anchor/docs/src/tutorials/tutorial-5.md

Errors

If you've ever programmed on a blockchain, you've probably been frustrated by either non existant or opaque error codes. Anchor attempts to address this by providing the #[error] attribute, which can be used to create typed Errors with descriptive messages that automatically propagate to the client.

Defining a Program

For example,

use anchor_lang::prelude::*;

#[program]
mod errors {
    use super::*;
    pub fn hello(_ctx: Context<Hello>) -> Result<()> {
        Err(ErrorCode::Hello.into())
    }
}

#[derive(Accounts)]
pub struct Hello {}

#[error]
pub enum ErrorCode {
    #[msg("This is an error message clients will automatically display")]
    Hello,
}

Observe the #[error] attribute on the ErrorCode enum. This macro generates two types: an Error and a Result, both of which can be used when returning from your program.

To use the Error, you can simply use the user defined ErrorCode with Rust's From trait. If you're unfamiliar with From, no worries. Just know that you need to either call .into() when using your ErrorCode. Or use Rust's ? operator, when returning an error. Both of these will automatically convert into the correct Error.

::: details What's the deal with this From stuff? Well, because the Solana runtime expects a ProgramError in the return value. The framework needs to wrap the user defined error code into a ProgramError::Code variant, before returning. The alternative would be to use the ProgramError directly. :::

Using the Client

When using the client, we get the error message.

try {
  const tx = await program.rpc.hello();
  assert.ok(false);
} catch (err) {
  const errMsg = "This is an error message clients will automatically display";
  assert.equal(err.toString(), errMsg);
}

It's that easy. :)

To run the full example, go here.