Do not run elapsed slots loop in tests #2411
Conversation
| #[cfg(not(feature = "dev-context-only-utils"))] | ||
| let run_loop = true; | ||
|
|
||
| if self.epoch.saturating_sub(account_rent_epoch) > 2u64.saturating_pow(30) { |
There was a problem hiding this comment.
2^30 still executes quickly. 2^35 takes more than a minute.
| #[cfg(feature = "dev-context-only-utils")] | ||
| let mut run_loop = true; | ||
| #[cfg(not(feature = "dev-context-only-utils"))] | ||
| let run_loop = true; |
There was a problem hiding this comment.
ok, so non-test cases (ie. normal validator), we will always run the loop.
Does the fuzzer run the validator (or sub-pieces) with dev-context-only-utils?
|
My preference is to have the test work around the production code, instead of vice versa. Is is possible to reduce the range for the epochs in the fuzzer? |
Yes, I can follow this approach. |
|
I'll follow @brooksprumo's advice to limit the range of rent epochs in the fuzzer. |
Problem
A loop that calculates the number of elapsed slots between two epochs depend on the difference between the start and the end epoch:
agave/sdk/src/rent_collector.rs
Line 93 in bb70670
This loop can get quite slow (>20 minutes) if the difference between them is too large (around 2^40). Although this case may not happen on the network, it slows down fuzzing efforts.
Summary of Changes
I added a condition to avoid running the loop when the number of iterations is too high.