Introduced in #17, this is because generated tests look up function body by signature, and do so using the raw argument OIDs. While type OIDs are consistent for built-in primitives, this is not true for those provided by extensions, causing tests to only work on the generating machine.
Even worse, this does not produce a test failure, instead a plan failure is raised as no body with the given parameters is found, and thus no assertion is run.
To reproduce: Create a function with an hstore parameter and generate tests. Load this function on another machine and run the tests. Expected: success. Actual: fails.
Recommend disabling this functionality until fully baked. I'm willing and able to do a PR to make it actually work, but it may be a month or two before I have time.
Introduced in #17, this is because generated tests look up function body by signature, and do so using the raw argument OIDs. While type OIDs are consistent for built-in primitives, this is not true for those provided by extensions, causing tests to only work on the generating machine.
Even worse, this does not produce a test failure, instead a plan failure is raised as no body with the given parameters is found, and thus no assertion is run.
To reproduce: Create a function with an hstore parameter and generate tests. Load this function on another machine and run the tests. Expected: success. Actual: fails.
Recommend disabling this functionality until fully baked. I'm willing and able to do a PR to make it actually work, but it may be a month or two before I have time.