Support sendmmsg/recvmmsg #1208
Conversation
|
Could you please fix the test failure before I review this PR? |
src/sys/socket/mod.rs
Outdated
| Errno::result(ret).map(|r| r as usize) | ||
| } | ||
|
|
||
| #[cfg(any(target_os = "linux", target_os = "freebsd"))] |
There was a problem hiding this comment.
It looks like sendmmsg is also supported on Android, Dragonfly, NetBSD, and OpenBSD.
There was a problem hiding this comment.
Thanks, I will fix it. I didn't find the support on Dragonfly.
|
Looks like you have some test failures to fix. |
glebpom
force-pushed
the
mmsg
branch
2 times, most recently
from
b8373d2
to
77421b1
Compare
src/sys/socket/mod.rs
Outdated
| { | ||
| let iter = data.into_iter(); | ||
|
|
||
| let (min_size, max_size) = iter.size_hint(); |
There was a problem hiding this comment.
How about adding a restriction to force IntoIterator::IntoIter always be an ExactSizeIterator? Additionally, use set_len right after vector initialization instead of MaybeUninit magic. It will simplify the whole logic of Vectors initializations, and prevent possible allocations. @asomers
There was a problem hiding this comment.
I think that's too great of a restriction. I don't know if I've seen a single Rust API that requires an ExactSizeIterator.
|
Oh, and please don't force-push until the end. It makes review harder. |
|
Looking a lot better. But it needs a more detailed review than my current level of sleepiness will allow. I'll set a reminder to do it tomorrow. Feel free to ping me if you don't hear anything for a few days. |
|
@glebpom I understand now the need for ExactSizeIterator in recvvmsg, and it seems reasonable. But I made a change to remove a mem::zeroed. Could you please review it? |
|
Build succeeded: |
… for sendmmsg() / recvmmsg() in nix-rust#1208. In nix-rust#1208, sendmmsg() / recvmmsg() were added, but OpenBSD(who doesn't support these) was included on the list of allowed operating systems for sendmmsg() related things. This broke the build on OpenBSD. For more Rust-world examples, see: rust-lang/libc@6f62973
1252: Add EV_DISPATCH, EV_RECEIPT items for EventFlags, and fix build on OpenBSD r=asomers a=nickpelone This pull request: - Adds `EV_DISPATCH` and `EV_RECEIPT` to `EventFlags`. OpenBSD supports these now. c379d1a - Fixes a regression that caused `nix` to be unable to build on OpenBSD since #1208. dd0a990 - Makes a change to avoid a deprecation warning from `libc` crate, which breaks tests. 0f9fcbd Since I know y'all don't have OpenBSD in CI right now, I've attached the results of a `cargo test` run on OpenBSD 6.7. Thanks for your time! [testresults.txt](https://github.com/nix-rust/nix/files/4684597/testresults.txt) Fixes #1251 Co-authored-by: Nick Pelone <nick.pelone@calyptix.com>
1252: Add EV_DISPATCH, EV_RECEIPT items for EventFlags, and fix build on OpenBSD r=asomers a=nickpelone This pull request: - Adds `EV_DISPATCH` and `EV_RECEIPT` to `EventFlags`. OpenBSD supports these now. c379d1a - Fixes a regression that caused `nix` to be unable to build on OpenBSD since #1208. dd0a990 - Makes a change to avoid a deprecation warning from `libc` crate, which breaks tests. 0f9fcbd Since I know y'all don't have OpenBSD in CI right now, I've attached the results of a `cargo test` run on OpenBSD 6.7. Thanks for your time! [testresults.txt](https://github.com/nix-rust/nix/files/4684597/testresults.txt) Fixes #1251 Co-authored-by: Nick Pelone <nick.pelone@calyptix.com>
This pull request adds
sendmmsgandrecvmmsgfunctions. They are used to send and receive multiple messages per one call. Additionally,AsMutimplemented for bothTimeSpecandTimeVal.This is still WIP, because we need to find the best way on how to pass multiple values to *mmsg functions. Current approach with iterators doesn't seem ergonomic.