Added unit test and fixed another occurances of the same issue. Unit test was failing before, so it definitely proves that there was an issue.

Thanks in advance for merging.

By reverting the changes to the library paths and keeping the tests, I cannot get this to fail on my own MBP with macOS Sequoia 15.3 (24D60) with either .NET 8 or .NET 9 SDKs. 🤔

I'm not an expert in macOS so I don't know why you can't reproduce it locally, but it definitely did fail in CI:

• https://github.com/SteamRE/SteamKit/actions/runs/13103559845/job/36554805608
• https://github.com/SteamRE/SteamKit/actions/runs/13103538944/job/36554764315

(Those were the other two issues that I discovered accidentally after writing unit test for verification of the first one)

Yeah, I saw that. I still have Questions.

Will have a think about this - seems harmless atm but not sure what the wider impact is exactly.

ah, got it.

This only seems to affect .NET 9.0.1 / SDK 9.0.102 or higher. Iiiiiinteresting.

If we have global.json for 8.0.100 latestMinor then why is dlopen on GitHub Actions looking in /Users/runner/.dotnet/shared/Microsoft.NETCore.App/9.0.1?

https://github.com/SteamRE/SteamKit/blob/master/SteamKit2/Tests/Tests.csproj#L5

sigh.

ok I think I've wrapped up the loose ends here. and I'm not exactly happy with where we are but I think a hotfix makes sense.

rollforward was specified in csprojs before global.json existed
ad531d9