Do not retry actions from CLI upon an Assertion Failed

[spalladino]

If a user tries sending a tx that they know will fail, the retrying-fetch we have will still retry it 3 times (and log it those 3 times). We should detect an Assertion Failed error as a server error, and thus unretriable, and not keep retrying.

Requested on Discord:

QoL request re: makeFetch errors
when testing, I send a tx that I know will fail. it will print the error foundation:retry ERROR Error: Assertion failed: '...' three times via the makeFetch retry. This doesn't have any positive purpose IMO since the ts interface will throw an error for the dev to handle. It should probably also short-circuit the retry loop if the error is Assertion failed:
the fetch failed error SocketError: other side closed is common and clutters the cli too. Instead of logging this error, it should just throw if all retries fail and let the dev handle it
I know the api optionally takes DebugLogger but I am not providing it. I think the ideal situation would be to log retry errors only if the debug logger is provided.

[benesjan]

I just tried reproducing it by putting an assertion into private token's transfer function and when I run the test I get correct behavior:
In the sandbox log I get one log: aztec:simulator:acvm Error in oracle callback callPrivateFunction +25ms
And in the failed test output I get a nicely looking assertion error:

@spalladino do you by any chance have any more info regarding how to reproduce this?

[spalladino]

Not really, all I got is that quote from Discord that Mike had shared.

[spalladino]

Can you try reproducing it from the CLI instead of from a test?

[benesjan]

Same result using the CLI.

Log output in Sandbox:

And the cute assertion message in CLI output:

Makes me think whether it was not some old version of sandbox which didn't yet use the NoRetry errors.

[spalladino]

Maybe follow up with the guy who reported this on Discord, to see if they can still reproduce it with the latest version, or they can share the logs?

[benesjan]

Marked this as blocked until I get a response from the user on Discord. Here is my message inquiring about more details.

It's also quite likely that this has already been addressed by our work on not retrying unrecoverable errors.

[spalladino]

Closing as stale