WIP: Make conferences persistent across restarts and reconnects.

[iphydf]

This is not new groupchats. This is just upgrade to old groupchats with
some advantages:

• Groupchats are now saved into tox_save.
• Clients can get groupchat unique id to save message log.
• Auto restore groupchats after restart even your friend uses
  non-upgraded version.

Submission will happen through #295. Review this PR only.

---

This change is 

[JFreegman]

toxcore/group.c, line 1181 at r1 (raw file):

    for (uint16_t i = groupnumber + 1; i < g_c->num_chats; ++i) {
        if (g_c->chats[i].live) {
            return 0;

When kill_groupchats() is called from tox_kill(), this function fails to free the memory associated with some of the groups (via realloc_conferences() below) because of this line. I'm not sure what the live variable indicates but all memory associated with conferences should be freed on shutdown regardless of the state. Commenting this line out fixes this particular memory leak, but causes others elsewhere.

---

Comments from Reviewable

[sudden6]

Reviewed 12 of 14 files at r1.
Review status: 12 of 14 files reviewed at latest revision, 4 unresolved discussions, some commit checks failed.

---

toxcore/group.c, line 2382 at r1 (raw file):

static bool self_peer_gid_collision(Group_c *g)
{
    uint32_t me = ~0;

use UINT32_MAX

---

toxcore/group.c, line 2586 at r1 (raw file):

        for (j = 0; j < DESIRED_CLOSE_CONNECTIONS; ++j) {
            if (closest(g, j) && g->closest_peers[j] == i) {
                j = DESIRED_CLOSE_CONNECTIONS + 100;

I think this magic + 100 thing is only used to get into the condition below, would be better to replace it with a flag variable

---

toxcore/group.c, line 2592 at r1 (raw file):

        }

        if (j == DESIRED_CLOSE_CONNECTIONS + 100) {

here

---

Comments from Reviewable

[iphydf]

Review status: 11 of 14 files reviewed at latest revision, 4 unresolved discussions.

---

toxcore/group.c, line 2382 at r1 (raw file):

Previously, sudden6 wrote…

use UINT32_MAX

Done.

---

toxcore/group.c, line 2586 at r1 (raw file):

Previously, sudden6 wrote…

I think this magic + 100 thing is only used to get into the condition below, would be better to replace it with a flag variable

Done.

---

toxcore/group.c, line 2592 at r1 (raw file):

Previously, sudden6 wrote…

here

Done.

---

Comments from Reviewable

[sudden6]

This were the last things I saw, from my side

---

Review status: 11 of 14 files reviewed at latest revision, 1 unresolved discussion.

---

Comments from Reviewable

[gongzisun]

Review a month, how is it going？

[tox-user]

I've been testing persistent groups with toxcore 0.2.2 for a week now. The biggest surprise for me was that they work even if I'm the only peer in a group who uses them. Other peers in a group don't need to have toxcore with persistent group chats to allow me to reconnect automatically. The biggest problem that I've noticed is that there is no indication that someone left the group while I was offline. Which means I will often wait to get reconnected, but it might never happen, because the other peers might have left. This could be partially fixed by clients. Instead of creating a persistent group only when we are invited to it, they should be always created on client start. That way it would be visible that we are in a group, even when none of the peers are online or they have left the group.

[envsh]

The biggest problem that I've noticed is that there is no indication that someone left the group while I was offline.

for persistent conference, seems need some way to get persistent peer list, and every peers online status.

[tox-user]

Yes, that would be great to have at some point.

I also have some weird issues with toxync and persistent groups. I don't get reconnected to toxync groups with autoinvite disabled when I come online, but when I then ask toxync to send me an invite to that group, it gets accepted without my permission. It's very strange, because it only happens with toxync groups.

[iphydf]

Superseded by #1069.