feat: Add back the tox_loop implementation for low latency.

[iphydf]

Copied code from #335.

Fixes #225.

---

This change is 

[JFreegman]

Reviewed 12 of 12 files at r1, all commit messages.
Reviewable status: 0 of 1 approvals obtained (waiting on @iphydf)

---

auto_tests/tox_loop_test.c, line 20 at r1 (raw file):

typedef struct {
    int start_count, stop_count;

Should be on separate lines

---

auto_tests/tox_loop_test.c, line 49 at r1 (raw file):

{
    pthread_t worker, worker_tcp;
    struct Tox_Options *opts = tox_options_new(NULL);

nullptr (this needs to be fixed PR-wide)

---

toxcore/Messenger.c, line 1911 at r1 (raw file):

#if defined(HAVE_LIBEV) || defined(HAVE_LIBEVENT)

    if (!m->dispatcher) {

== nullptr

---

toxcore/TCP_connection.c, line 63 at r1 (raw file):

 * Return number of elements of TCP connection array.
 *
 * @param tcp_c struct containing TCP_con array

Lots of missing .'s in the comments

---

toxcore/tox.c, line 875 at r1 (raw file):

    uint32_t len = tcp_connections_length(nc_get_tcp_c(m->net_crypto));

    for (uint32_t i = 0; i < len; i++) {

++i (one of many)

---

toxcore/tox.c, line 1025 at r1 (raw file):

#endif

bool tox_loop(Tox *tox, void *user_data, TOX_ERR_LOOP *error)

Would it make sense to split this up via helper functions for each ifdef condition?

---

toxcore/tox.c, line 1037 at r1 (raw file):

#ifdef HAVE_LIBEV
    bool ret = true;
    Event_Arg *tmp = (Event_Arg *) calloc(1, sizeof(Event_Arg));

Missing null check after calloc

---

toxcore/tox.c, line 1070 at r1 (raw file):

    free(tmp);
#elif defined(HAVE_LIBEVENT)
    Event_Arg *tmp = (Event_Arg *) calloc(1, sizeof(Event_Arg));

Missing null check after calloc

[sudden6]

Reviewable status: 1 change requests, 0 of 1 approvals obtained (waiting on @iphydf)

---

auto_tests/tox_loop_test.c, line 11 at r2 (raw file):

#define TCP_RELAY_PORT 33448
/* The Travis-CI container responds poorly to ::1 as a localhost address

We don't use Travis anymore, still needed? At least update the comment

---

toxcore/tox.c, line 947 at r2 (raw file):

    uint32_t i = 0;

    while (i < fdcount - 1 && i < len) {

this could be rewritten as for loop with upper bound of len since fdcount = len + 1 a few lines above, which is easier to read. This whole section could use some cleanup.

---

toxcore/tox.c, line 994 at r2 (raw file):

    // TODO(Ansa89): travis states that "ev_run" returns "void",
    // but "man 3 ev" states it returns "bool"
#if 0

don't commit dead code?

---

toxcore/tox.c, line 1083 at r2 (raw file):

    free(fdlist);
#endif

which #if terminates here? I'd generally add a comment if the preprocessor instructions are nested or span several lines

[Green-Sky]

options should have a prefix like TOX_ or TOXCORE_