conflicts_with_all ist accepts unacceptable entry and runtime panics

[thingsiplay]

I start a discussion, because I don't know if this is a bug or not. At the moment, I am in the process of upgrading some older dependencies in my CLI app. One of them is clap v3.2 to v4.5 (using derive). While doing so I noticed that with option conflicts including a - dash in the list will happily be accepted by clap. But that is clearly wrong and needs to be an _ underscore. Rust compiles this and running the app will crash with a panic.

I asked this in the Rust Discord and got the answer this is not a responsibility of Rust compiler itself, but maybe a bug in clap. Therefore before making an Issue, I would like to ask you guys if this is a bug or not.

I have two options config and open-config, which are mutual exclusive via the configuration:

conflicts_with_all = &["config", "open-config", "core"]

In clap 3 this worked fine, but with clap 4 I get following error message at runtime:

Argument or group 'open-config' specified in 'conflicts_with*' for 'noconfig' does not exist

This is solved by replacing the dash with the underscore:

conflicts_with_all = &["config", "open_config", "core"]

So either Rust or clap made a mistake to accept the unacceptable list entry with the dash. What do you guys think?

[epage]

This is not a bug.

The derive does not have all of the information it needs to properly error for this at compile time because a derive can only process the current struct and not any #[command(flatten)]ed structs. We could conditionalize the error based on whether flatten is present but that can create a false impression with users and requires duplicating a lot of code.

See also

• Duplicate short option (and other misconfigurations) gives runtime error, wouldn't it be cool, if it was compile time error #3133
• ArgAction::Count type error not caught at compile time #3864
• Clap panics on positional bools in debug builds #4467