Sysinfo improvements

[GuillaumeGomez]

Keeping a hard-set list of supported targets might not be the best, it'd force the code to be updated every time sysinfo supports a new target (which doesn't happen often but I guess it's better to consider it?). For example, sysinfo supports FreeBSD but it wasn't listed in the supported targets. So instead, just listing known problematic targets in the cfgs seems like a better strategy. But maybe it's not what you prefer? In any case, very curious to hear more about it.

For the rest, I simply updated sysinfo version since it got a lot of fixes (mostly Windows).

[GuillaumeGomez]

The miri failure is quite impressive. O.o

[mockersf]

The miri failure should be fixed in tomorrow nightly by rust-lang/rust#113946

[GuillaumeGomez]

I'm realizing that the dynamic_linking feature is used the wrong way: cargo features are supposed to be additive. Having the opposite (like no_dynamic_linking) would allow to write:

[features]
no_dynamic_linking = ["sysinfo"]
default-features = ["no_dynamic_linking"]

[dependencies]
sysinfo = { version = "0.29.6", optional = true }

And then in the code we could simply check with cfg(feature = "no_dynamic_linking"). Simplifying the cfg conditions even further.

[mockersf]

In other part of the engine dynamic_linking is additive, it's only an issue here because sysinfo fails with it.

As far as I know It's not possible to have a feature expose globally (dynamic_linking) that would enable no_dynamic_linking in a dependency when it's not present

[mockersf]

isn't this also needed for macOS? from sysinfo docs it seems it is https://docs.rs/sysinfo/latest/sysinfo/#use-in-binaries-running-inside-the-macos-or-ios-sandboxstores

[GuillaumeGomez]

Only if running it in a VM. Talk about a nightmare...

[GuillaumeGomez]

Closing as I don't think this is going anywhere for the time being. I'll send a PR once the new sysinfo release will be done.