We don't use board0 as a valid identifier.
Our mapping starts with 1 and goes up to
the maximum number of slots, which is 12
in the general case.

This was obsolete and was killing all dev_io/dev_mngr
instances

This script was used before systemd and killed
all of the dev_io dev_mngr instances. Now we let
systemd take care of this.

We now only pass the cfg_file as argument and
not the broker endpoint. The later is read from
the configuration file, among other options.

This allow systemd to take care of spawning different
dev_io services independently whenever the corresponding
device is added/removed from the system bus.

This is needed as we use a custom mapping between
board/bpm and the bpm_id used to spawn EPICS IOC
processes.

Now, we don't map our hash with <pattern>/afe or
<pattern>/dbe, just <pattern> is enough and we
split different configuration options in hash
structure fields.

Now, we don't specify the configuration filename
through the Makefile, but in the command-line of
dev_mngr.

This option is only valid if daemonize parameter
is "yes".

This is not needed anymore as we use system
services

This allows us to selective spawn the broker.
Useful in cases that the broker is an independent
system service, as Malamute.

This is related to commits a0f3f992 and 9d80fdf5

We were not multiplying the sample size by the
number of samples to give the number of total bytes
per block.

blink_leds and ad9510_cfg_defaults were not using
our simple messaging protocol, which causes some
confusion for the client programmer, as well as
for libbpmclient.

The timeout variable must support negative numbers, -1,
meaning an infinite wait.

Now, we can stop the acquisition FSM.
This is useful if we are acquiring data in
triggered mode and we want to abort the
acquisition.

For now, we reset the FPGA on PCIe open.