Refactor #30

[MabezDev]

Based on the comments from @nickray and my own thoughts:

• Remove stm32l47x feature, replaced by a more generic stm32l4x6, due to regressions, see cannot find type CRRCR in module rcc #32
• hsi48 will always be available in Clocks, but for stm32l4x6 devices the method to set it is not available
• remove duplicate freezes, opting to instead conditionally compile the specific code
• remove usb_rng, Clocks is used to detect if a peripheral has the right clocks or not* see rng.rs.

@mathk this will introduce some breaking changes for you, but nothing too severe I hope.

* The clock selection still needs to be implemented properly.

[nickray]

Looks like an improvement to me, thanks!

Maybe remove hsi48 completely for devices where it does not exist? And get rid of the || in checking if RNG clock is enabled? https://gist.github.com/nickray/e660f293c26ad34207aa86d51e9b578e

[MabezDev]

And get rid of the || in checking if RNG clock is enabled?

The || will still be required as AFAIK all l4 devices have the MSI, and all can use the msi as the 48mhz rng/usb clock.

Maybe remove hsi48 completely for devices where it does not exist

This can be done but then requires a bit more awkward cfg feature gating.

How about this instead? Doesn't look quite as weird.

let msi = match clocks.msi() {
    Some(msi) => msi == crate::rcc::MsiFreq::RANGE48M,
    None => false,
};
let hsi = clocks.hsi48();
assert!(msi || hsi);

[nickray]

Fine :) I was mistakenly under the impression that RNG and USB both needed HSI48 on the L432.

Not sure what happens if a device without HSI48 configures it to be turned on, plus it's wasting space on those devices, but I guess we can revisit minimalism when we're more feature complete, and the stm32-rs has come to a consensus on how best to organize multi-device HALs anyway.

[mathk]

@MabezDev Ok I see the stm32-rs/stm32-rs#141 has not been merge yet. If merged #32 should disappear. Do you still want to erfactor that ? Becasue the CRRCR is available on some 4x6 chip. I have double check all reference manual only the 47x does not have the CRRCR.

[mathk]

I still think that is interesting to have the usb_rng flags. as it is an independent clock line.

[MabezDev]

Do you still want to erfactor that ? Becasue the CRRCR is available on some 4x6 chip. I have double check all reference manual only the 47x does not have the CRRCR

At this point I time I think it's best to feature gate specific chips. e.g 476 475 471 instead of 47x. I also think it would be best to accept feature gate PR's as a full package where all/most peripherals are features gated for the device.

I'll try to feature gate a chip at some point as a reference PR for adding more devices.

Hopefully we can tick all the items of of this #29 not too far in the future :)

[MabezDev]

Fine :) I was mistakenly under the impression that RNG and USB both needed HSI48 on the L432.

Not sure what happens if a device without HSI48 configures it to be turned on, plus it's wasting space on those devices, but I guess we can revisit minimalism when we're more feature complete, and the stm32-rs has come to a consensus on how best to organize multi-device HALs anyway.

It shouldn't be able to turn it on once feature gated properly, the hsi(bool) is hidden. Not sure the byte for hsi matters to much but you are right there is probably a more efficient way to do this.