this is now following what Twincat does, as checked with tcpdump

not done for FastEtherCat yet, so it doesn't work

somehow the anybus thing wants to return a EoE frame when
asked for a CoE frame. We can just ignore it, then it works
fine.

some terminals do not support reading 8 bytes of the eeprom at
a time, we have to limit to 4 bytes at a time

this involves restructuring quite a bit, probably with compatibility
problems.

this is not actually ethercat specific

the format characters already include the information whether they
are signed or long, no need to add this information again.

this means we know what values were set in ebpf. It is a bit
magic how often this will get called.

all kinds of stuff updated:

- EEPROMs can now be written, especially the serial number
- tests should work again
- ebpf devices can now act cyclic like slow devices

See merge request !7

even when using fast ethercat, some information may be perfectly
valid slow as well. This way we have the best of both worlds.

and use it instead of our own random number generator for the
missed packages simulator

this tests now the actual initialization of the terminals.

I removed the ebpf tests for now, as I am about to change them anyways.

don't use RuntimeError, define an EtherCatError instead

those didn't work anymore since we have mmap

this script reads out all relevant data from the connected terminals
such that it can be used for offline tests. Also add the retrieved data.

if the status of the bus is unknown, we need to assure not
to address two terminals at the same time

This adds support to actually use the SDO data from terminals, and also add useful command line tools to inspect the bus.

See merge request !6

Beckhoff calls the input sync manager number 3, the output one
number 2. It's weird, but let's follow the standard.

bus couplers seem to be incomplete

that's what I found in the EEE lab and Brad's box