Download Latest Version
v19.14.0 source code.tar.gz (6.9 MB)
Get an email when there's a new version of stellar-core
| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| README.md | 2020-01-29 | 3.1 kB | |
| v12.3.0 source code.tar.gz | 2020-01-29 | 4.5 MB | |
| v12.3.0 source code.zip | 2020-01-29 | 4.9 MB | |
| Totals: 3 Items | 9.4 MB | 0 | |
Stability improvements
- Fixed a bug in meta data generation on older versions of the protocol, this would cause internal errors to attach invalid meta that could confuse downstream systems
- Fixed a bug in the liabilities invariant that would not detect certain classes of issues
- Removed index on account database, this increases performance post inflation removal
- Change default encoding used in postgres tables this helps increase performance of postgres deployments
- NB: this may cause stellar-core to hang on “Applying DB schema upgrade to version 12” for a few seconds, the first time this version runs.
- C++ compliance fixes
- Increase default page size for sqlite this improves performance when running with sqlite
- Add logging to help with diagnostics with corrupt state
- Fix a crash when stopping core while it’s catching up
- Change the way we track quorums in memory this reduces memory pressure when performing quorum intersection computation
- Improve the way we keep track of data used during consensus
- Fix build when building with both libc++ and the option
enable-extra-checks, this allows generating “checked” builds with more paranoid checks enabled - Fixed a race when catching up that could crash core in some cases
- Fixed the computation of percentiles when reporting metrics where the percentiles were computed from bad data
New features
- Added network topology survey that allows validators to perform analysis on the overall network’s connectivity graph. This will also allow us to make deeper changes to the P2P networking code.
- Added survey script that illustrates how to use the new survey code.
- Added an experimental feature for generating meta data without inserting it into the database while replaying in memory, this allows to speed up meta data generation (for historical data) by several orders of magnitude
- Added a warning message when maintenance is clearly misconfigured
- Added mode for testing network subsystems
- Added new metrics for reporting on time spent in I/O queues
Breaking changes
None