Although rare it is possible to trigger polling of a block after avalanche has
concluded that it is invalidated. This is already not the case for finalized
blocks so make the behavior the same for invalidated blocks.
Details
Details
- Reviewers
sdulfari - Group Reviewers
Restricted Project - Commits
- rABCf363d342d2bc: [avalanche] Do not poll avalanche invalidated blocks more than once
ninja check check-functional
Diff Detail
Diff Detail
- Repository
- rABC Bitcoin ABC
- Branch
- invalids-not-worth-polling
- Lint
Lint Passed - Unit
No Test Coverage - Build Status
Buildable 22155 Build 43945: Build Diff lint-circular-dependencies · build-diff · build-clang-tidy · build-debug · build-without-wallet · build-clang Build 43944: arc lint + arc unit
Event Timeline
Comment Actions
There is already a similar mechanism for the proof, and we will also need it for the txs, so we should try to factor the feature or at least use the same mechanism so it can be efficiently refactored later.
| src/avalanche/processor.cpp | ||
|---|---|---|
| 604 | Don't use a vector and you don't need that either. This is time consuming for the sake of avoiding the vector from growing unbound. | |
| 1000 | dito it should be random access | |
| src/avalanche/processor.h | ||
| 305 | Don't use a vector when all you want is to check if the block belongs to the container | |
Comment Actions
D13219 is a draft to give an idea what the followup can look like to generalize this. Still needs updates to tests and renaming pending acceptance of this diff.
| src/avalanche/processor.cpp | ||
|---|---|---|
| 583 ↗ | (On Diff #38216) | |
| src/avalanche/processor.h | ||
| 307 ↗ | (On Diff #38216) | The chosen values for the filter make the structure ~90kB. This is OK but suboptimal: we don't need a capacity of 10k blocks, but we need a low false positive rate. I suggest to divide the capacity and the false positive rate by x100 (so the structure size remains the same). |