CLI helper: CreateMint

[grod220]

The first of a number of CLI/JS helpers for users of the program. First up is the CreateMint instruction.

Takes large inspiration from the CLI helpers and tests in spl-single-pool.

Walkthrough for testing locally:

1. Build program

cargo build-sbf # generates .so file
cargo build # builds binary

2. Start solana test validator

solana-test-validator \
  --bpf-program TwRapQCDhWkZRrDaHfZGuHxkZ91gHDRkyuzNqeU5MgR target/deploy/spl_token_wrap.so \
  --reset

3. Create a new mint & token account

spl-token create-token
# Example will have G85DDQCoHchhDFWtUFT2MDkfUPYQt8BvViFyjGv167xb for this
spl-token create-account G85DDQCoHchhDFWtUFT2MDkfUPYQt8BvViFyjGv167xb
spl-token mint G85DDQCoHchhDFWtUFT2MDkfUPYQt8BvViFyjGv167xb 100000

4. Run CreateMint token-wrap command

cargo run --bin spl-token-wrap create-mint \
    G85DDQCoHchhDFWtUFT2MDkfUPYQt8BvViFyjGv167xb \  # unwrapped_mint
    TokenkegQfeZyiNwAJbNbGKPFXCWuBvf9Ss623VQ5DA \    # unwrapped_token_program
    TokenzQdBNbLqP5VEhdkAS6EPFLC1PHnBqCXEpPxuEb     # wrapped_token_program

5. See expected output

Creating wrapped mint for G85DDQCoHchhDFWtUFT2MDkfUPYQt8BvViFyjGv167xb
Funding wrapped_mint_account 7KTYY4ZAsVG2Tn1V4119ZGRLWRPzo3r6nD2pde9rtLsV with 1461600 lamports for rent
Funding backpointer_account 7ohy6W6gxAHg42pAawrUPeivRLyCBPXALgnq17eNVYFF with 1113600 lamports for rent
Unwrapped mint address: G85DDQCoHchhDFWtUFT2MDkfUPYQt8BvViFyjGv167xb
Wrapped mint address: 7KTYY4ZAsVG2Tn1V4119ZGRLWRPzo3r6nD2pde9rtLsV
Wrapped backpointer address:: 7ohy6W6gxAHg42pAawrUPeivRLyCBPXALgnq17eNVYFF
Funded wrapped mint lamports: 1461600
Funded backpointer lamports: 1113600
Signature: 3T5yjnCgDwcPAafyYrjxfFXxDZfPFRFr2R8K4kcPbZnEukKjsYxiwJrdAJAFyCU7guq5Yye5QEwobLbsMkGZNKEe

[buffalojoec]

Really small stuff from me. Looks great!

[joncinque]

Looks great overall! Mostly small things

[buffalojoec]

Lgtm, just a CI failure to address!

[grod220]

Since swapping from ProgramClient to RpcClient in the tests I am hitting flakiness with a ~30% success rate on the cli test. I think it may be because the ProgramClient perhaps waits for confirmation and the other doesn't. Not sure, investigating.

[buffalojoec]

Nice! 😊

[joncinque]

When using rpc_client.send_transaction, it only sends.

So rather than setting all of the data at startup to "fix" the race condition, the preference should be to use rpc_client.send_and_confirm_transaction to wait for a transaction to land before sending the next one, if the two transactions depend on each other.

You don't need to change setting the accounts into the validator, but can you at least use send_and_confirm_transaction in the tests? That also lets us find out if the transaction was successful or not.

Also, it's done now, but the workspace deps changes should have been done in a separate PR.

[grod220]

the preference should be to use rpc_client.send_and_confirm_transaction to wait for a transaction to land before sending the next one, if the two transactions depend on each other.

When digging into the ProgramClient, I did manage to find send_and_confirm_transaction which fixed the issue. However, if all accounts are loaded on startup, that method isn't needed anymore and there is only the single command now:

 Command::new(TOKEN_WRAP_CLI_BIN).args(["create-mint", ...

Though it's still not clear to me what strategy is preferred for CLI tests. Is it better to A) load all accounts within the validator on startup and make the single command you are testing or B) have a fresh startup state and make the sequential transactions to build the state the command your testing needs. If A, there shouldn't be additional changes needed. If B, I can revert back and make use of send_and_confirm_transaction to fix that race condition.

[joncinque]

I typically lean towards building things up in tests with transactions rather than setting up the accounts by hand, since it's possible for the setup to be incorrect, as we've already seen in some of the program tests earlier, but there's no need to revert the work here.

And right, sorry, I thought there was still some usage of send_transaction somewhere. This is all good to me then!

[grod220]

Helpful, in the next PR will probably re-establish the multiple-tx pattern again.