GitLab

Currently, libgda uses template1 default database to connect to before
a new database can be created. If more than one connection is
established to the template1 database a new database can't be created.

'postgres' database should be used to connect to before any other
database connection is established.

See relevant discussion here:
https://www.postgresql.org/list/pgsql-sql/2020-10

Message-ID: 390804abe88e4e9da9049f6e05c841783acaca47.camel@gmail.com

This bug affects the following issues:
GNOME/libgda#194
GNOME/libgda#221

Use uniqe DB for the test

See merge request GNOME/libgda!158

We don't need to worry about dropping the database in the test

This object is a convenience for returning information
during the creation of the database.

test_random_string() API was added and implemented in the separate file
to avoid repetition. meson.build file was updated to reflect this change.

To avoid conflict between tests we need to use a unique database name
every time the test starts.

Some minor corrections to improve code quality were also added.

This change will use two identical build/test
processes to stress multi-threaded/simultaneos access
to dabases

_gda_data_select_update_usage_flags() was used on one place
and its function was moved to gda-server-provider.c

Step required for #222

Fixing wrong assignment from GdaDataMetaWrapperMode to
GdaSqlIdentifierStyle

Explicitly pass type of operation to create a GdaServerOperation
instance.

1. Use gda_top_src/gda_top_build to be suitable
to use as a meson subproject

2. If no POSTGRESQL_CNC_PARAMS enviromental variable
is set, skip test

A simple correction in the example

See merge request GNOME/libgda!154

SQLite: Adding support for bigint type

See merge request GNOME/libgda!153

Add Documentation examples

See merge request GNOME/libgda!151

MySQL: Reimplementing CREATE_DB and DROP_DB

See merge request GNOME/libgda!149

They were moved to a separate test unit.