write() on a socket with the other end closed throws a SIGPIPE.
Let's ignore this signal. write() will return an error anyway...

fullread has been changed, so get_event too, and the callers also
for the moment the callers crash, enb.c will be a bit more clever,
the others to see...

The application (enb.c basically) will now monitor the socket
and reconnect if it dies. See following commits.

It was wrong.
Plus it is starting work so as to make enb.c not crashing if
the tracee goes away (maybe also textlog and vcd, but those are of
less importance).

The unit simulators may still need some checks.

(the T macro is then undefined)

if a read from the remote tracer's socket fails, let's smash
the socket and wait for another remote tracer

it's buggy. Next commit to fix it.

The problem is if the tracer exits while sending a message.
The local tracer will not get it full and will throw garbage
to the T in the tracee, leading to some bad things...

plus some cleanup here and there

get rid of forward.c, make everything static inside local.c

remove forward stuff from defs.h
remove defs.h from local.c and forward.c

The local tracer now waits for the remote tracer to connect,
that's no more the other way around.

All the remote tracers (textlog, enb, vcd) have also been converted.

first step: let the tracee connect before the remote tracer
and let's have a "no wait" mode where the local tracer starts
the tracee before the remote tracer is connected

get rid of T_send(), no use anymore

The T only works when it's TRUE, the version without shared memory
is not supported anymore.

can be deactivated on the command line or in the gui

- adding T macros here and there
- modify the T_messages.txt
- modify enb.c to trace them in the RRC view

plus add rrc connection setup complete tracing

as special events.

If we store all received events to a file, this dump is now
self-contained. Helps to analyze traces.

Some functions have also been factorized in the process, too much
duplication.

(Also tracer/Makefile was forgotten in previous commit, included
in this one...)

many assumptions about X backed were made, to be refined if does not
work properly in some setups