Implement locktower voting #3251
Conversation
aeyakovenko
force-pushed
the
locktower_voting
branch
from
c2d23ec
to
4976756
Compare
Codecov Report
@@ Coverage Diff @@
## master #3251 +/- ##
========================================
+ Coverage 78.2% 82.2% +3.9%
========================================
Files 153 133 -20
Lines 21746 20523 -1223
========================================
- Hits 17024 16883 -141
+ Misses 4722 3640 -1082 |
aeyakovenko
force-pushed
the
locktower_voting
branch
from
779f11e
to
b3101ae
Compare
carllin
force-pushed
the
locktower_voting
branch
6 times, most recently
from
f2b8778
to
1ac80cc
Compare
aeyakovenko
force-pushed
the
locktower_voting
branch
from
1ac80cc
to
16d30d8
Compare
|
@garious, any objections? |
core/src/bank_forks.rs
Outdated
| } | ||
|
|
||
| //produce a flat map of banks to all of its children | ||
| pub fn flat_children(&self) -> HashMap<u64, HashSet<u64>> { |
There was a problem hiding this comment.
The function name and the doc description need work.
core/src/locktower.rs
Outdated
| fn insert_fork_tree_lockouts( | ||
| stake_lockouts: &mut HashMap<u64, StakeLockout>, | ||
| vote: &Lockout, | ||
| flat_parents: &HashMap<u64, HashSet<u64>>, |
There was a problem hiding this comment.
We usually name HashMaps after the names of their values. In this case, I think that'd be parent_slots, but I can't tell for sure. That HashSet<u64> could use a type alias.
There was a problem hiding this comment.
it's such a simple hashset, i feel like an alias would just obscure it.
There was a problem hiding this comment.
Perhaps it's the u64 that really needs the type alias then.
There was a problem hiding this comment.
The combination of a bad parameter name and no alias for either HashSet or its contents makes this code pretty unreadable. Also, the function name itself is pretty tough on the brain to get the right context. Is it inserting "fork, tree-lockouts" or "fork-tree lockouts"?
|
@garious, Thanks! What do you think of the names? I think the naming could still use some work. can we handle it in another PR and have the fork-selection design doc reflect the naming? |
github comments
aeyakovenko
force-pushed
the
locktower_voting
branch
from
293bff0
to
8439f60
Compare
* locktower components and tests * integrate locktower into replay stage * track locktower duration * make sure threshold is checked after simulating the vote * check vote lockouts using the VoteState program * duplicate vote test * epoch stakes * disable impossible to verify tests
* locktower components and tests * integrate locktower into replay stage * track locktower duration * make sure threshold is checked after simulating the vote * check vote lockouts using the VoteState program * duplicate vote test * epoch stakes * disable impossible to verify tests
Problem
Fork selection in replay_stage does not use a sensible fork selection algorithm.
Summary of Changes
Implement locktower voting as described in the fork selection section of the book.
Fixes #