Fixed

• Fix startup regressions caused by PR #1949. Instead of checking all types by OID, we can return types for well known types PR #2257
• Backport PR #2148 Avoid leaking server error details through BatchUpdateException when logServerErrorDetail PR #2254
• Backpatch PR #2247 QueryExecutorImpl.receiveFastpathResult did not properly handle ParameterStatus messages. This in turn caused failures for some LargeObjectManager operations. Closes Issue #2237 Fixed by adding the missing code path, based on the existing handling in processResults. PR #2253
• Backpatch PR #2242 PgDatabaseMetaData.getIndexInfo() cast operands to smallint PR#2253 It is possible to break method PgDatabaseMetaData.getIndexInfo() by adding certain custom operators. This PR fixes it.
• Backpatching PR #2251 into 42.2 Clean up open connections to fix test failures on omni and appveyor use older syntax for COMMENT ON FUNCTION with explicit no-arg parameter parentheses as it is required on server versions before v10. Handle cleanup of connection creation in StatementTest, handle cleanup of privileged connection in DatabaseMetaDataTest
• Backpatch PR #2245 fixes case where duplicate tables are returned if there are duplicate descriptions oids are not guaranteed to be unique in the catalog PR #2248
• Change to updatable result set to use correctly primary or unique keys PR #2228 fixes issues introduced in PR #2199 closes Issue #2196
• Fix NPE calling getTypeInfo when alias is null PR #2220
• Backpatch PR #2217 to fix Issue #2215. OIDs are unsigned integers and were not being handled correctly when they exceeded the size of signed integers

Notable changes

Changed

• renewed the SSL keys for testing ### Added

Fixed

• getColumnPrecision for Numeric when scale and precision not specified now returns 0 instead of 131089 fixes: Issue #2188
• Calling refreshRow on an updateable resultset made the row readOnly. Fixes Issue #2193
• results should be updateable if there is a unique index available PR#2199 Fixes Issue #2196
• Rework sql type gathering to use OID instead of typname. This does not have the issue of name shadowing / qual-names, and has the added benefit of fixing #1948.

Notable changes

Changed

Added

Fixed

• Regression caused by https://github.com/pgjdbc/pgjdbc/commit/4fa2d5bc1ed8c0086a3a197fc1c28f7173d53cac. Unfortunately due to the blocking nature of the driver and issues with seeing if there is a byte available on a blocking stream when it is encrypted this introduces unacceptable delays in returning from peek(). At this time there is no simple solution to this.