Move kqueue-related definitions from nix to libc #415
Conversation
Change 11aa1f34243d5bbb7d6327a6607bd9d2530f3954 to libc added kqueue-related definitions. They are more accurate and more complete than nix's own definitions. Use them where possible. Also, rationalize Nix's definitions so its public API will be as similar as possible across all OSes.
|
Our master branch requires libc's github master branch. So you don't need to change our Cargo.toml. |
The final version of libc's PR nix-rust#379 removed a few definitions, and fixed OpenBSD's definition of fn kevent.
| nevents: size_t, | ||
| timeout: *const timespec) -> c_int; | ||
| } | ||
| // Redefine kevent in terms of programmer-friendly enums and bitfields. |
There was a problem hiding this comment.
I think it would be preferable here to just have an inner libc::kevent struct. Then we don't have to duplicate the definition, and can get rid of the test that they match. This would require some setters though, for the enum / flags parts. I guess I should look at some kqueue code to see how it's used first. Never done any.
There was a problem hiding this comment.
I tried that first. But I found it very inconvenient to require setters in my application code. It's much more convenient just to let ke = KEvent{whatever}
There was a problem hiding this comment.
The rest of nix just wraps libc's structs and provides constructor fuctions:
pub struct NixStruct {
struct: libc::struct
}
impl NixStruct {
pub fn new(all the fields as parameters) {
// do the dirty convertion here
}
}I like it and would like to keep it consistent.
There was a problem hiding this comment.
One issue here is that there are 6 args. I think it works ok for things with 2 maybe 3. With 6 it becomes hard to know what they are from reading.
There was a problem hiding this comment.
@fiveop what about testing fields? It's easy to do
let k = KEvent{...}
...
match k.filter {
nix.event.EVFILT_AIO => something
_ => something else
}
but that becomes harder if the filter field is no longer an enum.
There was a problem hiding this comment.
With an accessor you still have an enum. See for example EpollEvent::events().
There was a problem hiding this comment.
@fiveop that's only because EpollEvent and EpollEventKind are both defined within Nix. But kevent is defined within libc, and libc doesn't do enums, only constants. That's why I redefined KEvent within nix.
There was a problem hiding this comment.
I'm sorry. I forgot that my change to EpollEvent is not merged yet. You can see, how it can look like at https://github.com/nix-rust/nix/pull/410/files.
There was a problem hiding this comment.
Ok @fiveop I think I see what you want. I'll give it a shot.
|
@fiveop is this more like what you had in mind? |
There was a problem hiding this comment.
Yes this looks good. There are some things that do not follow our conventions yet. Since they are not introduced by this PR, I would suggest fixing them in a separate PR.
Could you also update the CHANGELOG.md with respect to the changes made here? I think we only need to mention, that KEvent is now opaque with respect to its internal structure and field access has been replaced by public read only accessors (so far) and a constructor function.
Thank you for all the work so far.
| type type_of_event_flag = u16; | ||
| #[cfg(any(target_os = "macos", | ||
| target_os = "freebsd", | ||
| target_os = "dragonfly"))] | ||
| bitflags!( |
There was a problem hiding this comment.
I think you can now use the libc_bitflags macro, that allows to reduce repetition. You can find examples elsewhere in nix.
Don't use it for FilterFlag, because it triggers recursion limit reached error
|
☔ The latest upstream changes (presumably bf00bf2) made this pull request unmergeable. Please resolve the merge conflicts. |
|
All three build failures seem unrelated to this PR. |
|
📌 Commit 3ccb313 has been approved by |
Move kqueue-related definitions from nix to libc See rust-lang/libc#379 for the companion commit to libc. If and when that commit is merged, I'll update nix's Cargo.toml file and add it to this PR.
|
💔 Test failed - status |
|
The failure is due to the gcc dependency not working with Rust 1.2. It is unrelated to this PR. |
|
I would have prefered to have waited until we get the build fully working again. A rebase instead of a merge from master would have also been nicer. |
|
Sorry about that. |
See rust-lang/libc#379 for the companion commit to libc. If and when that commit is merged, I'll update nix's Cargo.toml file and add it to this PR.