2022년 7월 31일 (일) 오전 6:32, Tom Lane <t...@sss.pgh.pa.us>님이 작성: > > Dong Wook Lee <sh95...@gmail.com> writes: > > I just wrote test about pg_rowlocks extension. > > I added sql and spec test for locking state. > > I think this could be cut down quite a bit. Do we really need > both a SQL test and an isolation test? Seems like you could > easily do everything in the isolation test.
I agree with your optionion. > Also, it is not a good idea to go creating superusers in a contrib > test: we support "make installcheck" for these tests, but people don't > especially like new superusers cropping up in their installations. > I doubt that we need *any* of the permissions-ish tests that you > propose adding here; those are not part of the module's own > functionality, and we don't generally have similar tests in other > contrib modules. I agree it's right to remove that part. > If you do keep any of it, remember to drop the roles you create --- > leaving global objects behind is not OK. (For one thing, it > breaks doing repeat "make installcheck"s.) > > Another thing that's bad style is the "drop table if exists". > This should be running in an empty database, and if somehow it's > not, destroying pre-existing objects would be pretty unfriendly. > Better to fail at the CREATE. Thank you for the good explanation. It will be very helpful to write a test in the future. > See also my comments about your pg_buffercache patch, which > largely apply here too. OK. I will add the `.gitignore` file. I will revise my patch and submit it again as soon as possible.
