GitLab

sql_serializer plugin can be built right now on machine where PostgreSQL installation is missing. This was required to avoid adding hard PostgreSQL dependency to the hived, in cases when customers does not want to use sql_serializer plugin.

Right now concurrent data writes are synchronized properly during live sync, to satisfy FK constraints.
Also single_transaction_controller is in use to hold all changes in one transaction.

(i.e. not account_history_api)

[ABW]: [Fix] LIB now is always loaded and written in unison with changes in DB (same with reindex point); no longer read from DB every block - cached version used
Note that it only masked problem that was solved by flushing after every change - that solution was dropped since it kills performance

[ABW]: [Fix] apparently explicit commit is needed when new irreversible data is stored in AH DB, otherwise recently stired LIB could be read as if it wasn't written

Fixed hived node initialization problems when it continue execution immediately after loading a state snapshot (chain-id and other non persistently stored data have been improperly initialized)
Improved snapshot related messages, to be more informative.

[ABW]: [Fix] Workaround for inconsistency in AH data after reindex (data was pushed directly as irreversible but state thinks some blocks are still reversible)
[Fix] Block api no longer gives blocks that were not processed into state even if longer block_log is available.
[Fix] hived compiles again with SMTs enabled.
[Ign] Cleaning old marked history data now follows the same pattern as everything else (with object reference instead of iterator copy - just in case).
Supplemented comment on how we can workaround problem of steem recovery account.

[ABW]: [Fix] Workaround for inconsistency in AH data after reindex (data was pushed directly as irreversible but state thinks some blocks are still reversible)
[Fix] Block api no longer gives blocks that were not processed into state even if longer block_log is available.
[Fix] hived compiles again with SMTs enabled.
[Ign] Cleaning old marked history data now follows the same pattern as everything else (with object reference instead of iterator copy - just in case).
Supplemented comment on how we can workaround problem of steem recovery account.

- Changed the STEEM_CHAIN_ID definition name to OLD_CHAIN_ID

WIP: created specific implementation of transaction controller to be just owned by given thread (and working concurrently in dedicated DB connection)

- Removed HIVE_CHAIN_ID from the defines parsed in the get_config
  generation CMake file

Runs on Ubuntu 18.04 LTS