Newer hash computation method is injected into the db with older hfm
version. Both naked new HAF db and old one use the same algorithm to
getting hash.

Now there is no need to extend list of hashed tables each time when
new table is added to schema hafd.

WARNING: previously people forget to  extend the list and there
are few tables which were not hashed. Because of this update from
previous version of HAF is impossible

There is no sense to pass schema parameter to hash db computation,
at the end there is a list of tables to take into computation and all of
them belongs to hafd schema.

 - hive: develop (a181ebfab21471951a369b138adb2f5003d7a642)

- repair an abandoned test for update and start it on CI
- fix for syntrax error during state provider update

 - hive: develop (015c4c0dc6b76d8b256363335609c5eadbcfaaeb)

This reverts commit e9fa365a

 - hive: develop (ef0e30e473fcc02ec5ca7f926fe1e3ca8301e02a)

Marcin authored 3 months ago and Dan Notestein committed 3 months ago

Previously, an application waited in app_next_iteration indefinitely for a moment when HAF reached the 'LIVE' stage what makes impossible to break with SIGINT and holds long-standing transaction. Now when HAF is before 'LIVE' then  app_next_iteration returns null block range after 500ms. Because no blocks range is returned application is forced to back ask again for blocks.

 - hive: develop (8d5912e639040fb8ec1452163d6bf166fd939d84)

Adjust functional tests, corrected genesis accounts, pow operation sets weight threshold to 1 instead of 0

 - hive: develop (bec4daffde0fb5768af80748679b884493c465de)

If the definitions of start providers functions are the same as in the old hfm version, then
it is possible to make update of state providers runtime function. The updatable runtime code needs to
placed in state providers functions hive.runtimecode_provider_<provider_name>, all elements which are created there shall be placed in schema hive and will be removed during update procedure and recreated with their new versions

Hash of start_provider functions bodies is computed and included into whole database hash.
It means that when the definitions of state providers tables are changed, then update is not possible.

indexes dissapeared because encoding functions were removed during
update

- all functional tests are running with hfm update between each step, this way we got all hfm feature tested including  checking how update procedure affects them

- fix for update system test which did not test update procedure because postgres did not start update when new version number  was the same as the old

 - hive: develop (5c7a9f728cb5454f16849900b50598a404e04cde)

Add functions for apps to request/wait for full vacuums and allow for concurrent index creation during live sync