Jerome Haltom

So the biggest difference seems to be the System.identityHashCode implementation. In one of the passes to rely more fully on OpenJDK, it was changed back to it's default implementation, which...

Oh. No idea. Why don't you profile it?

So I'm pretty sure this is us allowing this code to run: https://github.com/ikvmnet/jdk8u/blob/3f2b6ffee4b2a077f5fc8e00d58bce906a80cca0/jdk/src/windows/native/sun/windows/ShellFolder2.cpp#L243C3-L265C1 Trying to double initialize COM on a .NET thread which already has it initialized.

So this turns out to be pretty hard to solve. .NET will initialize ALL managed threads to MTA by default. There is no way to switch on the fly. There...

Looks like another option is available: not creating Java threads using System.Threading.Thread. Doing that will initialize CoInit them. However it looks like .NET does not CoInit them if they are...

I can't manage to reproduce this. You're going to need to provide some sort of reproduction of some kind.

> See also the OP in [dotnet/msbuild#2480](https://github.com/dotnet/msbuild/issues/2480) Would be interesting to investigate if there is some possibility of infering this scope, and thus simply being able to release the memory...

> [@wasabii](https://github.com/wasabii) I think that long-term that will need to be the solution (and future inference/tracking to make Target-invocation parallelism will likely _require_ this), but I want to start with...

Late to the conversation also, but to me, they should have been namespace scoped. I would have liked to use them for a use case I came up with: a...

So, here's what I notice. The path to the file is to the staging directory. That's within a single project. So, Project/obj/Release/net9.0. The IKVM build system assumes, like every part...