num_buffers controls the number of packets that are intercepted and
stored by tansiv-client until the next deadline. A too small value will
drop extraneous packets (silently unless you know how to activate the
tansiv logs).

This patch allow its parameterization at
- qemu level (passed in the global vsg option)
- boot.py level

https://fr.wikipedia.org/wiki/Mars_Climate_Orbiter

s4u::Comm::test_any takes a reference

At boot time we now match on the mac address to know which interface to
configure

Flatbuffers can return nullptr in several occasions. We don't allow this
in our protocol, so I've added some additional checks when decoding an
incoming SendPacket.

vsg: Remove dead code from the old wire protocol

See merge request !28

This gathers the function to send/recv reliably messages on a socket.
Also contains a helper function to receive a size prefixed flatbuffer
message.

The patch also remove the src/vsg directory as well as the need for
building libvsg.a.

Fix cpp tests

See merge request !27

tansiv-client: Introduce trait InnerBufferDisplay for buffer-type specific...

See merge request !26

RILLING Louis authored Jan 28, 2022 and SIMONIN Matthieu committed Feb 21, 2022

tansiv-client: Introduce trait InnerBufferDisplay for buffer-type specific implementation of Display

tests: Make possible to get child actor exit code

Closes #28

See merge request !23

Fix #28

To test client / actor protocol communication, actor execution is forked
while client test code is executed in the parent process (the cargo test
process).

Until now, as soon as the client (parent process) finishes its execution
the actor (child process) is dropped/killed. In some situation the actor
don't have the chance to cover all of its test logic.
Also actor test logic fails silently in the sense that only error log
message are displayed when an asertion fails. No error is propagated to
the parent process (parent process).

This patch:
- make possible to wait for the actor to finish before leaving the test.
  A `wait` method is added to the TestActor implementation to wait for
  the actor to terminate. This can be called by a test if deemed necessary.
  For instance it's added in `run_client_and_actor` helper method which
  covers most of the connector tests.

- propagate an error to the parent process.
  User of `wait` can check the return code of the child process.

- enable logging in connector's tests. They can be displayed in `cargo
  tests` using `TEST_FLAGS="--nocapture"`. This is the case in the
  Dockerfile.

This can keep track of:
- the total number of messages going through tansiv
- message identification, each message is assigned an id (in the current
implementation this is the value of the counter at the time of reception)

when running a TS, tansiv injects the socket name as the first argument
of the script that is fork/exec-ed.
notansiv.py is the symetric to tansiv but for nts system but use the
same wrapper script at the end (boot.py). So we make notansiv also
inject a (dummy) socket name as the first parameter of the boot.py
script.

- custom target to launch the tests
    - (tansiv/cpp) add a wrapper script
    - (client/rust) call the test target of the client's Makefile
- clean TANPROC_LIBS use

extra includes are described at the top level

There's no consensus yet on the right IDE use among the developers ;)

Also add send test to the CI

Fix #27

- [x] tansiv part
- [x] qemu part

tests: Fix hangs and improve error handling

See merge request !19

Once the fake actor scenario is over, the fake actor starts running
(again) the client-side of the test. Although this is usually harmless
because the client-side quickly blocks waiting for the server to
accept the connection to its socket, fix this by exiting after making
sure that the client side is done (either by destroying ScenarioRunner
or by exiting abnormally).

The signal handler of SIGQUIT calls exit() which is not signal safe.
This sometimes leads to deadlocks (see the backtrace below where
pthread_mutex_lock() is interrupted by the signal and called on the
same lock by exit()). To fix this just call _exit().

 (gdb) bt
 #0  0x0000792090f0f5cb in __lll_lock_wait () from target:/lib/x86_64-linux-gnu/libpthread.so.0
 #1  0x0000792090f08211 in pthread_mutex_lock () from target:/lib/x86_64-linux-gnu/libpthread.so.0
 #2  0x0000792090f4735f in ?? () from target:/lib64/ld-linux-x86-64.so.2
 #3  0x0000792090a0a537 in ?? () from target:/lib/x86_64-linux-gnu/libc.so.6
 #4  0x0000792090a0a6da in exit () from target:/lib/x86_64-linux-gnu/libc.so.6
 #5  0x00005a95f258ed39 in sigquit (signum=3) at /usr/include/c++/11/bits/catch.hpp:18
 #6  <signal handler called>
 #7  0x0000792090f08193 in pthread_mutex_lock () from target:/lib/x86_64-linux-gnu/libpthread.so.0
 #8  0x0000792090f2a40d in dlsym () from target:/lib/x86_64-linux-gnu/libdl.so.2
 #9  0x00005a95f24cba89 in std::sys::unix::weak::fetch (name=<optimized out>)
     at library/std/src/sys/unix/weak.rs:101
 #10 std::sys::unix::weak::Weak<F>::initialize (
     self=0x5a95f2847c88 <std::sys::unix::rand::imp::getrandom::getrandom::getrandom>)
     at library/std/src/sys/unix/weak.rs:85
 #11 0x00005a95f267f398 in std::sys::unix::weak::Weak<F>::get (self=<optimized out>)
     at library/std/src/sys/unix/weak.rs:54
 #12 std::sys::unix::rand::imp::getrandom::getrandom (buffer=0x7ffd9778bd10, length=16, flags=1)
     at library/std/src/sys/unix/weak.rs:135
 #13 std::sys::unix::rand::imp::getrandom (buf=...) at library/std/src/sys/unix/rand.rs:44
 #14 std::sys::unix::rand::imp::getrandom_fill_bytes (v=...) at library/std/src/sys/unix/rand.rs:69
 #15 std::sys::unix::rand::imp::fill_bytes (v=...) at library/std/src/sys/unix/rand.rs:102
 #16 0x00005a95f2691808 in std::sys::unix::rand::hashmap_random_keys ()
     at library/std/src/sys/unix/rand.rs:8
 #17 0x00005a95f25dbffa in std::thread::local::fast::Key<T>::try_initialize ()
 #18 0x00005a95f25ddccd in <clap::args::arg_matcher::ArgMatcher as core::default::Default>::default ()
 #19 0x00005a95f26ce92d in clap::app::App::get_matches_from_safe ()
 #20 0x00005a95f26c8171 in tansiv_client::init ()
 #21 0x00005a95f25915fc in vsg_init ()
 #22 0x00005a95f24f8373 in ____C_A_T_C_H____T_E_S_T____9 () at /usr/include/c++/11/bits/atomic:145
 #23 0x00005a95f24e486c in Catch::TestInvokerAsFunction::invoke (this=0x5a95f39586c0)
     at /srv/src/build/std_abs.h:14321
 #24 0x00005a95f24e3bb3 in Catch::TestCase::invoke (this=0x5a95f3969f68)
     at /srv/src/build/std_abs.h:14160
 #25 0x00005a95f24de7ff in Catch::RunContext::invokeActiveTestCase (this=0x7ffd9778cf30)
     at /srv/src/build/std_abs.h:13020
 #26 0x00005a95f24de583 in Catch::RunContext::runCurrentTest (this=0x7ffd9778cf30, redirectedCout=...,
     redirectedCerr=...) at /srv/src/build/std_abs.h:12993
 #27 0x00005a95f24dd211 in Catch::RunContext::runTest (this=0x7ffd9778cf30, testCase=...)
     at /srv/src/build/std_abs.h:12754
 #28 0x00005a95f24dff27 in Catch::(anonymous namespace)::TestGroup::execute (this=0x7ffd9778cf20)
     at /srv/src/build/std_abs.h:13347
 #29 0x00005a95f24e1203 in Catch::Session::runInternal (this=0x7ffd9778d230)
     at /srv/src/build/std_abs.h:13553
 #30 0x00005a95f24e0f48 in Catch::Session::run (this=0x7ffd9778d230) at /srv/src/build/std_abs.h:13509
 #31 0x00005a95f251ee17 in Catch::Session::run<char> (this=0x7ffd9778d230, argc=2, argv=0x7ffd9778d4c8)
     at /srv/src/build/std_abs.h:13231
 #32 0x00005a95f24f636f in main (argc=2, argv=0x7ffd9778d4c8) at /srv/src/build/std_abs.h:17526

tests: Fix logic of test "VSG send piggyback port"

See merge request !18