Mailer adapter does not behave as expected about isolator HOT 10 CLOSED

thanks a lot for your quick responses, I'm closing the ticket since both problems originated from my side

from isolator.

I believe the first problem is the mailer adapter doesn't recognize

What is delay? Sidekiq delay extensions or delayed_job? If you use delayed_job then it shouldn't raise, since delaying the execution means storing the record in the DB, and thus it only "survives" if the transaction commits.

my expectation is error shouldn't be thrown because send_second_email it's done after the transaction in after_commit

Correct. The bug is probably with the transaction detection in test env. Have you tried to run the create code in dev mode (i.e., in rails c)? Do you use transactional_tests or DatabaseCleaner?

from isolator.

What is delay?

It's sidekiq delay extensions

Have you tried to run the create code in dev mode (i.e., in rails c)?

Yes, things are working as expected in dev (aside from delay problem above)

Do you use transactional_tests or DatabaseCleaner?

DatabaseCleaner. I dug a little deeper and found that before the actual test portion, Isolator.transactions_threshold is 2 and Thread.current.fetch(:isolator_transactions, 0) is also 2. So it makes sense that send_second_email triggers an error. I believe this is a problem on my side, there's one more layer of transaction wrapped in my test setup somewhere.

from isolator.

Isolator.transactions_threshold is 2

Could you share your DatabaseCleaner configuration?

Which Rails version do you use, btw?

It's sidekiq delay extensions

We probably need some tests for this case first)

from isolator.

Could you share your DatabaseCleaner configuration?

RSpec.configure do |config|
  config.around do |example|
    # this is for ar_after_transaction: https://github.com/grosser/ar_after_transaction#transactional-fixtures---normally_open_transactions
    ActiveRecord::Base.normally_open_transactions = 0

    example.run
    DatabaseCleaner.clean_with(:truncation, pre_count: true)
  end
end

Which Rails version do you use, btw?

5.1.7 (ruby 2.6.5)

from isolator.

Could you try disabling ar_after_transaction? They do some AR monkey-patching, probably, incompatible with Isolator.

P.S. Btw, there is a better alternative to this gem which uses AR API with zero monkey-patching: https://github.com/Envek/after_commit_everywhere.

from isolator.

Could you try disabling ar_after_transaction?

Disabling it didn't help. But I workaround it by adding 1 more to Isolator.transactions_threshold in test environment. It's not a permanent solution but at least it allows me to run tests and identify the true violations.

Any word on supporting sidekiq delay extensions (the first problem)?

from isolator.

Any word on supporting sidekiq delay extensions (the first problem)?

Couldn't reproduce (see the test).
Could you please try to write a single file reproduction script (e.g., using a Rails bug report template)?

from isolator.

Which version of DatabaseCleaner do you use? Is it 1.8.0+ (it has a lot of internal changes)?

from isolator.

I figured out the mystery with sidekiq delay extension problem. We have an override that defers delay call to after transaction, so it makes sense that isolator doesn't throw error when enqueuing jobs with .delay

from isolator.