panic=abort compatible test harness

[alexcrichton]

Right now developing a panic=abort project can be relatively painful when running tests. A cargo build will compile everything with panic=abort but a cargo test will recompile everything with panic=unwind. This is currently due to the fact that libtest requires panic=unwind to correctly operate. Otherwise it can't catch test failures!

We may wish to explore a panic=abort compatible test harness to remove Cargo's special casing. Some possible ways to implement this may be:

• Instead of starting a thread for all tests we could instead start a process. This process could be a re-execution of the executable itself and we'd just do that once per test. The downside of this is that it may be super slow.
• We could continue to start a thread for each test but override the panic handler for the process. This panic handler would do "test related things" and provide an opportunity for a "clean exit". The downside of this is that you wouldn't get a set of all tests that failed, just the first few (maybe)

I think that we'll also need support in the compiler for this, we'd have to inject different test frameworks or backends depending on the -C panic=... configured mode.

[Mark-Simulacrum]

I believe this is #32512, closing in favor of that. I agree that this would be good, though I think it's more of a feature than an enhancement (but it's a fine line that doesn't matter that much).

[alexcrichton]

Aha indeed!