Respect verify partial doubles config when verifying job/mailer arguments

[odlp]

Following on from #2745, this PR skips the job/mailer verification that arguments match the signature when:

• RSpec Mocks is configured with verify_partial_doubles false
• The examples/expectations are wrapped with #without_partial_double_verification

This direction was suggested by @JonRowe here:

I'm leaving this open because I think this should respect the existing configuration for verifying doubles, we should check to see if we should verify arguments.

Closes #2801.

[JonRowe]

Rather than creating a new helper, I would rather the specs be setup with an around block that restores the original behaviour, and explicitly setting the values to true/false in a before in the context, I don't mind if you make a helper for the "restore" part (in an around hook) but I would strongly prefer it if the contexts set the correct values directly in a before.

[odlp]

Rather than creating a new helper, I would rather the specs be setup with an around block that restores the original behaviour, and explicitly setting the values to true/false in a before in the context, I don't mind if you make a helper for the "restore" part (in an around hook) but I would strongly prefer it if the contexts set the correct values directly in a before.

👍 Updated to match the requested style

FWIW I find a block-based helper superior when temporarily mutating global state because that interface guarantees the state is reset (like Timecop safe mode for example).

I've worked in many codebases which mutate global state in before hooks and don't have a corresponding after or around hook to reset said state, leading to order-dependent test failures. Appreciate this codebase is safeguarded with good code review with high standards... but to err is human and so on...

[JonRowe]

FWIW I find a block-based helper superior when temporarily mutating global state because that interface guarantees the state is reset (like Timecop safe mode for example).

I've worked in many codebases which mutate global state in before hooks and don't have a corresponding after or around hook to reset said state, leading to order-dependent test failures. Appreciate this codebase is safeguarded with good code review with high standards... but to err is human and so on...

An around hook is essentially a block based helper, the problem I had with it is the implementation obscured the actual config setting that was being set, as our specs act as part of our documentation its less obvious whats being tweaked, an alternative would be to use the without_partial_double_verification block helper thats actually part of rspec-mocks if you prefer.

[odlp]

@JonRowe

An around hook is essentially a block based helper

Very true.

an alternative would be to use the without_partial_double_verification block helper

I do use this in the other specs - but it sets a distinct flag Mocks.configuration.temporarily_suppress_partial_double_verification rather than Mocks.configuration.verify_partial_doubles. It's a pity in a way there are two flags at play here, but that's why we need to check both flags and test accordingly.

Overall I think the PR is fine now; I'm happy to adopt the house style and appreciate the trade-offs you mention between obscuring the config setting vs safety.

[JonRowe]

Thank you for this!

[JonRowe]

Released in 7.1.0