Fix flaky test_transaction_result_does_not_affect_bankhash

[apfitzge]

Problem

• test_transaction_result_does_not_affect_bankhash is flaky due to random keypairs/pubkeys
• The flakiness comes down to rent collection (rather updating rent epoch)
• In successful transactions, we collect rent during transaction processing for writable accounts
  • in this case, that will be the mint_keypair's pubkey
• In bank.freeze, we collect rent on partition(s)
  • in this case, that will be the mint_keypair IF the randomly generated key is in the partition
• IF the mint pubkey falls in the bank's rent collection partitions, the bankhash is unaffected since the rent is collected in both cases

Summary of Changes

• Use deterministic keys in the test, and in other tests

Fixes #

[apfitzge]

Hoping the change of keypairs/pubkeys in genesis config doens't break other tests, but we'll see. Killing randomness is good overall.

[apfitzge]

See @steviez recently saw this test failing as well: #3549

[steviez]

See @steviez recently saw this test failing as well: #3549

Do you have any buildkite links handy that failed on this test ? My PR that you mentioned included a bunch of logging that would help identify the offending account; that would help verify the theory you mentioned above (altho I guess I could also craft a Pubkey and re-run the unit test locally if I want to play around with it).

Either way - nice find and will get around to reviewing this shortly

[apfitzge]

See @steviez recently saw this test failing as well: #3549

Do you have any buildkite links handy that failed on this test ? My PR that you mentioned included a bunch of logging that would help identify the offending account; that would help verify the theory you mentioned above (altho I guess I could also craft a Pubkey and re-run the unit test locally if I want to play around with it).

Either way - nice find and will get around to reviewing this shortly

The output is pretty massive - I ended up adding a bunch of print statements and finding the issue that way while talking with brooks.

link: https://buildkite.com/organizations/anza/analytics/suites/agave/tests/c0ac6a06-88b4-8e28-834e-e232a122f763?branch=master#flaky

[brooksprumo]

I like removing the randomness. Almost feels like this could be two PRs: one for the changes to genesis utils, and another one for the changes to test_transaction_result_does_not_affect_bankhash.

[apfitzge]

I like removing the randomness. Almost feels like this could be two PRs: one for the changes to genesis utils, and another one for the changes to test_transaction_result_does_not_affect_bankhash.

The changes to genesis_utils change the result of the test **in most cases - or the equality at least. I'd rather just keep it all grouped as a single PR to keep it simple.

[brooksprumo]

[steviez]

LGTM

Did a scan through and looks like the changes in genesis_utils.rs are only affecting test code; can possibly throw some #[cfg(test)] in there in a separate PR

[mergify]

Backports to the beta branch are to be avoided unless absolutely necessary for fixing bugs, security issues, and perf regressions. Changes intended for backport should be structured such that a minimum effective diff can be committed separately from any refactoring, plumbing, cleanup, etc that are not strictly necessary to achieve the goal. Any of the latter should go only into master and ride the normal stabilization schedule. Exceptions include CI/metrics changes, CLI improvements and documentation updates on a case by case basis.