Now, each SMIO will deal with the traditional
Majordomo socket and the additional PIPE socket
for monitoring messages. For now, every message
is interpreted as a self-destruct one.

This fixed #79 github issue

For invalid temporary channels, such as TBT position and FOFB
position, the start address was not aligned to 8 or 16 bit, causing
a segmentation fault on reading from these address.

This fixes #80 github issue.

For invalid temporary channels, such as TBT position and FOFB
position, the start address was not aligned to 8 or 16 bit, causing
a segmentation fault on reading from these address.

This fixes #80 github issue.

The inclusion of the dependencies was deleted in a previous commit.
Now, it was readded.

Now, we don't explicitly depend on the PCIe driver for the
compilation of the dev_io and dev_mngr. However, the driver should,
of course, be loaded before starting any dev_io or dev_mngr
instances.

If the libmdp Makefile is not found, nothing is cleaned.

This target was added, as the dev_io would compile the
libclient and use it internally. Then, if the user also
wanted to install the libclient, the Makefile would not
regenerate the libclient objects, causing a possible
unwanted reuse of objects from a previous compilation.

Now, if the user wants to generate a clean libclient,
it should use make mrproper && make <desired flags> && make install.

Previously, we were using $?, meaning "all the prerequisites that
are newer than the target", which is wrong as the library is composed
of all of the objects specified in the prerequisites. Now, it was
changed to $^, meaning "all the prerequisites".

This fixes #78 github issue.

Now, the default frequency is 113.040445 MHz.

These functions were missing and are important for the PLL
lock monitoring and for selecting the reference clock input
(FMC front panel or FMC pin)

This fixes #77 github issue.

Now, the RW message parameter is used to select between
reading or writing the specified parameter.

Conflicts:
	foreign/libmdp

The correct default value is 0 (disabled).

It was added a a PCIe timeout detection, so we can try to
detect a invalid stream of bytes and reset the FPGA timeout counter.
On detecting the timeout, we just sleep some period and hope that
there is no other contention inside the FPGA firmware.

This is a first fix to github issue #73. The "correct" fix is
described there.

Now the function only returns valid bytes to the client.

This fixes #45 github issue.