Mercurial > pidgin
view src/protocols/oscar/rxqueue.c @ 12645:fc28451f5d96
[gaim-migrate @ 14983]
SF Patch #1314512 from Sadrul (who has a patch for everything)
"This patch introduces a flag for protocol plugins that
support offline messages (like Y!M and ICQ). This was
encouraged by the following conversation:
<sadrul> should offline buddies be listed/enabled in
the send-to menu?
<rekkanoryo> i would think only for protocols that
support offline messaging, if it's indicated that the
buddy is offline
-- <snip> --
<Bleeter> sadrul: personally, I'd like to see a
'supports offline' flag of some description
<Bleeter> one could then redirect (via plugins) through
email or alternative methods
<Bleeter> just a thought
<Paco-Paco> yeah, that sounds like a reasonble thing to have
This patch uses this flag to disable the buddies in the
send-to menu who are offline and the protocol doesn't
support offline messages."
I made this make the label insensitive instead of the whole menuitem. This
should address SimGuy's concerns about inconsistency (i.e. you could create a
conversation with someone via the buddy list that you couldn't create via the
Send To menu). I also hacked up some voodoo to show the label as sensitive when
moused-over, as that looks better (given the label-insensitive thing is itself a
hack). I think this works quite well.
BUG NOTE:
This makes more obvious an existing bug. The Send To menu isn't updated when
buddies sign on or off or change status (at least under some circumstances).
We need to fix that anyway, so I'm not going to let it hold up this commit.
Switching tabs will clear it up. I'm thinking we just might want to build the
contents of that menu when it is selected. That would save us a mess of
inefficient signal callbacks that update the Send To menus in open windows all
the time.
AIM NOTE:
This assumes that AIM can't offline message. That's not strictly true. You can
message invisible users on AIM. However, by design, we can't tell when a user
is invisible without resorting to dirty hackery. In practice, this isn't a
problem, as you can still select the AIM user from the menu. And really, how
often will you be choosing the Invisible contact, rather than the user going
Invisible in the middle of a conversation or IMing you while they're Invisible?
JABBER NOTE:
This assumes that Jabber can always offline message. This isn't strictly true.
Sadrul said:
I have updated Jabber according to this link which seems to
talk about how to determine the existence offline-message
support in a server:
http://www.jabber.org/jeps/jep-0013.html#discover
However, jabber.org doesn't seem to send the required
info. So I am not sure about it.
He later said:
I talked to Nathan and he said offline message support is
mostly assumed for most jabber servers. GTalk doesn't yet
support it, but they are working on it. So I have made
jabber to always return TRUE.
If there is truly no way to detect offline messaging capability, then this is
an acceptable solution. We could special case Google Talk because of its
popularity, and remove that later. It's probably not worth it though.
MSN NOTE:
This assumes that MSN can never offline message. That's effectively true, but
to be technically correct, MSN can offline message if there's already a
switchboard conversation open with a user. We could write an offline_message
function in the MSN prpl to detect that, but it'd be of limited usefulness,
especially given that under most circumstances (where this might matter), the
switchboard connection will be closed almost immediately.
CVS NOTE:
I'm writing to share a tragic little story.
I have a PC that I use for Gaim development. One day, I was writing a commit
message on it, when all of a suddent it went berserk. The screen started
flashing, and the whole commit message just disappeared. All of it. And it was
a good commit message! I had to cram and rewrite it really quickly. Needless to
say, my rushed commit message wasn't nearly as good, and I blame the PC for that.
Seriously, though, what kind of version control system loses your commit
message on a broken connection to the server? Stupid!
committer: Tailor Script <tailor@pidgin.im>
author | Richard Laager <rlaager@wiktel.com> |
---|---|
date | Fri, 23 Dec 2005 19:26:04 +0000 |
parents | c5c0f714d8bc |
children | af3d6c6aee6b |
line wrap: on
line source
/* * This file contains the management routines for the receive * (incoming packet) queue. The actual packet handlers are in * rxhandlers.c. */ #define FAIM_INTERNAL #include <aim.h> #ifndef _WIN32 #include <sys/socket.h> #endif /* * */ faim_internal int aim_recv(int fd, void *buf, size_t count) { int left, cur; for (cur = 0, left = count; left; ) { int ret; ret = recv(fd, ((unsigned char *)buf)+cur, left, 0); /* Of course EOF is an error, only morons disagree with that. */ if (ret <= 0) return -1; cur += ret; left -= ret; } return cur; } /* * Read into a byte stream. Will not read more than count, but may read * less if there is not enough room in the stream buffer. */ faim_internal int aim_bstream_recv(aim_bstream_t *bs, int fd, size_t count) { int red = 0; if (!bs || (fd < 0)) return -1; if (count > (bs->len - bs->offset)) count = bs->len - bs->offset; /* truncate to remaining space */ if (count) { red = aim_recv(fd, bs->data + bs->offset, count); if (red <= 0) return -1; } bs->offset += red; return red; } /** * Free an aim_frame_t * * @param frame The frame to free. * @return -1 on error; 0 on success. */ faim_internal void aim_frame_destroy(aim_frame_t *frame) { free(frame->data.data); /* XXX aim_bstream_free */ free(frame); return; } /* * Read a FLAP header from conn into fr, and return the number of * bytes in the payload. * * @return -1 on error, otherwise return the length of the payload. */ static int aim_get_command_flap(aim_session_t *sess, aim_conn_t *conn, aim_frame_t *fr) { fu8_t hdr_raw[6]; aim_bstream_t hdr; fr->hdrtype = AIM_FRAMETYPE_FLAP; /* * Read FLAP header. Six bytes total. * * Byte # | Description * -------|------------- * 0x00 | Always 0x2a * 0x01 | Channel number, usually "2." "1" is used during login, * | 4 is used during logoff. * 0x02 | Sequence number, 2 bytes. * 0x04 | Number of data bytes that follow, 2 bytes. */ aim_bstream_init(&hdr, hdr_raw, sizeof(hdr_raw)); if (aim_bstream_recv(&hdr, conn->fd, 6) < 6) { aim_conn_close(conn); return -1; } aim_bstream_rewind(&hdr); /* * This shouldn't happen unless the socket breaks, the server breaks, * or we break. We must handle it just in case. */ if (aimbs_get8(&hdr) != 0x2a) { gaim_debug_misc("oscar", "Invalid FLAP frame received on FLAP connection!"); aim_conn_close(conn); return -1; } fr->hdr.flap.channel = aimbs_get8(&hdr); fr->hdr.flap.seqnum = aimbs_get16(&hdr); return aimbs_get16(&hdr); } /* * Read a rendezvous header from conn into fr, and return the number of * bytes in the payload. * * @return -1 on error, otherwise return the length of the payload. */ static int aim_get_command_rendezvous(aim_session_t *sess, aim_conn_t *conn, aim_frame_t *fr) { fu8_t hdr_raw[8]; aim_bstream_t hdr; fr->hdrtype = AIM_FRAMETYPE_OFT; /* * Read rendezvous header */ aim_bstream_init(&hdr, hdr_raw, sizeof(hdr_raw)); if (aim_bstream_recv(&hdr, conn->fd, 8) < 8) { aim_conn_close(conn); return -1; } aim_bstream_rewind(&hdr); aimbs_getrawbuf(&hdr, fr->hdr.rend.magic, 4); fr->hdr.rend.hdrlen = aimbs_get16(&hdr); fr->hdr.rend.type = aimbs_get16(&hdr); return fr->hdr.rend.hdrlen - 8; } /* * Grab a single command sequence off the socket, and enqueue it in the incoming event queue * in a separate struct. * * @return 0 on success, otherwise return the error number. */ faim_export int aim_get_command(aim_session_t *sess, aim_conn_t *conn) { aim_frame_t *fr; int payloadlen; if (!sess || !conn) return -EINVAL; if (conn->fd == -1) return -1; /* it's an aim_conn_close()'d connection */ /* If stdin is closed, then zero becomes a valid fd if (conn->fd < 3) return -1; */ if (conn->status & AIM_CONN_STATUS_INPROGRESS) return aim_conn_completeconnect(sess, conn); if (!(fr = (aim_frame_t *)calloc(sizeof(aim_frame_t), 1))) return -ENOMEM; /* * Rendezvous (client to client) connections do not speak FLAP, so this * function will break on them. */ if (conn->type == AIM_CONN_TYPE_RENDEZVOUS) payloadlen = aim_get_command_rendezvous(sess, conn, fr); else if (conn->type == AIM_CONN_TYPE_LISTENER) { gaim_debug_misc("oscar", "AIM_CONN_TYPE_LISTENER on fd %d\n", conn->fd); free(fr); return -1; } else payloadlen = aim_get_command_flap(sess, conn, fr); if (payloadlen < 0) { free(fr); return -1; } if (payloadlen > 0) { fu8_t *payload = NULL; if (!(payload = (fu8_t *) malloc(payloadlen))) { aim_frame_destroy(fr); return -1; } aim_bstream_init(&fr->data, payload, payloadlen); /* read the payload */ if (aim_bstream_recv(&fr->data, conn->fd, payloadlen) < payloadlen) { aim_frame_destroy(fr); /* free's payload */ aim_conn_close(conn); return -1; } } else aim_bstream_init(&fr->data, NULL, 0); aim_bstream_rewind(&fr->data); fr->conn = conn; /* Enqueue this puppy */ fr->next = NULL; /* this will always be at the bottom */ if (sess->queue_incoming == NULL) sess->queue_incoming = fr; else { aim_frame_t *cur; for (cur = sess->queue_incoming; cur->next; cur = cur->next); cur->next = fr; } fr->conn->lastactivity = time(NULL); return 0; } /* * Purge receive queue of all handled commands (->handled==1). * */ faim_export void aim_purge_rxqueue(aim_session_t *sess) { aim_frame_t *cur, **prev; for (prev = &sess->queue_incoming; (cur = *prev); ) { if (cur->handled) { *prev = cur->next; aim_frame_destroy(cur); } else prev = &cur->next; } return; } /* * Since aim_get_command will aim_conn_kill dead connections, we need * to clean up the rxqueue of unprocessed connections on that socket. * * XXX: this is something that was handled better in the old connection * handling method, but eh. */ faim_internal void aim_rxqueue_cleanbyconn(aim_session_t *sess, aim_conn_t *conn) { aim_frame_t *currx; for (currx = sess->queue_incoming; currx; currx = currx->next) { if ((!currx->handled) && (currx->conn == conn)) currx->handled = 1; } return; }