[Bug] CLI flags do not take precedence over env var DBT_DEFER_STATE

[Siete-F]

Is this a new bug in dbt-core?

• I believe this is a new bug in dbt-core
• I have searched the existing issues, and I could not find an existing issue for this bug

Current Behavior

When specifying the env var DBT_DEFER_STATE=stateprod, it appears to be considered even when you try to overrule the state manually using dbt test --defer --state state (non-prod)

Do I miss something regarding

Expected Behavior

Expected to be able to overwrite the state using the flags, as stated in the docs: https://docs.getdbt.com/reference/node-selection/syntax#establishing-state

Steps To Reproduce

The following example results in stateprod being used as state.

When doing the same with DBT_STATE (and unsetting DBT_DEFER_STATE), it works as expected, and the flag gets precedence. i.e. the used state is 'state'

Relevant log output

NA

Environment

- OS: Windows 11
- Python: 3.10.5
- dbt: 1.8.7

Which database adapter are you using with dbt?

spark

Additional Context

I noticed that the on-run-end DOES use the expected state (provided via the cli flags) when having DBT_DEFER_STATE set! Causing inconsistent catalogs being referred in my case.

[dbeatty10]

Thanks for reaching out @Siete-F !

Did you try the --defer-state flag? That's the one that overrides the DBT_DEFER_STATE environment variable:

More info / examples in the docs here and here.

[Siete-F]

Hi @dbeatty10 ,
Thanks for your reply!
no, I didn't but I did expect the --defer and --state to overwrite also --defer-state when it has been set. It felt like one of those options was probably legacy, but maybe I am mistaken?

The second doc's link you shared I didn't check yet, so thanks for sharing! State seems to be used in two different contexts of which I am unfamiliar with the state:modified+ select syntax. But I dont understand (from a short read) from the docs how --defer-state differs from the combination --defer and --state, so that's why I would expect the latter two (as cli commands) to also overwrite the former as an env var.

As a sidenote, the behavior that I am sure is undesired is that the on-run-end uses one state while the dbt call uses another. See the last sentence of the bug report.

(have to go)