Mercurial > pidgin.yaz
view src/protocols/oscar/txqueue.c @ 5205:fefad67de2c7
[gaim-migrate @ 5573]
I had a damn good commit message, but it was eaten. Let's try it again.
Announcing, Gaim Plugin API version 2.0, or GPAPIV2.0 for short.
There are lots'a cool thingies here.
Okay now, this isn't as cool as the previous message, but:
1) There's now a single entry function for all plugin types. It returns a
detailed information structure on the plugin. This removes a lot of the
ugliness from old plugins. Oh yeah, libicq wasn't converted to this, so
if you use it, well, you shouldn't have used it anyway, but now you
can't! bwahahaha. Use AIM/ICQ.
2) There are now 3 types of plugins: Standard, Loader, and Protocol
plugins.
Standard plugins are, well, standard, compiled plugins.
Loader plugins load other plugins. For example, the perl support is now
a loader plugin. It loads perl scripts. In the future, we'll have
Ruby and Python loader plugins.
Protocol plugins are, well, protocol plugins... yeah...
3) Plugins have unique IDs, so they can be referred to or automatically
updated from a plugin database in the future. Neat, huh?
4) Plugins will have dependency support in the future, and can be hidden,
so if you have, say, a logging core plugin, it won't have to show up,
but then you load the GTK+ logging plugin and it'll auto-load the core
plugin. Core/UI split plugins!
5) There will eventually be custom plugin signals and RPC of some sort, for
the core/ui split plugins.
So, okay, back up .gaimrc.
I'd like to thank my parents for their support, javabsp for helping convert
a bunch of protocol plugins, and Etan for helping convert a bunch of
standard plugins.
Have fun. If you have any problems, please let me know, but you probably
won't have anything major happen. You will have to convert your plugins,
though, and I'm not guaranteeing that all perl scripts will still work.
I'll end up changing the perl script API eventually, so I know they won't
down the road. Don't worry, though. It'll be mass cool.
faceprint wants me to just commit the damn code already. So, here we go!!!
..
..
I need a massage. From a young, cute girl. Are there any young, cute girls
in the audience? IM me plz k thx.
committer: Tailor Script <tailor@pidgin.im>
author | Christian Hammond <chipx86@chipx86.com> |
---|---|
date | Fri, 25 Apr 2003 06:47:33 +0000 |
parents | 858979ab3867 |
children | c86f075b269a |
line wrap: on
line source
/* * txqueue.c * * Herein lies all the mangement routines for the transmit (Tx) queue. * */ #define FAIM_INTERNAL #include <aim.h> #ifndef _WIN32 #include <sys/socket.h> #else #include "win32dep.h" #endif /* * Allocate a new tx frame. * * This is more for looks than anything else. * * Right now, that is. If/when we implement a pool of transmit * frames, this will become the request-an-unused-frame part. * * framing = AIM_FRAMETYPE_OFT/FLAP * chan = channel for FLAP, hdrtype for OFT * */ faim_internal aim_frame_t *aim_tx_new(aim_session_t *sess, aim_conn_t *conn, fu8_t framing, fu16_t chan, int datalen) { aim_frame_t *fr; if (!conn) { faimdprintf(sess, 0, "aim_tx_new: ERROR: no connection specified\n"); return NULL; } /* For sanity... */ if ((conn->type == AIM_CONN_TYPE_RENDEZVOUS) || (conn->type == AIM_CONN_TYPE_LISTENER)) { if (framing != AIM_FRAMETYPE_OFT) { faimdprintf(sess, 0, "aim_tx_new: attempted to allocate inappropriate frame type for rendezvous connection\n"); return NULL; } } else { if (framing != AIM_FRAMETYPE_FLAP) { faimdprintf(sess, 0, "aim_tx_new: attempted to allocate inappropriate frame type for FLAP connection\n"); return NULL; } } if (!(fr = (aim_frame_t *)malloc(sizeof(aim_frame_t)))) return NULL; memset(fr, 0, sizeof(aim_frame_t)); fr->conn = conn; fr->hdrtype = framing; if (fr->hdrtype == AIM_FRAMETYPE_FLAP) { fr->hdr.flap.type = chan; } else if (fr->hdrtype == AIM_FRAMETYPE_OFT) { fr->hdr.rend.type = chan; } else faimdprintf(sess, 0, "tx_new: unknown framing\n"); if (datalen > 0) { fu8_t *data; if (!(data = (unsigned char *)malloc(datalen))) { aim_frame_destroy(fr); return NULL; } aim_bstream_init(&fr->data, data, datalen); } return fr; } /* * aim_tx_enqeue__queuebased() * * The overall purpose here is to enqueue the passed in command struct * into the outgoing (tx) queue. Basically... * 1) Make a scope-irrelevent copy of the struct * 3) Mark as not-sent-yet * 4) Enqueue the struct into the list * 6) Return * * Note that this is only used when doing queue-based transmitting; * that is, when sess->tx_enqueue is set to &aim_tx_enqueue__queuebased. * */ static int aim_tx_enqueue__queuebased(aim_session_t *sess, aim_frame_t *fr) { if (!fr->conn) { faimdprintf(sess, 1, "aim_tx_enqueue: WARNING: enqueueing packet with no connecetion\n"); fr->conn = aim_getconn_type(sess, AIM_CONN_TYPE_BOS); } if (fr->hdrtype == AIM_FRAMETYPE_FLAP) { /* assign seqnum -- XXX should really not assign until hardxmit */ fr->hdr.flap.seqnum = aim_get_next_txseqnum(fr->conn); } fr->handled = 0; /* not sent yet */ /* see overhead note in aim_rxqueue counterpart */ if (!sess->queue_outgoing) sess->queue_outgoing = fr; else { aim_frame_t *cur; for (cur = sess->queue_outgoing; cur->next; cur = cur->next) ; cur->next = fr; } return 0; } /* * aim_tx_enqueue__immediate() * * Parallel to aim_tx_enqueue__queuebased, however, this bypasses * the whole queue mess when you want immediate writes to happen. * * Basically the same as its __queuebased couterpart, however * instead of doing a list append, it just calls aim_tx_sendframe() * right here. * */ static int aim_tx_enqueue__immediate(aim_session_t *sess, aim_frame_t *fr) { if (!fr->conn) { faimdprintf(sess, 1, "aim_tx_enqueue: ERROR: packet has no connection\n"); aim_frame_destroy(fr); return 0; } if (fr->hdrtype == AIM_FRAMETYPE_FLAP) fr->hdr.flap.seqnum = aim_get_next_txseqnum(fr->conn); fr->handled = 0; /* not sent yet */ aim_tx_sendframe(sess, fr); aim_frame_destroy(fr); return 0; } faim_export int aim_tx_setenqueue(aim_session_t *sess, int what, int (*func)(aim_session_t *, aim_frame_t *)) { if (what == AIM_TX_QUEUED) sess->tx_enqueue = &aim_tx_enqueue__queuebased; else if (what == AIM_TX_IMMEDIATE) sess->tx_enqueue = &aim_tx_enqueue__immediate; else if (what == AIM_TX_USER) { if (!func) return -EINVAL; sess->tx_enqueue = func; } else return -EINVAL; /* unknown action */ return 0; } faim_internal int aim_tx_enqueue(aim_session_t *sess, aim_frame_t *fr) { /* * If we want to send a connection thats inprogress, we have to force * them to use the queue based version. Otherwise, use whatever they * want. */ if (fr && fr->conn && (fr->conn->status & AIM_CONN_STATUS_INPROGRESS)) { return aim_tx_enqueue__queuebased(sess, fr); } return (*sess->tx_enqueue)(sess, fr); } /* * aim_get_next_txseqnum() * * This increments the tx command count, and returns the seqnum * that should be stamped on the next FLAP packet sent. This is * normally called during the final step of packet preparation * before enqueuement (in aim_tx_enqueue()). * */ faim_internal flap_seqnum_t aim_get_next_txseqnum(aim_conn_t *conn) { flap_seqnum_t ret; ret = ++conn->seqnum; return ret; } static int aim_send(int fd, const void *buf, size_t count) { int left, cur; for (cur = 0, left = count; left; ) { int ret; ret = send(fd, ((unsigned char *)buf)+cur, left, 0); if (ret == -1) return -1; else if (ret == 0) return cur; cur += ret; left -= ret; } return cur; } static int aim_bstream_send(aim_bstream_t *bs, aim_conn_t *conn, size_t count) { int wrote = 0; if (!bs || !conn || (count < 0)) return -EINVAL; if (count > aim_bstream_empty(bs)) count = aim_bstream_empty(bs); /* truncate to remaining space */ if (count) { if ((conn->type == AIM_CONN_TYPE_RENDEZVOUS) && (conn->subtype == AIM_CONN_SUBTYPE_OFT_DIRECTIM)) { /* I strongly suspect that this is a horrible thing to do * and I feel really guilty doing it. */ const char *sn = aim_odc_getsn(conn); aim_rxcallback_t userfunc; while (count - wrote > 1024) { wrote = wrote + aim_send(conn->fd, bs->data + bs->offset + wrote, 1024); if ((userfunc=aim_callhandler(conn->sessv, conn, AIM_CB_FAM_SPECIAL, AIM_CB_SPECIAL_IMAGETRANSFER))) userfunc(conn->sessv, NULL, sn, count-wrote>1024 ? ((double)wrote / count) : 1); } } if (count - wrote) { wrote = wrote + aim_send(conn->fd, bs->data + bs->offset + wrote, count - wrote); } } if (((aim_session_t *)conn->sessv)->debug >= 2) { int i; aim_session_t *sess = (aim_session_t *)conn->sessv; faimdprintf(sess, 2, "\nOutgoing data: (%d bytes)", wrote); for (i = 0; i < wrote; i++) { if (!(i % 8)) faimdprintf(sess, 2, "\n\t"); faimdprintf(sess, 2, "0x%02x ", *(bs->data + bs->offset + i)); } faimdprintf(sess, 2, "\n"); } bs->offset += wrote; return wrote; } static int sendframe_flap(aim_session_t *sess, aim_frame_t *fr) { aim_bstream_t obs; fu8_t *obs_raw; int payloadlen, err = 0, obslen; payloadlen = aim_bstream_curpos(&fr->data); if (!(obs_raw = malloc(6 + payloadlen))) return -ENOMEM; aim_bstream_init(&obs, obs_raw, 6 + payloadlen); /* FLAP header */ aimbs_put8(&obs, 0x2a); aimbs_put8(&obs, fr->hdr.flap.type); aimbs_put16(&obs, fr->hdr.flap.seqnum); aimbs_put16(&obs, payloadlen); /* payload */ aim_bstream_rewind(&fr->data); aimbs_putbs(&obs, &fr->data, payloadlen); obslen = aim_bstream_curpos(&obs); aim_bstream_rewind(&obs); if (aim_bstream_send(&obs, fr->conn, obslen) != obslen) err = -errno; free(obs_raw); /* XXX aim_bstream_free */ fr->handled = 1; fr->conn->lastactivity = time(NULL); return err; } static int sendframe_rendezvous(aim_session_t *sess, aim_frame_t *fr) { aim_bstream_t bs; fu8_t *bs_raw; int err = 0; int totlen = 8 + aim_bstream_curpos(&fr->data); if (!(bs_raw = malloc(totlen))) return -1; aim_bstream_init(&bs, bs_raw, totlen); aimbs_putraw(&bs, fr->hdr.rend.magic, 4); aimbs_put16(&bs, 8 + fr->hdr.rend.hdrlen); aimbs_put16(&bs, fr->hdr.rend.type); /* payload */ aim_bstream_rewind(&fr->data); aimbs_putbs(&bs, &fr->data, totlen - 8); aim_bstream_rewind(&bs); if (aim_bstream_send(&bs, fr->conn, totlen) != totlen) err = -errno; free(bs_raw); /* XXX aim_bstream_free */ fr->handled = 1; fr->conn->lastactivity = time(NULL); return err; } faim_internal int aim_tx_sendframe(aim_session_t *sess, aim_frame_t *fr) { if (fr->hdrtype == AIM_FRAMETYPE_FLAP) return sendframe_flap(sess, fr); else if (fr->hdrtype == AIM_FRAMETYPE_OFT) return sendframe_rendezvous(sess, fr); return -1; } faim_export int aim_tx_flushqueue(aim_session_t *sess) { aim_frame_t *cur; for (cur = sess->queue_outgoing; cur; cur = cur->next) { if (cur->handled) continue; /* already been sent */ if (cur->conn && (cur->conn->status & AIM_CONN_STATUS_INPROGRESS)) continue; /* * And now for the meager attempt to force transmit * latency and avoid missed messages. */ if ((cur->conn->lastactivity + cur->conn->forcedlatency) >= time(NULL)) { /* * XXX should be a break! we dont want to block the * upper layers * * XXX or better, just do this right. * */ sleep((cur->conn->lastactivity + cur->conn->forcedlatency) - time(NULL)); } /* XXX this should call the custom "queuing" function!! */ aim_tx_sendframe(sess, cur); } /* purge sent commands from queue */ aim_tx_purgequeue(sess); return 0; } /* * aim_tx_purgequeue() * * This is responsable for removing sent commands from the transmit * queue. This is not a required operation, but it of course helps * reduce memory footprint at run time! * */ faim_export void aim_tx_purgequeue(aim_session_t *sess) { aim_frame_t *cur, **prev; for (prev = &sess->queue_outgoing; (cur = *prev); ) { if (cur->handled) { *prev = cur->next; aim_frame_destroy(cur); } else prev = &cur->next; } return; } /** * aim_tx_cleanqueue - get rid of packets waiting for tx on a dying conn * @sess: session * @conn: connection that's dying * * for now this simply marks all packets as sent and lets them * disappear without warning. * */ faim_internal void aim_tx_cleanqueue(aim_session_t *sess, aim_conn_t *conn) { aim_frame_t *cur; for (cur = sess->queue_outgoing; cur; cur = cur->next) { if (cur->conn == conn) cur->handled = 1; } return; }