Standardize pg_regress usage

Standard PGXS `make installcheck` usage depends on a existing postgres
server. Tests will run in a contrib_regression database created by
pg_regress.

The pg_regress option --use-existing makes pg_regress require an
existing contrib_regression database, but the Makefile does not create
it. Removing that option makes using the testsuite easier.

Similarly, the init.sql file was not loaded by the Makefile, but it is
required to run the tests. Move it into tests/sql/ and name it
00_init.sql so it gets run first.

With these changes, the testsuite runs without any further tweaking
during the Debian package builds for apt.postgresql.org.

Signed-off-by: Christoph Berg <myon@debian.org>

Author: df7cb

Makefile

@@ -34,7 +34,7 @@ EXTRA_CLEAN = sql/$(EXTENSION)--$(EXTVERSION).sql $(EXTENSION).control
 
 TESTS = $(wildcard test/sql/*.sql)
 REGRESS = $(patsubst test/sql/%.sql,%,$(TESTS))
-REGRESS_OPTS = --use-existing --inputdir=test
+REGRESS_OPTS = --inputdir=test
 
 MODULE_big = $(EXTENSION)
 SRC = $(wildcard $(SRC_DIR)/*.c)

test/expected/00_init.out

@@ -0,0 +1,25 @@
+CREATE TABLE projects
+( id integer
+, name text
+, project_name text
+, client_id integer
+, subclient_id int
+);
+-- ensure these dropped column cases are tested
+ALTER TABLE projects DROP COLUMN project_name;
+ALTER TABLE projects DROP COLUMN subclient_id;
+INSERT INTO projects VALUES (1, 'Windows 7', 1);
+INSERT INTO projects VALUES (2, 'has,comma', 1);
+INSERT INTO projects VALUES (NULL, NULL, NULL);
+INSERT INTO projects VALUES (4, 'OSX', 2);
+INSERT INTO projects VALUES (NULL, 'has"quote', NULL);
+INSERT INTO projects VALUES (5, 'has,comma and "quote"', 7);
+INSERT INTO projects VALUES (6, E'has \n LF', 7);
+INSERT INTO projects VALUES (7, E'has \r CR', 8);
+INSERT INTO projects VALUES (8, E'has \r\n CRLF"', 8);
+create extension if not exists pg_csv;
+CREATE TABLE nasty (
+  "unusual"",names" INTEGER GENERATED ALWAYS AS IDENTITY,
+  text TEXT
+);
+INSERT INTO nasty (text) VALUES ('test');