Hi, so I managed to implement a simple midi controller yesterday, using usbd-midi and atmega-usbd. I'm not new to Rust and have done some arduino embedded before, but this was the first time I joined the two. Frankly, I'm still a newbie at embedded in general. So I have a few questions if you don't mind.

• My device is working when connected to windoze, but I'd like to send a prototype to a mac/ipad user. Have you tested this lib connected to mac/ipad? Ignore the midi part, I just mean the USB side. My theory is that it should just work(tm). I'm sending them a prototype board, so I'll be able to report back on this in a week or so.
• The example code you've provided is much more verbose than something like https://github.com/btrepp/usbd-midi (STM32). As a suggestion to you, is it possible to move much of this interrupt stuff into your library such that the user side implementation is easier? I did try to move the polling out of the interrupt and into my program loop, but that caused windoze to fail to recognise the device, which I found odd given the btrepp STM32 example works fine in the main thread.
• What is the purpose of the PLL section? I tested my code with and without and it works either way.

I realise this library is still fledgeling. Thank you for implementing it. You saved me a big headache!

The example link in the README.me points to an archived git repository. That archived git repo points back to this repository.

Please update the README.md so there is a less confusing link.

In the process of tracking down why I can't send/receive packets via other endpoints, I've discovered that writes to EPNUM don't stick sometimes. If you read the value of EPNUM immediately after writing to it and compare to the value you wrote, it might not be the same.

I've been tracing the code and it seems to work fine for a while and then stops working at some point. Some data points:

• It works just fine while configuring endpoints near the start of the program, in enable() (which is called when the UsbDevice is constructed, finalizing endpoint allocations).

• It stops working sometime before the USB reset signal (i.e. before the first EORSTI interrupt), e.g. you may notice it fail in poll when the endpoint interrupt flags are being checked for pending data.

• Polling the endpoint interrupt flags does work at least once before failing

• When it fails, if you read the value of EPNUM before and after the write, the before/after values are always equal, and they might always be zero (I may have encountered a situation where they were non zero but I don't quite remember)

Basically i have usbd-serial device in the main loop, and i want to use serial device to log errors in the panic handler.

There are two solutions for this:

1. Make a global serial singleton.
2. Reset the device somehow.

First one should just work, but i want to implement second one, and have encountered problems with this.

I have tried setting a static mut var in the .noinit section, similar to how lufa jumps to bootloader, to keep that variable between reboots, then have used watchdog to reset the program, and then on the start of the program checked if that variable is set do the actual panic handling, but even that doesn't work, for some unknown reason, i have to wait until PC powercycles the devices itself.

I have tried .reset() method on the bus, but that doesn't work.

I have found .force_reset() method that is currently unimplemented and i think that is what i need probably, so what i ask is for its implementation.

The reports for the two "L" keypresses are identical, which supposedly makes the host PC think that the "L" key has been held down.

Possible solutions:

• Send an empty report in between identical characters
• (Possibly simpler) Send an empty report between all characters

Their default configuration allows them to function as the USB data pins, but it should still consume them to make sure they are configured correctly.