deps: Move from winapi to windows

[poliorcetics]

winapi is in maintenance mode and the new blessed way to access Windows APIs are the windows
and windows-sys crates. I was forced to use windows because sysinfo uses the COM APIs
which are not present in windows-sys.

[poliorcetics]

Currently blocked by microsoft/windows-rs#2370 because ULARGE_INTEGER is not available yet and I didn't want to make it myself here. I could though ?

Also, it should be noted ntapi still uses winapi behind the scene. I plan to work on that but it will be much harder than this since ntapi exposes structures with members that are marked as Reserved in windows-rs.

There is microsoft/windows-rs#2358 (comment) which could probably help, and it's now possible to build crates above windows-rs using the metadata and generator, I'll investigate that too

[riverar]

Can you point to what's blocking you?

ULARGE_INTEGER => u64
LARGE_INTEGER => i64

[GuillaumeGomez]

Currently blocked by microsoft/windows-rs#2370 because ULARGE_INTEGER is not available yet and I didn't want to make it myself here. I could though ?

I'd rather wait for the alias to be available in the dependency rather than adding it in sysinfo.

Overall, it seems to go into the right direction, but windows doesn't seem ready yet.

[riverar]

@GuillaumeGomez Don't think we have any plans to add that alias. It's not needed as the PR demonstrates?

[GuillaumeGomez]

If it's used by the windows API, then it's needed, even if it's just an alias. Just like libc provides aliases for int64_t and other sized types, it's not because "we can't do without them" but rather to make it much easier for maintainers to compare between their code and the API.

[poliorcetics]

error: package windows v0.44.0 cannot be built because it requires rustc 1.64 or newer, while the currently active rustc version is 1.59.0

Ah, that's sad.

[GuillaumeGomez]

We can update the minimum required version. It'll just require us to change sysinfo major version number.

[poliorcetics]

We can update the minimum required version. It'll just require us to change sysinfo major version number.

What do you prefer ? Doing it in this MR (in new commits) or in a separate one ?

[GuillaumeGomez]

In this one, otherwise CI will always fail. Just put it into its own commit.

[kennykerr]

error: package windows v0.44.0 cannot be built because it requires rustc 1.64 or newer, while the currently active rustc version is 1.59.0

Ah, that's sad.

The latest version of the windows crate supports Rust 1.48.

https://kennykerr.ca/rust-getting-started/windows-or-windows-sys.html

[poliorcetics]

running 27 tests
test common::tests::check_display_impl_mac_address ... ok
test common::tests::check_display_impl_process_status ... ok
test common::tests::check_mac_address_is_unspecified_false ... ok
test common::tests::check_mac_address_is_unspecified_true ... ok
test system::tests::check_hostname_has_no_nuls ... ok
test common::tests::check_uid_gid_from_impls ... ok
test system::tests::check_if_send_and_sync ... ok
test system::tests::test_consecutive_cpu_usage_update ... ignored
test system::tests::test_get_process ... ok
test system::tests::test_refresh_process ... ok
test system::tests::test_refresh_system ... ok
test system::tests::check_uptime ... ok
error: test failed, to rerun pass `--lib`

Caused by:
  process didn't exit successfully: `D:\a\sysinfo\sysinfo\target\debug\deps\sysinfo-3e0191f40579cb26.exe` (exit code: 0xc0000005, STATUS_ACCESS_VIOLATION)
note: test exited abnormally; to see the full output pass --nocapture to the harness.

I'm not managing to reproduce locally, makes searching for the bug much more annoying than it should be

[GuillaumeGomez]

Try to update CI script for windows to run with RUST_BACKTRACE=full and with --test-threads 1 maybe?

[poliorcetics]

Rebased and green CI, should be good to go 🎉

[GuillaumeGomez]

Apart from the few code style nits, it looks great, thanks for working on this!

[poliorcetics]

All fixed :)

[GuillaumeGomez]

Looks all good to me, thanks a lot!