Potential bug in getLogs

[benesjan]

Hello,
I am fetching logs and I get a log from an older block than fromBlock param. I am testing the code against Anvil. Is this a bug or is this for whatever reason an expected behavior? Thank you

[jxom]

Interesting! We definitely do not modify the block number sent over RPC. Could be worth raising this as an issue in Foundry.

https://github.com/foundry-rs/foundry

[fubhy]

Anvil has caching for fork requests. Maybe the issue is there?

[fubhy]

Are you using a fork? Can you try to run anvil with the --no-storage-caching option?

[benesjan]

I am not using a fork and the --no-storage-caching param is not available without fork.

When I place the log in rpc.js I get the same incorrect output and when I copy the request body from the output and run it with curl I get the correct result. For this reason I would say it has to be some issue on the request side and not in Anvil. I tried explicitly setting no-cache in fetch call but it didn't help. It's really weird. Will try to do the reproducible example.

[benesjan]

I tried logging all http traffic with sudo tcpflow -p -c -i lo0 port 9000 (started anvil on port 9000) and then when I pasted the exact request to curl again I got the incorrectly cached response directly from anvil. I guess I didn't catch this before because maybe the json was slightly different or something. Will talk to Foundry devs and close this if it will actually turn out to be anvil issue.

[benesjan]

I didn't catch the anvil issue before because it seems that when I don't kill the test immediately after the bug occurs the test does further communication with anvil and it makes the bug not occur again. So I managed to only reproduce the bug with curl when I carefully watch the logs and kill the test fast.

[benesjan]

It was a bug in Anvil as per this comment.