Create list of high-value error messages to improve

[nicc]

We are intent on improving our error messages in general. There are some initivates that unblock this. In preparation for these improvement, we'd like a list of the most impactful error messages to focus on first.

Consider all errors (protocol, crypto, snarkyjs). Timeboxed to an estimate of 5. No more than 20 items.

Preemptive candidates:

• Out of memory
• Circuit size

Resources for when we are ready to improve the error messages:

• https://www.nngroup.com/articles/error-message-guidelines
• https://uxplanet.org/how-to-write-good-error-messages-858e4551cd4
• https://bootcamp.uxdesign.cc/7-steps-to-design-error-messages-49e509f03d18

[barriebyron]

@ymekuria shared https://github.com/MinaProtocol/docs/blob/main/pages/en/node-operators/transaction-failures.mdx that has been floating around, not published in the docs as of April 11, 2023

[nicc]

Another candidate:

Account_action_state_precondition_unsatisfied error during Reducer.reduce usually when you reduce the wrong set of actions (because you hash the wrong actions and the target hash doesn't match the committed on-chain hash anymore)

[nicc]

IRI: @nicc

[mitschabaude]

Here's a truly bad one: #808
If you forget to add tx.prove() when doing a zkApp call, the error is just:

Assert_failure src/lib/transaction_logic/zkapp_command_logic.ml:1266:17

This is because the authorizationKind is implicitly set to "proof" but then the actual authorization doesn't match that (missing proof).
(thrown from inside the transaction logic, called by LocalBlockchain)

[mitschabaude]

Basically, every error thrown from inside snarky is bad, because of the long and noisy stack trace. @mrmr1993 had some good ideas to address this.

Here's an example (cut off) for when a constraint fails inside the prover (for example, during tx.prove()):

constraint-unsatisfied.txt

[Kirol54]

I'm assuming it's coming from graphql but it's not a really helpful error message regardless

[rpanic]

I guess Can't evaluate prover code outside an as_prover block can be improved

[mitschabaude]

'Permutation failed' for when you create a different circuit every time because you use a constant instead of a variable

[nicc]

@nicc to review / create new issues

[shimkiv]

Another messages for consideration:

• https://github.com/o1-labs/e2e-tests/issues/56
• "Cannot start new transaction within another transaction" after an attempt to produce txn using unknown to the network account #425
• Better error handling of values pulled out from variables within the circuits #478
• Make PrivateKey.fromBase58 throw human readable error #601
• Handle assertions errors in JS with better error message #743
• Check limits before proceeding with the heavy computations #782
• Swallowed errors during comms with Archive-Node-API #841

cc @nicc

[nicc]

All improvements identified here have been reflected as new issues with the error-messages label. Closing this issue. Thank you, everyone!