Fix pg db cancel missing

[rweikusat]

This fixes a probably accidental editing error which cause the pg_db_send_cancel and pg_db_cancel to be "garbled together" while losing parts of the code (issue noticed by Ed Sabol).

It also changes the cancel path in handle_old_async to only do a rollback if there's actually a transaction running because I noticed a warning on make test and wanted to investigate that to ensure that it's harmless.

This doesn't address the other issue noted by Ed Sabol ("failed to create dbh ...") yet.

[turnstep]

Thanks. I confess I haven't had time to really sit down and sort all this out - and the garbled code was my fault I'm sure - so thanks for all your efforts here!

[rweikusat]

Well, I have the time as I'm unemployed since my employer collapsed in January. And I actually enjoy working with code. That's at least hopefully more useful than doing job applications which - at best - get boilerplate "HR deleted this from the system" rejections.

:-)

[turnstep]

Could you update the Changes file with all the new async stuff?

[rweikusat]

No problem.

But there's still the other error Ed Sabol uncovered (the "could not create dbh") and I'd like to get that resolved first. I expect that this will take a few days (starting Sunday or Monday) because ATM, I first need to find a way how I can reproduce that in an environment I can make changes to other than by creating dummy pull requests for that.

[esabol]

But there's still the other error Ed Sabol uncovered (the "could not create dbh") and I'd like to get that resolved first. I expect that this will take a few days (starting Sunday or Monday) because ATM, I first need to find a way how I can reproduce that in an environment I can make changes to other than by creating dummy pull requests for that.

Speaking from experience, GitHub workflows can be a painful to debug. Check out https://nektosact.com/ and https://github.com/nektos/act for one way to make it easier, I hope.

[rweikusat]

Thanks. I'll have a look at that.

[esabol]

@rweikusat : Looks like @oetiker fixed that failing test in 708542e in PR #150 ?

[rweikusat]

Sort-of. That's probably the reason for the failed test but it should just use $ENV{DBI_PASS} without the intermediate variables and useless '' assignment, as done in t/dbd_test_setup.pl (and everywhere else). Or rather, I should have written it this way from the start ... :-(

I'll create a pull request with that.