Update linux-raw-sys to 0.6

[oherrala]

This required changing fields of some structs under io_uring thus probably braking semantic versioning.

[sunfishcode]

Yes, this is unfortunately semver breaking. If there's a specific feature you need in linux-raw-sys 0.6, we can backport or add compatible definitions it to 0.4.

[raftario]

I'm personally interested in this for the futex support in io_uring

[sunfishcode]

0.38.x releases are now being done in a dedicated 0.38 branch, so main can now accept semver-breaking changes.

The CI failure looks like rustix's io_uring_sync_cancel_reg needs to be updated to match the changes in Linux.

[raftario]

@oherrala Do you plan on updating this to fix CI ? If not I'd be happy to take over

[sunfishcode]

The CI failure here looks due to changes in the layout of xdp_umem_reg, so it looks like rustix's XdpUmemReg needs to be updated to match.

[sunfishcode]

The compile errors are fixed on main in #1173. However the linux-i686 failure here looks relevant:

thread 'net::types::test_sizes' panicked at src/net/types.rs:1817:5:
assertion `left == right` failed
  left: 32
 right: 28

[arctic-alpaca]

However the linux-i686 failure here looks relevant:

This is caused by the libc crate still using the previous/removed version of the struct:

• https://github.com/rust-lang/libc/blob/main/src/unix/linux_like/linux/gnu/mod.rs#L396-L402
• https://github.com/rust-lang/libc/blob/main/src/unix/linux_like/linux/musl/mod.rs#L308-L314

[arctic-alpaca]

Starting with libc version 0.2.165, the struct is updated and this PR shouldn't be blocked on this anymore.

[oherrala]

Rebased my changes on top of ef39477. Let's see if the tests pass now.

[cgwalters]

We hit on the listxattr call exposing c_char and hence I think this PR would roll in sunfishcode/linux-raw-sys@769c988#diff-b1a35a68f14e696205874893c07fd24fdb88882b47c23cc0e0c80a30c7d53759R20 which would fix that

[sunfishcode]

Thanks!