Hi all!

I'm doing a little experiment (mainly as a learning experience) where I'm frankensteining together a way to run a piece of firmware directly on the host, deferring operations that can only be run on the MCU to probe-rs. Idea is to be able to shorten the feedback loop of embedded developers, by skipping the flash erase and program steps. Disclaimer: as mentioned, this is an experiment, to explore if this is even possible, and making things portable and beautiful is explicitly not on my TODO list.

Currently, I've got a simple PAC and `cortex_m::asm::Delay` based blinky application working, which runs on my linux pc, but has the LED of my nucleo board flashing. Pretty neat! The MCU runs a firmware that does basically this:

```rust
fn main() -> ! {
    loop {
        call_target();
    }
}

[..]

#[no_mangle]
#[used]
#[link_section = ".target_fn_data"]
static mut TARGET_FN: u32 = 0;

#[no_mangle]
#[used]
#[link_section = ".target_fn_data"]
static mut TARGET_FN_INPUT: [u8; 1024] = [0; 1024];

#[no_mangle]
#[used]
#[link_section = ".target_fn_data"]
static mut TARGET_FN_OUTPUT: [u8; 1024] = [0; 1024];

pub fn call_target() {
    let target_fn = unsafe { &mut TARGET_FN };
    if *target_fn != 0 {
        let f: fn() = unsafe { core::mem::transmute(*target_fn) };
        log::trace!("Calling target {:08x}", *target_fn);
        f();
        log::trace!("Done calling target {:08x}", *target_fn);
        *target_fn = 0;
    }
}

```


Next step of course is brining in the big guns, i.e. Embassy. My goal is to have the [embassy stm32L4 blinky example](https://github.com/embassy-rs/embassy/blob/main/examples/stm32l4/src/bin/blinky.rs) run on my host. Loads of stuff to figure out, but the thing I'm stuck on currently, is the step where `embassy::init` configures the RTC (I think?). Specifically, my application hangs on the following line: https://github.com/embassy-rs/embassy/blob/main/embassy-stm32/src/rcc/bd.rs#L69 

Now in my logs I can see that the call to `modify` on the line prior to this one is indeed executed, with a value that seems correct (bit 8 is indeed 1), but reading it never yields the value just written. 

Is the DAP doing something magical here that I'm not grasping?