…)` when that function is called by `col_isnt_pk()`. Very weird.
* Eliminated trailing white space.
* Refactored the above to cut down on redundant code.
* Added `hasnt_pk()` and `hasnt_fk()`. * Added missing `DROP` statements to `uninstall_pgtap.sql.in`. * Refactored the code for checking for the presence or absence of tables and columns so that it's much simpler.
…last commit, but I didn't.
* Fixed an error in the test count on 8.0 in `sql/todotap.sql`. * Changed the `Makefile` so that options to `pg_regress` are passed with an equals sign, as required by pg_regress 8.1 and earlier. * Added a cast from `oidvector` to `regtype` on 8.0.
* Converted `sql/unique.sql` to use `check_test()`. * Changed `sql/todotap.sql` to use `ok()` to compare arrays, instead of using `is()` with the undocumented `textin()` and `array_out()` functions. We lose the better diagnostics, but as long as the test pass, it should be fine, and we can switch back if the test ever starts to fail again.
* Converted `sql/istap.sql` to use `check_test()`.
* Fixed a couple of `$Id` keywords. * Fixed a typo in the README.
…n. So use one we wrote ourselves.
…s well as the error code. * Moved testing of multiline description from `sql/throwtap.sql` to `sql/moretap.sql`, where it belongs. * Switched to `check_test()` in `sql/throwtap.sql`. * Now testing for the exception message for an invalid call to `todo_end()` in `t/todotap.sql`. * Changed the documentation and comments for `throws_ok()` and `lives_ok()` to use the word "query" for their first augments, rather than "text" or "sql".