Sentinel Errors

[jhernandezb]

Summary of Bug

sdk errors package allows new modules to register their own sentinel errors through sdkerrors.Register(ModuleName, Code, Message) but documentation around how to use it is inconsistent since all sdk modules start with code 2 without explaining why it should.

I found a while back that if you use code 1 for a sentinel error while building a module there is nothing that prevents from doing it, but if the module ever returns such error when you submit a tx that will trigger it the raw_log message its replaced, because its a reserved code for ABCI(?)

Example:
{"height":"104","txhash":"B4D2A9C80ADB2C6147FBD9336423A67CD65E9D072B511DCF085D3EAC9BD28FAC","codespace":"sdkerrordemo","code":1,"data":"","raw_log":"internal","logs":[],"info":"","gas_wanted":"200000","gas_used":"44695","tx":null,"timestamp":""}

Either this should have better docs or sdkerrors.Register should prevent it (breaking change)

Documentation
https://docs.cosmos.network/v0.42/building-modules/errors.html#registration

Modules using CodeError 1
https://github.com/tendermint/liquidity/blob/develop/x/liquidity/types/errors.go#L9
https://github.com/desmos-labs/desmos/blob/master/x/profiles/types/error.go#L10 (other modules from same codebase)

Version

v0.42+

Steps to Reproduce

• create a sentinel error with code 1
• submit a tx that triggers this error
• query the tx and see the message
  NOTE: This doesn't happens during tests since the expected error from the msg handlers its the one expected.

---

For Admin Use

• Not duplicate issue
• Appropriate labels applied
• Appropriate contributors tagged
• Contributor assigned/self-assigned

[alexanderbez]

@jhernandezb what exactly are you looking for concretely here? The docs state 1 is reserved and that you cannot use it.

[jhernandezb]

not really sure if thats been there but I missed it but the code snippet is wrong and like me I think that's what people might see first as an example see the the links of modules I posted.

Probably this is the wrong repo for docs issues

[alexanderbez]

Good catch, the code snippet should be fixed. But the use of not allowing 1 has been there for a while.

/cc @alessio