[Bug] RetryBackoffLayer can trigger a div-by-zero if CU misconfigured

[mothran]

Component

transports

What version of Alloy are you on?

0.6.4

Operating System

Linux

Describe the bug

When running the RetryBackOffLayer with a low (lower than AVG_COST) it is possible trigger a div-by-zero in the layer:

thread 'main' panicked at /home/user/.cargo/registry/src/index.crates.io-6f17d22bba15001f/alloy-transport-0.6.4/src/layers/retry.rs:230:53:
attempt to divide by zero

With the configured params:

        RetryBackoffLayer::new(10, 1000, 10);

Having a sanity check for the configured CU, or bounding that div would be nice to avoid the panic.

[mattsse]

makes sense,

do you want to provide a fix for this?

[mothran]

makes sense,

do you want to provide a fix for this?

I can take a crack at it (maybe next week), is there a example of unit testing tower layers in the codebase I can use to work off?

[yash-atreya]

makes sense,
do you want to provide a fix for this?

I can take a crack at it (maybe next week), is there a example of unit testing tower layers in the codebase I can use to work off?

@mothran ptal here

alloy/crates/provider/src/provider/trait.rs

Line 1432 in 28ad3f7

async fn test_custom_retry_policy() {

[Evalir]

What's the fix we want to do here? bound the div to always be at least one? Or just handle the case where CU < AVG_COST separately?