build: use windows-{core,bindgen} 0.59, windows-targets 0.53

[ErichDonGubler]

• Depends on build: allow windows-core 0.59, windows-targets 0.53 nagisa/rust_libloading#169.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 86.28%. Comparing base (63f322f) to head (8b708e3).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #65   +/-   ##
=======================================
  Coverage   86.28%   86.28%           
=======================================
  Files           5        5           
  Lines         554      554           
  Branches      554      554           
=======================================
  Hits          478      478           
  Misses         48       48           
  Partials       28       28           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ErichDonGubler]

Not sure what the Linux-only nightly debug failure is, but it's something to do with LSAN. 😬 I don't think I have the bandwidth to figure it out right this moment, though.

[ErichDonGubler]

This PR only upgrades windows-* dependencies, though @mxinden had invited me to make them a range dependency instead. The reason I didn't is that while things seem source-compatible between versions of windows-core releases, windows-bindgen's output supposedly changes significantly between the 0.58 and 0.59 versions. I wasn't sure if it was safe to use a range in that case.

It seems that bindgen is/was used to avoid bringing in the windows crate, but I think that should actually be fine to use now in mozilla-central. Was there another reason not to use the windows crate directly? If not, we might be able to use a range over the windows crate itself.

[mxinden]

Not sure what the Linux-only nightly debug failure is, but it's something to do with LSAN. 😬 I don't think I have the bandwidth to figure it out right this moment, though.

Will be fixed with #69.

[mxinden]

It seems that bindgen is/was used to avoid bringing in the windows crate, but I think that should actually be fine to use now in mozilla-central. Was there another reason not to use the windows crate directly? If not, we might be able to use a range over the windows crate itself.

@larseggert can you explain why bindgen is used?

[mxinden]

For the record: nagisa/rust_libloading#169

[larseggert]

In earlier versions, we tried to avoid pulling in the other largerwindows crates by having a static Rust binding file for the few Windows APIs we called, together with a CI test that called bindgen to detect any changes.

We abandoned that at some point, and converted to calling bindgen during build. I don't think there was any reason other than it was a simply change.

If we can avoid bindgen by just using the other windows crates, let's do so, esp. if it makes integration easier.

[larseggert]

@ErichDonGubler I don't seem to have the ability to rebase your PR; did you maybe not give owners the relevant permission?

[ErichDonGubler]

@larseggert: Huh, that's strange, I'm not sure why you can't rebase. I don't see the Allow maintainers to edit this PR checkbox, but that's normally what I'd check. 🤔

I've rebased using the GH UI.

[larseggert]

@ErichDonGubler Thanks for rebasing. Would you want to switch us from the bindgen dep to the other windows deps in this PR?

[ErichDonGubler]

@larseggert: Sure!