Reactor Scram

The crash handler is supposed to get backtraces too, they end up in the log directory Ubuntu 20.04 (where we build the binaries) has glibc 2.31, and I thought glibc...

The headless client can do our auth scheme, the reason we have all these compatibility problems is the GUI for the settings window and the tray menu

Whatever it takes to avoid 2.x lol

@thomaseizinger Will do. There was a performance regression when I changed from `blocking_send` to `try_send` and since that was redundant with just closing the MPSC channel, I reverted to `blocking_send`....

Or I guess they just run faster overnight since they're VMs? this one completed in 13 minutes without any change to the MSI build https://github.com/firezone/firezone/actions/runs/9707332885/job/26792383572

In #5672 I'm adding a custom test runner for smoke-testing the GUI and IPC service at the same time. This test could be built in the same way or even...

But `cargo test` only compiles binaries with the libtest entry point, so I can't run normal code in there for the smoke tests, right? Unless I make a phony test...

I'd like to know whether the IPC service / GUI started near the same time Windows started or later on. That would tell us whether it likely auto-started (maybe intended...

I want to know whether the IPC service auto-started or not. If the logs show, e.g. that one instance did not shut down cleanly, and the next instance started up...

I agree that `windows` being 0.x still is really annoying. I actually see 7 different versions of it in my `cargo-tree` right now. `uptime_lib` is under 100 lines and MIT...