Before we were not calling the proper close() function in the destructor
of the TrtlDevice object.

We were also also not checking that the device handle (token) was
defined before using it.

Last but not least, we were not wrapping the device handle as a Python
Ctypes 'c_void_p', which caused Ctypes to convert it to a 32-bit int,
pointing to nowhere in a 64-bit system.

Otherwise our FECs might complain

Resolve "makefile for firmware does not work with gcc version 11.2"

Closes #33

See merge request be-cem-edl/common/mockturtle!28

Otherwise we might get multiple matches from grep if more than one
device is programmed with the same FPGA bitstream.

This provides an option to pytest to look at a particular folder (e.g.
`/lib/firmware`) for all required test firmware.

see commit diff for comment with explanation

Resolve "Package firmware for distribution"

Closes #36

See merge request be-cem-edl/common/mockturtle!27

Resolve "Migrate CI software build and publishing to new common pipeline"

Closes #35

See merge request be-cem-edl/common/mockturtle!26

Resolve "Conflict between multiple sync users on same HMQ"

Closes #34

See merge request be-cem-edl/common/mockturtle!25

This modification makes sure that all device handles (to the same MT
devid) are opened at the start of the test and kept open for the
duration of the test. This allows catching race conditions between the
read/write pointers in the HMQ buffers (because the read pointer is
initialised to the value of the write pointer when the device handle is
first acquired, so the sooner we open it, the more such errors we will
detect).

Resolve "Fix broken pytest CI for releases"

Closes #32

See merge request be-cem-edl/common/mockturtle!24

Resolve "Make v4.2.0 release"

Closes #30

See merge request be-cem-edl/common/mockturtle!23

This makes the "code" more readable, target-specific variables can be
confused as additional targets (they are not).

This new target will create two compressed tarballs, one containing the
kernel modules and another one containing software libraries and tools.

Also update gitlab CI to use this new target to pulbish the software
and use it during testing on real hardware.

Resolve "SW reset signals for mock-turtle CPUs"

Closes #31

See merge request be-cem-edl/common/mockturtle!22

There is one reset signal per CPU. Their behavior is the same as the reset bit in the control registers (it resets the corresponding CPU while leaving the RAM intact).

Resolve "Document external RMQ interface"

Closes #28

See merge request be-cem-edl/common/mockturtle!21

Resolve "Run pytests in CI"

Closes #29

See merge request be-cem-edl/common/mockturtle!20

Resolve "Remove driver code to handle sequence numbers"

Closes #27 and #24

See merge request be-cem-edl/common/mockturtle!19