Hello All,
I am working on a project that aims to produce test cases that improve mutation coverage of a dbms's test suite.
The rough workflow of the project goes as follows:
(a) apply mutation at a souce code level
(b) compile and check if the mutated installation passed existing testsuite
(c) If not, fuzz the mutated installation with SQL fuzzer
(d) if a fuzzor successfully produce a test case that crash or trigger bugs in the mutated installation, use a reduction tools to reduce the test case
(e) add the reduced test case to existing test suite
For postgres, I am looking at adding test cases to test suite in test/src/regress/. I have gone through (a)-(e), and managed to produced some test cases. As an example, I claim the test case
```
CREATE RECURSIVE VIEW a(b) AS SELECT'' ;
SELECT FROM a WHERE NULL;
```
could kill the following mutation at optimizer/plan/setrefs.c, 502:5--502:33
Original binary operator expression:
```
rte->rtekind == RTE_SUBQUERY
````
Replacement expression:
```
(rte->rtekind) >= RTE_SUBQUERY
```
I have a few questions about adding these test cases:
(a) The regression test suite is run by a parallel scheduler, with some test cases dependent on previous test cases. If I just add my test case as part of the parallel scheduler’s tests, it might not work, since previous test cases in the scheduler might already create the same table, for instance.
(b) How do I get my test cases reviewed and ultimately included in a future release of PostgreSQL?
Thank you for your time.
Regards,
Jon
