain init
* commit
* rustfmt
* only fast sync once
then treat as full sync
* commit
* add some debug logging so we can track progress when verifying rangeproofs
* rustfmt
Quentin Le Sceller