Fix DB connection leak

[martonarbocz]

https://kaligo.slack.com/archives/C01LK7XLGE9/p1704682961528359
We discovered that using Sequel.connect to return the connection actually leaks it. Let's use a block instead.

• sqlite depends on pkg-config, so I added that to bin/setup

[Vuta]

CI is failing, can you help fix it

[martonarbocz]

CI is failing, can you help fix it

Yeah I'm looking at it, quite weird because originally tests passed for me with the system Ruby, but they fail when 3.1.2.

[hieuk09]

Overall LGTM, could you add a test?

[fill32]

how do we carry out the test? trigger the worker and see if there's any idle connection?

[martonarbocz]

I wanted to just do expect ... not_change db.pool.size, but it seems that db connections stay constant during the test run, even when I deliberately created extra connections by not using a block 🤔

[martonarbocz]

Maybe this is because of the simpler DB setup (just using SQLite)? Hieu, do you have any advice how to carry out the test reliably?

[Vuta]

@martonarbocz you can try assert the value of Sequel::DATABASES before and after migrator.migrate is called

[martonarbocz]

@martonarbocz you can try assert the value of Sequel::DATABASES before and after migrator.migrate is called

Thank you sir, that works 💯

[hieuk09]

We don't need to create the table because the migrator will create the table if it does not exist.