Joel Dice (Dec 06 2023 at 19:02): Joel Dice (Dec 06 2023 at 19:02):@Badeend I'm upgrading my wasi-libc fork to use the latest WASI RC. I see you removed TCP no-delay. What should we do if setsockopt is called with TCP_NODELAY? Just ignore it?
Joel Dice (Dec 06 2023 at 19:03):Or return an error if the value being set is zero?
Dave Bakker (badeend) (Dec 06 2023 at 19:03):Did you see the pr I've posted? I already updated it for you
Joel Dice (Dec 06 2023 at 19:04):oops, no
Joel Dice (Dec 06 2023 at 19:05):ugh, I didn't get an email about it for some reason. Thanks for the PR.
Joel Dice (Dec 06 2023 at 19:09):I think removing support for TCP_NODELAY is going to break one of my test cases. tokio-postgres sets it and will error out and refuse to connect if that fails. Could we accept nonzero values as a no-op and only return an error if someone tries to set it to zero?
Joel Dice (Dec 06 2023 at 19:41):Ah, I didn't get an email because I wasn't "watching" my own fork on GitHub :face_palm: Just fixed that
Notification Bot (Dec 06 2023 at 19:41):A message was moved here from #wasi > Container containing WASI-SDK and wasmtime? by Joel Dice.
Dave Bakker (badeend) (Dec 07 2023 at 09:36):Could we accept nonzero values as a no-op (...) ?
Yes :+1: , but not for the reason you've documented. :joy_cat:
I've created a follow-up PR. Hopefully that clears it up.
Joel Dice (Dec 07 2023 at 14:49):Thanks (as always) for the PR. Out of curiosity: what's the reason this is not yet supported in wasi-sockets? Are there portability concerns?
Dave Bakker (badeend) (Dec 07 2023 at 15:39):Well the TLDR would be: preview2 had to be shipped, and it wasn't clear yet (nor specified) how TCP_NODELAY would affect wasi:io input- and output streams
Last updated: Dec 23 2024 at 12:05 UTC