Mercurial > pidgin.yaz
changeset 20402:00e7d8ded567
propagate from branch 'im.pidgin.rlaager.merging.msnp13-and-pidgin' (head 8cfbe223ca6e514146a4d46aeab4c7816de17b79)
to branch 'im.pidgin.cpw.khc.msnp14' (head d64cf2078da03ec91620a88bb3408d3eb14a19e2)
author | Richard Laager <rlaager@wiktel.com> |
---|---|
date | Sun, 15 Apr 2007 05:01:55 +0000 |
parents | 4ddc27c18781 (current diff) ea9a5566a156 (diff) |
children | 599b5a4aebab |
files | libpurple/protocols/msn/msn-utils.c libpurple/protocols/msn/msn-utils.h libpurple/protocols/msn/state.c libpurple/protocols/msn/user.h |
diffstat | 12 files changed, 1938 insertions(+), 574 deletions(-) [+] |
line wrap: on
line diff
--- a/libpurple/protocols/msn/dialog.c Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/dialog.c Sun Apr 15 05:01:55 2007 +0000 @@ -117,6 +117,7 @@ buddy = purple_find_buddy_in_group(account, passport, group); else buddy = purple_find_buddy(account, passport); + } if (buddy != NULL) purple_blist_remove_buddy(buddy);
--- a/libpurple/protocols/msn/msn-utils.c Sun Apr 15 04:59:43 2007 +0000 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,440 +0,0 @@ -/** - * @file msn-utils.c Utility functions - * - * purple - * - * Purple is the legal property of its developers, whose names are too numerous - * to list here. Please refer to the COPYRIGHT file distributed with this - * source distribution. - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA - */ -#include "msn.h" -#include "msn-utils.h" - -void -msn_parse_format(const char *mime, char **pre_ret, char **post_ret) -{ - char *cur; - GString *pre = g_string_new(NULL); - GString *post = g_string_new(NULL); - unsigned int colors[3]; - - if (pre_ret != NULL) *pre_ret = NULL; - if (post_ret != NULL) *post_ret = NULL; - - cur = strstr(mime, "FN="); - - if (cur && (*(cur = cur + 3) != ';')) - { - pre = g_string_append(pre, "<FONT FACE=\""); - - while (*cur && *cur != ';') - { - pre = g_string_append_c(pre, *cur); - cur++; - } - - pre = g_string_append(pre, "\">"); - post = g_string_prepend(post, "</FONT>"); - } - - cur = strstr(mime, "EF="); - - if (cur && (*(cur = cur + 3) != ';')) - { - while (*cur && *cur != ';') - { - pre = g_string_append_c(pre, '<'); - pre = g_string_append_c(pre, *cur); - pre = g_string_append_c(pre, '>'); - post = g_string_prepend_c(post, '>'); - post = g_string_prepend_c(post, *cur); - post = g_string_prepend_c(post, '/'); - post = g_string_prepend_c(post, '<'); - cur++; - } - } - - cur = strstr(mime, "CO="); - - if (cur && (*(cur = cur + 3) != ';')) - { - int i; - - i = sscanf(cur, "%02x%02x%02x;", &colors[0], &colors[1], &colors[2]); - - if (i > 0) - { - char tag[64]; - - if (i == 1) - { - colors[1] = 0; - colors[2] = 0; - } - else if (i == 2) - { - unsigned int temp = colors[0]; - - colors[0] = colors[1]; - colors[1] = temp; - colors[2] = 0; - } - else if (i == 3) - { - unsigned int temp = colors[2]; - - colors[2] = colors[0]; - colors[0] = temp; - } - - g_snprintf(tag, sizeof(tag), - "<FONT COLOR=\"#%02hhx%02hhx%02hhx\">", - colors[0], colors[1], colors[2]); - - pre = g_string_append(pre, tag); - post = g_string_prepend(post, "</FONT>"); - } - } - - cur = strstr(mime, "RL="); - - if (cur && (*(cur = cur + 3) != ';')) - { - if (*cur == '1') - { - /* RTL text was received */ - pre = g_string_append(pre, "<SPAN style=\"direction:rtl;text-align:right;\">"); - post = g_string_prepend(post, "</SPAN>"); - } - } - - cur = g_strdup(purple_url_decode(pre->str)); - g_string_free(pre, TRUE); - - if (pre_ret != NULL) - *pre_ret = cur; - else - g_free(cur); - - cur = g_strdup(purple_url_decode(post->str)); - g_string_free(post, TRUE); - - if (post_ret != NULL) - *post_ret = cur; - else - g_free(cur); -} - -/* - * We need this because we're only supposed to encode spaces in the font - * names. purple_url_encode() isn't acceptable. - */ -static const char * -encode_spaces(const char *str) -{ - static char buf[BUF_LEN]; - const char *c; - char *d; - - g_return_val_if_fail(str != NULL, NULL); - - for (c = str, d = buf; *c != '\0'; c++) - { - if (*c == ' ') - { - *d++ = '%'; - *d++ = '2'; - *d++ = '0'; - } - else - *d++ = *c; - } - - return buf; -} - -/* - * Taken from the zephyr plugin. - * This parses HTML formatting (put out by one of the gtkimhtml widgets - * and converts it to msn formatting. It doesn't deal with the tag closing, - * but gtkimhtml widgets give valid html. - * It currently deals properly with <b>, <u>, <i>, <font face=...>, - * <font color=...>, <span dir=...>, <span style="direction: ...">. - * It ignores <font back=...> and <font size=...> - */ -void -msn_import_html(const char *html, char **attributes, char **message) -{ - int len, retcount = 0; - const char *c; - char *msg; - char *fontface = NULL; - char fonteffect[4]; - char fontcolor[7]; - char direction = '0'; - - gboolean has_bold = FALSE; - gboolean has_italic = FALSE; - gboolean has_underline = FALSE; - gboolean has_strikethrough = FALSE; - - g_return_if_fail(html != NULL); - g_return_if_fail(attributes != NULL); - g_return_if_fail(message != NULL); - - len = strlen(html); - msg = g_malloc0(len + 1); - - memset(fontcolor, 0, sizeof(fontcolor)); - strcat(fontcolor, "0"); - memset(fonteffect, 0, sizeof(fonteffect)); - - for (c = html; *c != '\0';) - { - if (*c == '<') - { - if (!g_ascii_strncasecmp(c + 1, "br>", 3)) - { - msg[retcount++] = '\r'; - msg[retcount++] = '\n'; - c += 4; - } - else if (!g_ascii_strncasecmp(c + 1, "i>", 2)) - { - if (!has_italic) - { - strcat(fonteffect, "I"); - has_italic = TRUE; - } - c += 3; - } - else if (!g_ascii_strncasecmp(c + 1, "b>", 2)) - { - if (!has_bold) - { - strcat(fonteffect, "B"); - has_bold = TRUE; - } - c += 3; - } - else if (!g_ascii_strncasecmp(c + 1, "u>", 2)) - { - if (!has_underline) - { - strcat(fonteffect, "U"); - has_underline = TRUE; - } - c += 3; - } - else if (!g_ascii_strncasecmp(c + 1, "s>", 2)) - { - if (!has_strikethrough) - { - strcat(fonteffect, "S"); - has_strikethrough = TRUE; - } - c += 3; - } - else if (!g_ascii_strncasecmp(c + 1, "a href=\"", 8)) - { - c += 9; - - if (!g_ascii_strncasecmp(c, "mailto:", 7)) - c += 7; - - while ((*c != '\0') && g_ascii_strncasecmp(c, "\">", 2)) - msg[retcount++] = *c++; - - if (*c != '\0') - c += 2; - - /* ignore descriptive string */ - while ((*c != '\0') && g_ascii_strncasecmp(c, "</a>", 4)) - c++; - - if (*c != '\0') - c += 4; - } - else if (!g_ascii_strncasecmp(c + 1, "span", 4)) - { - /* Bi-directional text support using CSS properties in span tags */ - c += 5; - - while (*c != '\0' && *c != '>') - { - while (*c == ' ') - c++; - if (!g_ascii_strncasecmp(c, "dir=\"rtl\"", 9)) - { - c += 9; - direction = '1'; - } - else if (!g_ascii_strncasecmp(c, "style=\"", 7)) - { - /* Parse inline CSS attributes */ - char *attributes; - int attr_len = 0; - c += 7; - while (*(c + attr_len) != '\0' && *(c + attr_len) != '"') - attr_len++; - if (*(c + attr_len) == '"') - { - char *attr_dir; - attributes = g_strndup(c, attr_len); - attr_dir = purple_markup_get_css_property(attributes, "direction"); - if (attr_dir && (!strncasecmp(attr_dir, "RTL", 3))) - direction = '1'; - g_free(attr_dir); - g_free(attributes); - } - - } - else - { - c++; - } - } - if (*c == '>') - c++; - } - else if (!g_ascii_strncasecmp(c + 1, "font", 4)) - { - c += 5; - - while ((*c != '\0') && !g_ascii_strncasecmp(c, " ", 1)) - c++; - - if (!g_ascii_strncasecmp(c, "color=\"#", 7)) - { - c += 8; - - fontcolor[0] = *(c + 4); - fontcolor[1] = *(c + 5); - fontcolor[2] = *(c + 2); - fontcolor[3] = *(c + 3); - fontcolor[4] = *c; - fontcolor[5] = *(c + 1); - - c += 8; - } - else if (!g_ascii_strncasecmp(c, "face=\"", 6)) - { - const char *end = NULL; - const char *comma = NULL; - unsigned int namelen = 0; - - c += 6; - end = strchr(c, '\"'); - comma = strchr(c, ','); - - if (comma == NULL || comma > end) - namelen = (unsigned int)(end - c); - else - namelen = (unsigned int)(comma - c); - - fontface = g_strndup(c, namelen); - c = end + 2; - } - else - { - /* Drop all unrecognized/misparsed font tags */ - while ((*c != '\0') && g_ascii_strncasecmp(c, "\">", 2)) - c++; - - if (*c != '\0') - c += 2; - } - } - else - { - while ((*c != '\0') && (*c != '>')) - c++; - if (*c != '\0') - c++; - } - } - else if (*c == '&') - { - if (!g_ascii_strncasecmp(c, "<", 4)) - { - msg[retcount++] = '<'; - c += 4; - } - else if (!g_ascii_strncasecmp(c, ">", 4)) - { - msg[retcount++] = '>'; - c += 4; - } - else if (!g_ascii_strncasecmp(c, " ", 6)) - { - msg[retcount++] = ' '; - c += 6; - } - else if (!g_ascii_strncasecmp(c, """, 6)) - { - msg[retcount++] = '"'; - c += 6; - } - else if (!g_ascii_strncasecmp(c, "&", 5)) - { - msg[retcount++] = '&'; - c += 5; - } - else if (!g_ascii_strncasecmp(c, "'", 6)) - { - msg[retcount++] = '\''; - c += 6; - } - else - msg[retcount++] = *c++; - } - else - msg[retcount++] = *c++; - } - - if (fontface == NULL) - fontface = g_strdup("MS Sans Serif"); - - *attributes = g_strdup_printf("FN=%s; EF=%s; CO=%s; PF=0; RL=%c", - encode_spaces(fontface), - fonteffect, fontcolor, direction); - *message = g_strdup(msg); - - g_free(fontface); - g_free(msg); -} - -void -msn_parse_socket(const char *str, char **ret_host, int *ret_port) -{ - char *host; - char *c; - int port; - - host = g_strdup(str); - - if ((c = strchr(host, ':')) != NULL) - { - *c = '\0'; - port = atoi(c + 1); - } - else - port = 1863; - - *ret_host = host; - *ret_port = port; -}
--- a/libpurple/protocols/msn/msn-utils.h Sun Apr 15 04:59:43 2007 +0000 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,51 +0,0 @@ -/** - * @file msn-utils.h Utility functions - * - * purple - * - * Purple is the legal property of its developers, whose names are too numerous - * to list here. Please refer to the COPYRIGHT file distributed with this - * source distribution. - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA - */ -#ifndef _MSN_UTILS_H_ -#define _MSN_UTILS_H_ - -/** - * Parses the MSN message formatting into a format compatible with Purple. - * - * @param mime The mime header with the formatting. - * @param pre_ret The returned prefix string. - * @param post_ret The returned postfix string. - * - * @return The new message. - */ -void msn_parse_format(const char *mime, char **pre_ret, char **post_ret); - -/** - * Parses the Purple message formatting (html) into the MSN format. - * - * @param html The html message to format. - * @param attributes The returned attributes string. - * @param message The returned message string. - * - * @return The new message. - */ -void msn_import_html(const char *html, char **attributes, char **message); - -void msn_parse_socket(const char *str, char **ret_host, int *ret_port); - -#endif /* _MSN_UTILS_H_ */
--- a/libpurple/protocols/msn/msn.c Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/msn.c Sun Apr 15 05:01:55 2007 +0000 @@ -114,7 +114,7 @@ return PURPLE_CMD_RET_OK; } -static void +void msn_act_id(PurpleConnection *gc, const char *entry) { MsnCmdProc *cmdproc; @@ -450,6 +450,7 @@ session = gc->proto_data; xfer = purple_xfer_new(gc->account, PURPLE_XFER_SEND, who); + if (xfer) { slplink = msn_session_get_slplink(session, who); @@ -501,44 +502,6 @@ return "msn"; } -static void -msn_list_emblems(PurpleBuddy *b, const char **se, const char **sw, - const char **nw, const char **ne) -{ - MsnUser *user; - PurplePresence *presence; - const char *emblems[4] = { NULL, NULL, NULL, NULL }; - int i = 0; - - user = b->proto_data; - presence = purple_buddy_get_presence(b); - - if (!purple_presence_is_online(presence)) - emblems[i++] = "offline"; - else if (purple_presence_is_status_active(presence, "busy") || - purple_presence_is_status_active(presence, "phone")) - emblems[i++] = "occupied"; - else if (!purple_presence_is_available(presence)) - emblems[i++] = "away"; - - if (user == NULL) - { - emblems[0] = "offline"; - } - else - { - if (user->mobile) - emblems[i++] = "wireless"; - if (!(user->list_op & (1 << MSN_LIST_RL))) - emblems[i++] = "nr"; - } - - *se = emblems[0]; - *sw = emblems[1]; - *nw = emblems[2]; - *ne = emblems[3]; -} - /* * Set the User status text * Add the PSM String Using "Name - PSM String" format @@ -548,20 +511,30 @@ { PurplePresence *presence; PurpleStatus *status; - const char *msg, *name; + const char *msg, *name, *cmedia; char *psm_str, *tmp2, *text; presence = purple_buddy_get_presence(buddy); status = purple_presence_get_active_status(presence); msg = purple_status_get_attr_string(status, "message"); + cmedia=purple_status_get_attr_string(status, "currentmedia"); + if (!purple_presence_is_available(presence) && !purple_presence_is_idle(presence)){ name = purple_status_get_name(status); }else{ name = NULL; } - if (msg != NULL) { + if (cmedia != NULL) { + if(name) { + tmp2 = g_strdup_printf("%s - %s", name, cmedia); + text = g_markup_escape_text(tmp2, -1); + } else { + text = g_markup_escape_text(cmedia, -1); + } + return text; + } else if (msg != NULL) { tmp2 = purple_markup_strip_html(msg); if (name){ psm_str = g_strdup_printf("%s - %s", name, tmp2); @@ -596,14 +569,23 @@ if (purple_presence_is_online(presence)) { - char *psm; - psm = msn_status_text(buddy); + const char *psm, *currentmedia; + char *tmp; + + psm = purple_status_get_attr_string(status, "message"); + currentmedia = purple_status_get_attr_string(status, "currentmedia"); purple_notify_user_info_add_pair(user_info, _("Status"), (purple_presence_is_idle(presence) ? _("Idle") : purple_status_get_name(status))); if (psm) { - purple_notify_user_info_add_pair(user_info, _("PSM"), psm); - g_free(psm); + tmp = g_markup_escape_text(psm, -1); + purple_notify_user_info_add_pair(user_info, _("PSM"), tmp); + g_free(tmp); + } + if (currentmedia) { + tmp = g_markup_escape_text(currentmedia, -1); + purple_notify_user_info_add_pair(user_info, _("Current media"), tmp); + g_free(tmp); } } @@ -636,34 +618,40 @@ status = purple_status_type_new_with_attrs( PURPLE_STATUS_AVAILABLE, NULL, NULL, TRUE, TRUE, FALSE, "message", _("Message"), purple_value_new(PURPLE_TYPE_STRING), + "currentmedia", _("Current media"), purple_value_new(PURPLE_TYPE_STRING), NULL); types = g_list_append(types, status); status = purple_status_type_new_with_attrs( PURPLE_STATUS_AWAY, NULL, NULL, TRUE, TRUE, FALSE, "message", _("Message"), purple_value_new(PURPLE_TYPE_STRING), + "currentmedia", _("Current media"), purple_value_new(PURPLE_TYPE_STRING), NULL); types = g_list_append(types, status); status = purple_status_type_new_with_attrs( PURPLE_STATUS_AWAY, "brb", _("Be Right Back"), TRUE, TRUE, FALSE, "message", _("Message"), purple_value_new(PURPLE_TYPE_STRING), + "currentmedia", _("Current media"), purple_value_new(PURPLE_TYPE_STRING), NULL); types = g_list_append(types, status); status = purple_status_type_new_with_attrs( PURPLE_STATUS_AWAY, "busy", _("Busy"), TRUE, TRUE, FALSE, "message", _("Message"), purple_value_new(PURPLE_TYPE_STRING), + "currentmedia", _("Current media"), purple_value_new(PURPLE_TYPE_STRING), NULL); types = g_list_append(types, status); status = purple_status_type_new_with_attrs( PURPLE_STATUS_AWAY, "phone", _("On the Phone"), TRUE, TRUE, FALSE, "message", _("Message"), purple_value_new(PURPLE_TYPE_STRING), + "currentmedia", _("Current media"), purple_value_new(PURPLE_TYPE_STRING), NULL); types = g_list_append(types, status); status = purple_status_type_new_with_attrs( PURPLE_STATUS_AWAY, "lunch", _("Out to Lunch"), TRUE, TRUE, FALSE, "message", _("Message"), purple_value_new(PURPLE_TYPE_STRING), + "currentmedia", _("Current media"), purple_value_new(PURPLE_TYPE_STRING), NULL); types = g_list_append(types, status); @@ -674,7 +662,11 @@ status = purple_status_type_new_full(PURPLE_STATUS_OFFLINE, NULL, NULL, FALSE, TRUE, FALSE); types = g_list_append(types, status); - + + status = purple_status_type_new_full(PURPLE_STATUS_MOBILE, + "mobile", NULL, FALSE, FALSE, TRUE); + types = g_list_append(types, status); + return types; } @@ -795,9 +787,7 @@ gc->wants_to_die = TRUE; purple_connection_error(gc, _("SSL support is needed for MSN. Please install a supported " - "SSL library. See http://purple.sf.net/faq-ssl.php for more " - "information.")); - + "SSL library.")); return; } @@ -844,6 +834,7 @@ PurpleMessageFlags flags) { PurpleAccount *account; + PurpleBuddy *buddy = purple_find_buddy(gc->account, who); MsnMessage *msg; char *msgformat; char *msgtext; @@ -851,6 +842,16 @@ purple_debug_info("MaYuan","send IM {%s} to %s\n",message,who); account = purple_connection_get_account(gc); + if (buddy) { + PurplePresence *p = purple_buddy_get_presence(buddy); + if (purple_presence_is_status_primitive_active(p, PURPLE_STATUS_MOBILE)) { + char *text = purple_markup_strip_html(message); + send_to_mobile(gc, who, text); + g_free(text); + return 1; + } + } + msn_import_html(message, &msgformat, &msgtext); if(msn_user_is_online(account, who)|| msn_user_is_yahoo(account, who)){ @@ -1908,7 +1909,7 @@ "This either means that the user does not exist, " "or that the user exists " "but has not created a public profile.") : - _("Purple could not find " /* This should never happen */ + _("Could not find " /* This should never happen */ "any information in the user's profile. " "The user most likely does not exist.")))); } @@ -2051,6 +2052,71 @@ return TRUE; } +static PurpleAccount *find_acct(const char *prpl, const char *acct_id) +{ + PurpleAccount *acct = NULL; + + /* If we have a specific acct, use it */ + if (acct_id) { + acct = purple_accounts_find(acct_id, prpl); + if (acct && !purple_account_is_connected(acct)) + acct = NULL; + } else { /* Otherwise find an active account for the protocol */ + GList *l = purple_accounts_get_all(); + while (l) { + if (!strcmp(prpl, purple_account_get_protocol_id(l->data)) + && purple_account_is_connected(l->data)) { + acct = l->data; + break; + } + l = l->next; + } + } + + return acct; +} + +static gboolean msn_uri_handler(const char *proto, const char *cmd, GHashTable *params) +{ + char *acct_id = g_hash_table_lookup(params, "account"); + PurpleAccount *acct; + + if (g_ascii_strcasecmp(proto, "msnim")) + return FALSE; + + acct = find_acct("prpl-msn", acct_id); + + if (!acct) + return FALSE; + + /* msnim:chat?contact=user@domain.tld */ + if (!g_ascii_strcasecmp(cmd, "Chat")) { + char *sname = g_hash_table_lookup(params, "contact"); + if (sname) { + PurpleConversation *conv = purple_find_conversation_with_account( + PURPLE_CONV_TYPE_IM, sname, acct); + if (conv == NULL) + conv = purple_conversation_new(PURPLE_CONV_TYPE_IM, acct, sname); + purple_conversation_present(conv); + } + /*else + **If pidgindialogs_im() was in the core, we could use it here. + * It is all purple_request_* based, but I'm not sure it really belongs in the core + pidgindialogs_im();*/ + + return TRUE; + } + /* msnim:add?contact=user@domain.tld */ + else if (!g_ascii_strcasecmp(cmd, "Add")) { + char *name = g_hash_table_lookup(params, "contact"); + purple_blist_request_add_buddy(acct, name, NULL, NULL); + return TRUE; + } + + return FALSE; +} + + static PurplePluginProtocolInfo prpl_info = { OPT_PROTO_MAIL_CHECK, @@ -2058,7 +2124,7 @@ NULL, /* protocol_options */ {"png", 0, 0, 96, 96, 0, PURPLE_ICON_SCALE_SEND}, /* icon_spec */ msn_list_icon, /* list_icon */ - msn_list_emblems, /* list_emblems */ + NULL, /* list_emblems */ msn_status_text, /* status_text */ msn_tooltip_text, /* tooltip_text */ msn_status_types, /* away_states */ @@ -2132,10 +2198,10 @@ "MSN", /**< name */ VERSION, /**< version */ /** summary */ - N_("MSN Protocol Plugin"), + N_("Windows Live Messenger Protocol Plugin"), /** description */ - N_("MSN Protocol Plugin"), - "Christian Hammond <chipx86@gnupdate.org>", /**< author */ + N_("Windows Live Messenger Protocol Plugin"), + "MaYuan <mayuan2006@gmail.com>", /**< author */ PURPLE_WEBSITE, /**< homepage */ msn_load, /**< load */
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/libpurple/protocols/msn/msnutils.c Sun Apr 15 05:01:55 2007 +0000 @@ -0,0 +1,1592 @@ +/** + * @file msnutils.c Utility functions + * + * purple + * + * Purple is the legal property of its developers, whose names are too numerous + * to list here. Please refer to the COPYRIGHT file distributed with this + * source distribution. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ +#include "msn.h" +#include "msnutils.h" +#include "time.h" +//#include <openssl/md5.h> + +char *rand_guid(void); + +/************************************************************************** + * Util + **************************************************************************/ +char * +rand_guid() +{ + return g_strdup_printf("%4X%4X-%4X-%4X-%4X-%4X%4X%4X", + rand() % 0xAAFF + 0x1111, + rand() % 0xAAFF + 0x1111, + rand() % 0xAAFF + 0x1111, + rand() % 0xAAFF + 0x1111, + rand() % 0xAAFF + 0x1111, + rand() % 0xAAFF + 0x1111, + rand() % 0xAAFF + 0x1111, + rand() % 0xAAFF + 0x1111); +} + +void +msn_parse_format(const char *mime, char **pre_ret, char **post_ret) +{ + char *cur; + GString *pre = g_string_new(NULL); + GString *post = g_string_new(NULL); + unsigned int colors[3]; + + if (pre_ret != NULL) *pre_ret = NULL; + if (post_ret != NULL) *post_ret = NULL; + + cur = strstr(mime, "FN="); + + if (cur && (*(cur = cur + 3) != ';')) + { + pre = g_string_append(pre, "<FONT FACE=\""); + + while (*cur && *cur != ';') + { + pre = g_string_append_c(pre, *cur); + cur++; + } + + pre = g_string_append(pre, "\">"); + post = g_string_prepend(post, "</FONT>"); + } + + cur = strstr(mime, "EF="); + + if (cur && (*(cur = cur + 3) != ';')) + { + while (*cur && *cur != ';') + { + pre = g_string_append_c(pre, '<'); + pre = g_string_append_c(pre, *cur); + pre = g_string_append_c(pre, '>'); + post = g_string_prepend_c(post, '>'); + post = g_string_prepend_c(post, *cur); + post = g_string_prepend_c(post, '/'); + post = g_string_prepend_c(post, '<'); + cur++; + } + } + + cur = strstr(mime, "CO="); + + if (cur && (*(cur = cur + 3) != ';')) + { + int i; + + i = sscanf(cur, "%02x%02x%02x;", &colors[0], &colors[1], &colors[2]); + + if (i > 0) + { + char tag[64]; + + if (i == 1) + { + colors[1] = 0; + colors[2] = 0; + } + else if (i == 2) + { + unsigned int temp = colors[0]; + + colors[0] = colors[1]; + colors[1] = temp; + colors[2] = 0; + } + else if (i == 3) + { + unsigned int temp = colors[2]; + + colors[2] = colors[0]; + colors[0] = temp; + } + + g_snprintf(tag, sizeof(tag), + "<FONT COLOR=\"#%02hhx%02hhx%02hhx\">", + colors[0], colors[1], colors[2]); + + pre = g_string_append(pre, tag); + post = g_string_prepend(post, "</FONT>"); + } + } + + cur = strstr(mime, "RL="); + + if (cur && (*(cur = cur + 3) != ';')) + { + if (*cur == '1') + { + /* RTL text was received */ + pre = g_string_append(pre, "<SPAN style=\"direction:rtl;text-align:right;\">"); + post = g_string_prepend(post, "</SPAN>"); + } + } + + cur = g_strdup(purple_url_decode(pre->str)); + g_string_free(pre, TRUE); + + if (pre_ret != NULL) + *pre_ret = cur; + else + g_free(cur); + + cur = g_strdup(purple_url_decode(post->str)); + g_string_free(post, TRUE); + + if (post_ret != NULL) + *post_ret = cur; + else + g_free(cur); +} + +/*encode the str to RFC2047 style + * Currently only support the UTF-8 and base64 encode + */ +char * +msn_encode_mime(const char *str) +{ + char *base64; + + base64 = gaim_base64_encode((guchar *)str, strlen(str)); + return g_strdup_printf("=?utf-8?B?%s?=", base64); +} + +/* + * We need this because we're only supposed to encode spaces in the font + * names. purple_url_encode() isn't acceptable. + */ +static const char * +encode_spaces(const char *str) +{ + static char buf[BUF_LEN]; + const char *c; + char *d; + + g_return_val_if_fail(str != NULL, NULL); + + for (c = str, d = buf; *c != '\0'; c++) + { + if (*c == ' ') + { + *d++ = '%'; + *d++ = '2'; + *d++ = '0'; + } + else + *d++ = *c; + } + + return buf; +} + +/* + * Taken from the zephyr plugin. + * This parses HTML formatting (put out by one of the gtkimhtml widgets + * and converts it to msn formatting. It doesn't deal with the tag closing, + * but gtkimhtml widgets give valid html. + * It currently deals properly with <b>, <u>, <i>, <font face=...>, + * <font color=...>, <span dir=...>, <span style="direction: ...">. + * It ignores <font back=...> and <font size=...> + */ +void +msn_import_html(const char *html, char **attributes, char **message) +{ + int len, retcount = 0; + const char *c; + char *msg; + char *fontface = NULL; + char fonteffect[4]; + char fontcolor[7]; + char direction = '0'; + + gboolean has_bold = FALSE; + gboolean has_italic = FALSE; + gboolean has_underline = FALSE; + gboolean has_strikethrough = FALSE; + + g_return_if_fail(html != NULL); + g_return_if_fail(attributes != NULL); + g_return_if_fail(message != NULL); + + len = strlen(html); + msg = g_malloc0(len + 1); + + memset(fontcolor, 0, sizeof(fontcolor)); + strcat(fontcolor, "0"); + memset(fonteffect, 0, sizeof(fonteffect)); + + for (c = html; *c != '\0';) + { + if (*c == '<') + { + if (!g_ascii_strncasecmp(c + 1, "br>", 3)) + { + msg[retcount++] = '\r'; + msg[retcount++] = '\n'; + c += 4; + } + else if (!g_ascii_strncasecmp(c + 1, "i>", 2)) + { + if (!has_italic) + { + strcat(fonteffect, "I"); + has_italic = TRUE; + } + c += 3; + } + else if (!g_ascii_strncasecmp(c + 1, "b>", 2)) + { + if (!has_bold) + { + strcat(fonteffect, "B"); + has_bold = TRUE; + } + c += 3; + } + else if (!g_ascii_strncasecmp(c + 1, "u>", 2)) + { + if (!has_underline) + { + strcat(fonteffect, "U"); + has_underline = TRUE; + } + c += 3; + } + else if (!g_ascii_strncasecmp(c + 1, "s>", 2)) + { + if (!has_strikethrough) + { + strcat(fonteffect, "S"); + has_strikethrough = TRUE; + } + c += 3; + } + else if (!g_ascii_strncasecmp(c + 1, "a href=\"", 8)) + { + c += 9; + + if (!g_ascii_strncasecmp(c, "mailto:", 7)) + c += 7; + + while ((*c != '\0') && g_ascii_strncasecmp(c, "\">", 2)) + msg[retcount++] = *c++; + + if (*c != '\0') + c += 2; + + /* ignore descriptive string */ + while ((*c != '\0') && g_ascii_strncasecmp(c, "</a>", 4)) + c++; + + if (*c != '\0') + c += 4; + } + else if (!g_ascii_strncasecmp(c + 1, "span", 4)) + { + /* Bi-directional text support using CSS properties in span tags */ + c += 5; + + while (*c != '\0' && *c != '>') + { + while (*c == ' ') + c++; + if (!g_ascii_strncasecmp(c, "dir=\"rtl\"", 9)) + { + c += 9; + direction = '1'; + } + else if (!g_ascii_strncasecmp(c, "style=\"", 7)) + { + /* Parse inline CSS attributes */ + char *attributes; + int attr_len = 0; + c += 7; + while (*(c + attr_len) != '\0' && *(c + attr_len) != '"') + attr_len++; + if (*(c + attr_len) == '"') + { + char *attr_dir; + attributes = g_strndup(c, attr_len); + attr_dir = purple_markup_get_css_property(attributes, "direction"); + if (attr_dir && (!strncasecmp(attr_dir, "RTL", 3))) + direction = '1'; + g_free(attr_dir); + g_free(attributes); + } + + } + else + { + c++; + } + } + if (*c == '>') + c++; + } + else if (!g_ascii_strncasecmp(c + 1, "font", 4)) + { + c += 5; + + while ((*c != '\0') && !g_ascii_strncasecmp(c, " ", 1)) + c++; + + if (!g_ascii_strncasecmp(c, "color=\"#", 7)) + { + c += 8; + + fontcolor[0] = *(c + 4); + fontcolor[1] = *(c + 5); + fontcolor[2] = *(c + 2); + fontcolor[3] = *(c + 3); + fontcolor[4] = *c; + fontcolor[5] = *(c + 1); + + c += 8; + } + else if (!g_ascii_strncasecmp(c, "face=\"", 6)) + { + const char *end = NULL; + const char *comma = NULL; + unsigned int namelen = 0; + + c += 6; + end = strchr(c, '\"'); + comma = strchr(c, ','); + + if (comma == NULL || comma > end) + namelen = (unsigned int)(end - c); + else + namelen = (unsigned int)(comma - c); + + fontface = g_strndup(c, namelen); + c = end + 2; + } + else + { + /* Drop all unrecognized/misparsed font tags */ + while ((*c != '\0') && g_ascii_strncasecmp(c, "\">", 2)) + c++; + + if (*c != '\0') + c += 2; + } + } + else + { + while ((*c != '\0') && (*c != '>')) + c++; + if (*c != '\0') + c++; + } + } + else if (*c == '&') + { + if (!g_ascii_strncasecmp(c, "<", 4)) + { + msg[retcount++] = '<'; + c += 4; + } + else if (!g_ascii_strncasecmp(c, ">", 4)) + { + msg[retcount++] = '>'; + c += 4; + } + else if (!g_ascii_strncasecmp(c, " ", 6)) + { + msg[retcount++] = ' '; + c += 6; + } + else if (!g_ascii_strncasecmp(c, """, 6)) + { + msg[retcount++] = '"'; + c += 6; + } + else if (!g_ascii_strncasecmp(c, "&", 5)) + { + msg[retcount++] = '&'; + c += 5; + } + else if (!g_ascii_strncasecmp(c, "'", 6)) + { + msg[retcount++] = '\''; + c += 6; + } + else + msg[retcount++] = *c++; + } + else + msg[retcount++] = *c++; + } + + if (fontface == NULL) + fontface = g_strdup("MS Sans Serif"); + + *attributes = g_strdup_printf("FN=%s; EF=%s; CO=%s; PF=0; RL=%c", + encode_spaces(fontface), + fonteffect, fontcolor, direction); + *message = g_strdup(msg); + + g_free(fontface); + g_free(msg); +} + +void +msn_parse_socket(const char *str, char **ret_host, int *ret_port) +{ + char *host; + char *c; + int port; + + host = g_strdup(str); + + if ((c = strchr(host, ':')) != NULL){ + *c = '\0'; + port = atoi(c + 1); + }else{ + port = 1863; + } + + *ret_host = host; + *ret_port = port; +} +/*************************************************************************** + * MSN Time Related Funciton + ***************************************************************************/ +#if 0 +int +msn_convert_iso8601(const char *timestr,struct tm tm_time) +{ + char temp[64]; + struct tm ctime; + time_t ts; + + gaim_debug_info("MaYuan","convert string is{%s}\n",timestr); + tzset(); + /*copy string first*/ + memset(temp, 0, sizeof(temp)); + strncpy(temp, timestr, strlen(timestr)); + + /*convert via strptime()*/ + memset(&ctime, 0, sizeof(struct tm)); + strptime(temp, "%d %b %Y %T %Z", &ctime); + ts = mktime(&ctime) - timezone; + localtime_r(&ts, tm_time); +} +#endif + +/*************************************************************************** + * MSN Challenge Computing Function + ***************************************************************************/ +/*check the edian of system*/ +int +isBigEndian(void) +{ + short int word = 0x0100; + char *byte = (char *)&word; + + return(byte[0]); +} + +/*swap utility*/ +unsigned int +swapInt(unsigned int dw) +{ + unsigned int tmp; + tmp = (dw & 0x000000FF); + tmp = ((dw & 0x0000FF00) >> 0x08) | (tmp << 0x08); + tmp = ((dw & 0x00FF0000) >> 0x10) | (tmp << 0x08); + tmp = ((dw & 0xFF000000) >> 0x18) | (tmp << 0x08); + return(tmp); +} + +/* + * Handle MSN Chanllege computation + *This algorithm reference with http://msnpiki.msnfanatic.com/index.php/MSNP11:Challenges + */ +#define BUFSIZE 256 +void +msn_handle_chl(char *input, char *output) +{ + GaimCipher *cipher; + GaimCipherContext *context; + char *productKey = MSNP13_WLM_PRODUCT_KEY, + *productID = MSNP13_WLM_PRODUCT_ID, + *hexChars = "0123456789abcdef", + buf[BUFSIZE]; + unsigned char md5Hash[16], *newHash; + unsigned int *md5Parts, *chlStringParts, newHashParts[5]; + + long long nHigh=0, nLow=0; + + int i, bigEndian; + + /* Determine our endianess */ + bigEndian = isBigEndian(); + + /* Create the MD5 hash by using Gaim MD5 algorithm*/ + cipher = gaim_ciphers_find_cipher("md5"); + context = gaim_cipher_context_new(cipher, NULL); + + gaim_cipher_context_append(context, (const guchar *)input, + strlen(input)); + gaim_cipher_context_append(context, (const guchar *)productKey, + strlen(productKey)); + gaim_cipher_context_digest(context, sizeof(md5Hash), md5Hash, NULL); + gaim_cipher_context_destroy(context); + + /* Split it into four integers */ + md5Parts = (unsigned int *)md5Hash; + for(i=0; i<4; i++){ + /* check for endianess */ + if(bigEndian) + md5Parts[i] = swapInt(md5Parts[i]); + + /* & each integer with 0x7FFFFFFF */ + /* and save one unmodified array for later */ + newHashParts[i] = md5Parts[i]; + md5Parts[i] &= 0x7FFFFFFF; + } + + /* make a new string and pad with '0' */ + snprintf(buf, BUFSIZE-5, "%s%s", input, productID); + i = strlen(buf); + memset(&buf[i], '0', 8 - (i % 8)); + buf[i + (8 - (i % 8))]='\0'; + + /* split into integers */ + chlStringParts = (unsigned int *)buf; + + /* this is magic */ + for (i=0; i<(strlen(buf)/4)-1; i+=2){ + long long temp; + + if(bigEndian){ + chlStringParts[i] = swapInt(chlStringParts[i]); + chlStringParts[i+1] = swapInt(chlStringParts[i+1]); + } + + temp=(md5Parts[0] * (((0x0E79A9C1 * (long long)chlStringParts[i]) % 0x7FFFFFFF)+nHigh) + md5Parts[1])%0x7FFFFFFF; + nHigh=(md5Parts[2] * (((long long)chlStringParts[i+1]+temp) % 0x7FFFFFFF) + md5Parts[3]) % 0x7FFFFFFF; + nLow=nLow + nHigh + temp; + } + nHigh=(nHigh+md5Parts[1]) % 0x7FFFFFFF; + nLow=(nLow+md5Parts[3]) % 0x7FFFFFFF; + + newHashParts[0]^=nHigh; + newHashParts[1]^=nLow; + newHashParts[2]^=nHigh; + newHashParts[3]^=nLow; + + /* swap more bytes if big endian */ + for(i=0; i<4 && bigEndian; i++) + newHashParts[i] = swapInt(newHashParts[i]); + + /* make a string of the parts */ + newHash = (unsigned char *)newHashParts; + + /* convert to hexadecimal */ + for (i=0; i<16; i++) + { + output[i*2]=hexChars[(newHash[i]>>4)&0xF]; + output[(i*2)+1]=hexChars[newHash[i]&0xF]; + } + + output[32]='\0'; + +// gaim_debug_info("MaYuan","chl output{%s}\n",output); +} + +#if (!defined(_XOPEN_SOURCE))||defined(_WIN32) + +#ifndef __P +# if defined (__GNUC__) || (defined (__STDC__) && __STDC__) +# define __P(args) args +# else +# define __P(args) () +# endif /* GCC. */ +#endif /* Not __P. */ + +#if defined(HAVE_LOCALTIME_R) && ! HAVE_LOCALTIME_R && ! defined localtime_r +# ifdef _LIBC +# define localtime_r __localtime_r +# else +/* Approximate localtime_r as best we can in its absence. */ +# define localtime_r my_localtime_r +static struct tm *localtime_r __P ((const time_t *, struct tm *)); +static struct tm * +localtime_r (t, tp) + const time_t *t; + struct tm *tp; +{ + struct tm *l = localtime (t); + if (! l) + return 0; + *tp = *l; + return tp; +} +# endif /* ! _LIBC */ +#endif /* ! HAVE_LOCALTIME_R && ! defined (localtime_r) */ + + +#define match_char(ch1, ch2) if (ch1 != ch2) return NULL + +#if defined __GNUC__ && __GNUC__ >= 2 +# define match_string(cs1, s2) \ + ({ size_t len = strlen (cs1); \ + int result = strncasecmp ((cs1), (s2), len) == 0; \ + if (result) (s2) += len; \ + result; }) +#else +/* Oh come on. Get a reasonable compiler. */ +# define match_string(cs1, s2) \ + (strncasecmp ((cs1), (s2), strlen (cs1)) ? 0 : ((s2) += strlen (cs1), 1)) +#endif + +/* We intentionally do not use isdigit() for testing because this will + lead to problems with the wide character version. */ +#define get_number(from, to, n) \ + do { \ + int __n = n; \ + val = 0; \ + while (*rp == ' ') \ + ++rp; \ + if ((*rp < '0') || (*rp > '9')) \ + return NULL; \ + do { \ + val *= 10; \ + val += *rp++ - '0'; \ + } while ((--__n > 0) && (val * 10 <= to) && (*rp >= '0') && (*rp <= '9')); \ + if ((val < from) || (val > to)) \ + return NULL; \ + } while (0) + +#ifdef _NL_CURRENT +# define get_alt_number(from, to, n) \ + ({ \ + __label__ do_normal; \ + if (*decided != raw) \ + { \ + const char *alts = _NL_CURRENT (LC_TIME, ALT_DIGITS); \ + int __n = n; \ + int any = 0; \ + while (*rp == ' ') \ + ++rp; \ + val = 0; \ + do { \ + val *= 10; \ + while (*alts != '\0') \ + { \ + size_t len = strlen (alts); \ + if (strncasecmp (alts, rp, len) == 0) \ + break; \ + alts += len + 1; \ + ++val; \ + } \ + if (*alts == '\0') \ + { \ + if (*decided == not && ! any) \ + goto do_normal; \ + /* If we haven't read anything it's an error. */ \ + if (! any) \ + return NULL; \ + /* Correct the premature multiplication. */ \ + val /= 10; \ + break; \ + } \ + else \ + *decided = loc; \ + } while (--__n > 0 && val * 10 <= to); \ + if (val < from || val > to) \ + return NULL; \ + } \ + else \ + { \ + do_normal: \ + get_number (from, to, n); \ + } \ + 0; \ + }) +#else +# define get_alt_number(from, to, n) \ + /* We don't have the alternate representation. */ \ + get_number(from, to, n) +#endif + +#define recursive(new_fmt) \ + (*(new_fmt) != '\0' \ + && (rp = strptime_internal (rp, (new_fmt), tm, decided, era_cnt)) != NULL) + + +#ifdef _LIBC +/* This is defined in locale/C-time.c in the GNU libc. */ +extern const struct locale_data _nl_C_LC_TIME; +extern const unsigned short int __mon_yday[2][13]; + +# define weekday_name (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (DAY_1)].string) +# define ab_weekday_name \ + (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (ABDAY_1)].string) +# define month_name (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (MON_1)].string) +# define ab_month_name (&_nl_C_LC_TIME.values[_NL_ITEM_INDEX (ABMON_1)].string) +# define HERE_D_T_FMT (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (D_T_FMT)].string) +# define HERE_D_FMT (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (D_FMT)].string) +# define HERE_AM_STR (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (AM_STR)].string) +# define HERE_PM_STR (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (PM_STR)].string) +# define HERE_T_FMT_AMPM \ + (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (T_FMT_AMPM)].string) +# define HERE_T_FMT (_nl_C_LC_TIME.values[_NL_ITEM_INDEX (T_FMT)].string) + +# define strncasecmp(s1, s2, n) __strncasecmp (s1, s2, n) +#else +static char const weekday_name[][10] = + { + "Sunday", "Monday", "Tuesday", "Wednesday", + "Thursday", "Friday", "Saturday" + }; +static char const ab_weekday_name[][4] = + { + "Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat" + }; +static char const month_name[][10] = + { + "January", "February", "March", "April", "May", "June", + "July", "August", "September", "October", "November", "December" + }; +static char const ab_month_name[][4] = + { + "Jan", "Feb", "Mar", "Apr", "May", "Jun", + "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" + }; +# define HERE_D_T_FMT "%a %b %e %H:%M:%S %Y" +# define HERE_D_FMT "%m/%d/%y" +# define HERE_AM_STR "AM" +# define HERE_PM_STR "PM" +# define HERE_T_FMT_AMPM "%I:%M:%S %p" +# define HERE_T_FMT "%H:%M:%S" + +const unsigned short int __mon_yday[2][13] = + { + /* Normal years. */ + { 0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334, 365 }, + /* Leap years. */ + { 0, 31, 60, 91, 121, 152, 182, 213, 244, 274, 305, 335, 366 } + }; +#endif + +/* Status of lookup: do we use the locale data or the raw data? */ +enum locale_status { not, loc, raw }; + + +#ifndef __isleap +/* Nonzero if YEAR is a leap year (every 4 years, + except every 100th isn't, and every 400th is). */ +# define __isleap(year) \ + ((year) % 4 == 0 && ((year) % 100 != 0 || (year) % 400 == 0)) +#endif + +/* Compute the day of the week. */ +static void +day_of_the_week (struct tm *tm) +{ + /* We know that January 1st 1970 was a Thursday (= 4). Compute the + the difference between this data in the one on TM and so determine + the weekday. */ + int corr_year = 1900 + tm->tm_year - (tm->tm_mon < 2); + int wday = (-473 + + (365 * (tm->tm_year - 70)) + + (corr_year / 4) + - ((corr_year / 4) / 25) + ((corr_year / 4) % 25 < 0) + + (((corr_year / 4) / 25) / 4) + + __mon_yday[0][tm->tm_mon] + + tm->tm_mday - 1); + tm->tm_wday = ((wday % 7) + 7) % 7; +} + +/* Compute the day of the year. */ +static void +day_of_the_year (struct tm *tm) +{ + tm->tm_yday = (__mon_yday[__isleap (1900 + tm->tm_year)][tm->tm_mon] + + (tm->tm_mday - 1)); +} + +static char * +#ifdef _LIBC +internal_function +#endif +strptime_internal __P ((const char *rp, const char *fmt, struct tm *tm, + enum locale_status *decided, int era_cnt)); + +static char * +#ifdef _LIBC +internal_function +#endif +strptime_internal (rp, fmt, tm, decided, era_cnt) + const char *rp; + const char *fmt; + struct tm *tm; + enum locale_status *decided; + int era_cnt; +{ + const char *rp_backup; + int cnt; + size_t val; + int have_I, is_pm; + int century, want_century; + int want_era; + int have_wday, want_xday; + int have_yday; + int have_mon, have_mday; +#ifdef _NL_CURRENT + size_t num_eras; +#endif + struct era_entry *era; + + have_I = is_pm = 0; + century = -1; + want_century = 0; + want_era = 0; + era = NULL; + + have_wday = want_xday = have_yday = have_mon = have_mday = 0; + + while (*fmt != '\0') + { + /* A white space in the format string matches 0 more or white + space in the input string. */ + if (isspace (*fmt)) + { + while (isspace (*rp)) + ++rp; + ++fmt; + continue; + } + + /* Any character but `%' must be matched by the same character + in the iput string. */ + if (*fmt != '%') + { + match_char (*fmt++, *rp++); + continue; + } + + ++fmt; +#ifndef _NL_CURRENT + /* We need this for handling the `E' modifier. */ + start_over: +#endif + + /* Make back up of current processing pointer. */ + rp_backup = rp; + + switch (*fmt++) + { + case '%': + /* Match the `%' character itself. */ + match_char ('%', *rp++); + break; + case 'a': + case 'A': + /* Match day of week. */ + for (cnt = 0; cnt < 7; ++cnt) + { +#ifdef _NL_CURRENT + if (*decided !=raw) + { + if (match_string (_NL_CURRENT (LC_TIME, DAY_1 + cnt), rp)) + { + if (*decided == not + && strcmp (_NL_CURRENT (LC_TIME, DAY_1 + cnt), + weekday_name[cnt])) + *decided = loc; + break; + } + if (match_string (_NL_CURRENT (LC_TIME, ABDAY_1 + cnt), rp)) + { + if (*decided == not + && strcmp (_NL_CURRENT (LC_TIME, ABDAY_1 + cnt), + ab_weekday_name[cnt])) + *decided = loc; + break; + } + } +#endif + if (*decided != loc + && (match_string (weekday_name[cnt], rp) + || match_string (ab_weekday_name[cnt], rp))) + { + *decided = raw; + break; + } + } + if (cnt == 7) + /* Does not match a weekday name. */ + return NULL; + tm->tm_wday = cnt; + have_wday = 1; + break; + case 'b': + case 'B': + case 'h': + /* Match month name. */ + for (cnt = 0; cnt < 12; ++cnt) + { +#ifdef _NL_CURRENT + if (*decided !=raw) + { + if (match_string (_NL_CURRENT (LC_TIME, MON_1 + cnt), rp)) + { + if (*decided == not + && strcmp (_NL_CURRENT (LC_TIME, MON_1 + cnt), + month_name[cnt])) + *decided = loc; + break; + } + if (match_string (_NL_CURRENT (LC_TIME, ABMON_1 + cnt), rp)) + { + if (*decided == not + && strcmp (_NL_CURRENT (LC_TIME, ABMON_1 + cnt), + ab_month_name[cnt])) + *decided = loc; + break; + } + } +#endif + if (match_string (month_name[cnt], rp) + || match_string (ab_month_name[cnt], rp)) + { + *decided = raw; + break; + } + } + if (cnt == 12) + /* Does not match a month name. */ + return NULL; + tm->tm_mon = cnt; + want_xday = 1; + break; + case 'c': + /* Match locale's date and time format. */ +#ifdef _NL_CURRENT + if (*decided != raw) + { + if (!recursive (_NL_CURRENT (LC_TIME, D_T_FMT))) + { + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + if (*decided == not && + strcmp (_NL_CURRENT (LC_TIME, D_T_FMT), HERE_D_T_FMT)) + *decided = loc; + want_xday = 1; + break; + } + *decided = raw; + } +#endif + if (!recursive (HERE_D_T_FMT)) + return NULL; + want_xday = 1; + break; + case 'C': + /* Match century number. */ +#ifdef _NL_CURRENT + match_century: +#endif + get_number (0, 99, 2); + century = val; + want_xday = 1; + break; + case 'd': + case 'e': + /* Match day of month. */ + get_number (1, 31, 2); + tm->tm_mday = val; + have_mday = 1; + want_xday = 1; + break; + case 'F': + if (!recursive ("%Y-%m-%d")) + return NULL; + want_xday = 1; + break; + case 'x': +#ifdef _NL_CURRENT + if (*decided != raw) + { + if (!recursive (_NL_CURRENT (LC_TIME, D_FMT))) + { + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + if (*decided == not + && strcmp (_NL_CURRENT (LC_TIME, D_FMT), HERE_D_FMT)) + *decided = loc; + want_xday = 1; + break; + } + *decided = raw; + } +#endif + /* Fall through. */ + case 'D': + /* Match standard day format. */ + if (!recursive (HERE_D_FMT)) + return NULL; + want_xday = 1; + break; + case 'k': + case 'H': + /* Match hour in 24-hour clock. */ + get_number (0, 23, 2); + tm->tm_hour = val; + have_I = 0; + break; + case 'I': + /* Match hour in 12-hour clock. */ + get_number (1, 12, 2); + tm->tm_hour = val % 12; + have_I = 1; + break; + case 'j': + /* Match day number of year. */ + get_number (1, 366, 3); + tm->tm_yday = val - 1; + have_yday = 1; + break; + case 'm': + /* Match number of month. */ + get_number (1, 12, 2); + tm->tm_mon = val - 1; + have_mon = 1; + want_xday = 1; + break; + case 'M': + /* Match minute. */ + get_number (0, 59, 2); + tm->tm_min = val; + break; + case 'n': + case 't': + /* Match any white space. */ + while (isspace (*rp)) + ++rp; + break; + case 'p': + /* Match locale's equivalent of AM/PM. */ +#ifdef _NL_CURRENT + if (*decided != raw) + { + if (match_string (_NL_CURRENT (LC_TIME, AM_STR), rp)) + { + if (strcmp (_NL_CURRENT (LC_TIME, AM_STR), HERE_AM_STR)) + *decided = loc; + break; + } + if (match_string (_NL_CURRENT (LC_TIME, PM_STR), rp)) + { + if (strcmp (_NL_CURRENT (LC_TIME, PM_STR), HERE_PM_STR)) + *decided = loc; + is_pm = 1; + break; + } + *decided = raw; + } +#endif + if (!match_string (HERE_AM_STR, rp)) + if (match_string (HERE_PM_STR, rp)) + is_pm = 1; + else + return NULL; + break; + case 'r': +#ifdef _NL_CURRENT + if (*decided != raw) + { + if (!recursive (_NL_CURRENT (LC_TIME, T_FMT_AMPM))) + { + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + if (*decided == not && + strcmp (_NL_CURRENT (LC_TIME, T_FMT_AMPM), + HERE_T_FMT_AMPM)) + *decided = loc; + break; + } + *decided = raw; + } +#endif + if (!recursive (HERE_T_FMT_AMPM)) + return NULL; + break; + case 'R': + if (!recursive ("%H:%M")) + return NULL; + break; + case 's': + { + /* The number of seconds may be very high so we cannot use + the `get_number' macro. Instead read the number + character for character and construct the result while + doing this. */ + time_t secs = 0; + if (*rp < '0' || *rp > '9') + /* We need at least one digit. */ + return NULL; + + do + { + secs *= 10; + secs += *rp++ - '0'; + } + while (*rp >= '0' && *rp <= '9'); + + if (localtime_r (&secs, tm) == NULL) + /* Error in function. */ + return NULL; + } + break; + case 'S': + get_number (0, 61, 2); + tm->tm_sec = val; + break; + case 'X': +#ifdef _NL_CURRENT + if (*decided != raw) + { + if (!recursive (_NL_CURRENT (LC_TIME, T_FMT))) + { + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + if (strcmp (_NL_CURRENT (LC_TIME, T_FMT), HERE_T_FMT)) + *decided = loc; + break; + } + *decided = raw; + } +#endif + /* Fall through. */ + case 'T': + if (!recursive (HERE_T_FMT)) + return NULL; + break; + case 'u': + get_number (1, 7, 1); + tm->tm_wday = val % 7; + have_wday = 1; + break; + case 'g': + get_number (0, 99, 2); + /* XXX This cannot determine any field in TM. */ + break; + case 'G': + if (*rp < '0' || *rp > '9') + return NULL; + /* XXX Ignore the number since we would need some more + information to compute a real date. */ + do + ++rp; + while (*rp >= '0' && *rp <= '9'); + break; + case 'U': + case 'V': + case 'W': + get_number (0, 53, 2); + /* XXX This cannot determine any field in TM without some + information. */ + break; + case 'w': + /* Match number of weekday. */ + get_number (0, 6, 1); + tm->tm_wday = val; + have_wday = 1; + break; + case 'y': +#ifdef _NL_CURRENT + match_year_in_century: +#endif + /* Match year within century. */ + get_number (0, 99, 2); + /* The "Year 2000: The Millennium Rollover" paper suggests that + values in the range 69-99 refer to the twentieth century. */ + tm->tm_year = val >= 69 ? val : val + 100; + /* Indicate that we want to use the century, if specified. */ + want_century = 1; + want_xday = 1; + break; + case 'Y': + /* Match year including century number. */ + get_number (0, 9999, 4); + tm->tm_year = val - 1900; + want_century = 0; + want_xday = 1; + break; + case 'Z': + /* XXX How to handle this? */ + break; + case 'E': +#ifdef _NL_CURRENT + switch (*fmt++) + { + case 'c': + /* Match locale's alternate date and time format. */ + if (*decided != raw) + { + const char *fmt = _NL_CURRENT (LC_TIME, ERA_D_T_FMT); + + if (*fmt == '\0') + fmt = _NL_CURRENT (LC_TIME, D_T_FMT); + + if (!recursive (fmt)) + { + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + if (strcmp (fmt, HERE_D_T_FMT)) + *decided = loc; + want_xday = 1; + break; + } + *decided = raw; + } + /* The C locale has no era information, so use the + normal representation. */ + if (!recursive (HERE_D_T_FMT)) + return NULL; + want_xday = 1; + break; + case 'C': + if (*decided != raw) + { + if (era_cnt >= 0) + { + era = _nl_select_era_entry (era_cnt); + if (match_string (era->era_name, rp)) + { + *decided = loc; + break; + } + else + return NULL; + } + else + { + num_eras = _NL_CURRENT_WORD (LC_TIME, + _NL_TIME_ERA_NUM_ENTRIES); + for (era_cnt = 0; era_cnt < (int) num_eras; + ++era_cnt, rp = rp_backup) + { + era = _nl_select_era_entry (era_cnt); + if (match_string (era->era_name, rp)) + { + *decided = loc; + break; + } + } + if (era_cnt == (int) num_eras) + { + era_cnt = -1; + if (*decided == loc) + return NULL; + } + else + break; + } + + *decided = raw; + } + /* The C locale has no era information, so use the + normal representation. */ + goto match_century; + case 'y': + if (*decided == raw) + goto match_year_in_century; + + get_number(0, 9999, 4); + tm->tm_year = val; + want_era = 1; + want_xday = 1; + break; + case 'Y': + if (*decided != raw) + { + num_eras = _NL_CURRENT_WORD (LC_TIME, + _NL_TIME_ERA_NUM_ENTRIES); + for (era_cnt = 0; era_cnt < (int) num_eras; + ++era_cnt, rp = rp_backup) + { + era = _nl_select_era_entry (era_cnt); + if (recursive (era->era_format)) + break; + } + if (era_cnt == (int) num_eras) + { + era_cnt = -1; + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + *decided = loc; + era_cnt = -1; + break; + } + + *decided = raw; + } + get_number (0, 9999, 4); + tm->tm_year = val - 1900; + want_century = 0; + want_xday = 1; + break; + case 'x': + if (*decided != raw) + { + const char *fmt = _NL_CURRENT (LC_TIME, ERA_D_FMT); + + if (*fmt == '\0') + fmt = _NL_CURRENT (LC_TIME, D_FMT); + + if (!recursive (fmt)) + { + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + if (strcmp (fmt, HERE_D_FMT)) + *decided = loc; + break; + } + *decided = raw; + } + if (!recursive (HERE_D_FMT)) + return NULL; + break; + case 'X': + if (*decided != raw) + { + const char *fmt = _NL_CURRENT (LC_TIME, ERA_T_FMT); + + if (*fmt == '\0') + fmt = _NL_CURRENT (LC_TIME, T_FMT); + + if (!recursive (fmt)) + { + if (*decided == loc) + return NULL; + else + rp = rp_backup; + } + else + { + if (strcmp (fmt, HERE_T_FMT)) + *decided = loc; + break; + } + *decided = raw; + } + if (!recursive (HERE_T_FMT)) + return NULL; + break; + default: + return NULL; + } + break; +#else + /* We have no information about the era format. Just use + the normal format. */ + if (*fmt != 'c' && *fmt != 'C' && *fmt != 'y' && *fmt != 'Y' + && *fmt != 'x' && *fmt != 'X') + /* This is an illegal format. */ + return NULL; + + goto start_over; +#endif + case 'O': + switch (*fmt++) + { + case 'd': + case 'e': + /* Match day of month using alternate numeric symbols. */ + get_alt_number (1, 31, 2); + tm->tm_mday = val; + have_mday = 1; + want_xday = 1; + break; + case 'H': + /* Match hour in 24-hour clock using alternate numeric + symbols. */ + get_alt_number (0, 23, 2); + tm->tm_hour = val; + have_I = 0; + break; + case 'I': + /* Match hour in 12-hour clock using alternate numeric + symbols. */ + get_alt_number (1, 12, 2); + tm->tm_hour = val - 1; + have_I = 1; + break; + case 'm': + /* Match month using alternate numeric symbols. */ + get_alt_number (1, 12, 2); + tm->tm_mon = val - 1; + have_mon = 1; + want_xday = 1; + break; + case 'M': + /* Match minutes using alternate numeric symbols. */ + get_alt_number (0, 59, 2); + tm->tm_min = val; + break; + case 'S': + /* Match seconds using alternate numeric symbols. */ + get_alt_number (0, 61, 2); + tm->tm_sec = val; + break; + case 'U': + case 'V': + case 'W': + get_alt_number (0, 53, 2); + /* XXX This cannot determine any field in TM without + further information. */ + break; + case 'w': + /* Match number of weekday using alternate numeric symbols. */ + get_alt_number (0, 6, 1); + tm->tm_wday = val; + have_wday = 1; + break; + case 'y': + /* Match year within century using alternate numeric symbols. */ + get_alt_number (0, 99, 2); + tm->tm_year = val >= 69 ? val : val + 100; + want_xday = 1; + break; + default: + return NULL; + } + break; + default: + return NULL; + } + } + + if (have_I && is_pm) + tm->tm_hour += 12; + + if (century != -1) + { + if (want_century) + tm->tm_year = tm->tm_year % 100 + (century - 19) * 100; + else + /* Only the century, but not the year. Strange, but so be it. */ + tm->tm_year = (century - 19) * 100; + } + +#ifdef _NL_CURRENT + if (era_cnt != -1) + { + era = _nl_select_era_entry(era_cnt); + if (want_era) + tm->tm_year = (era->start_date[0] + + ((tm->tm_year - era->offset) + * era->absolute_direction)); + else + /* Era start year assumed. */ + tm->tm_year = era->start_date[0]; + } + else +#endif + if (want_era) + return NULL; + + if (want_xday && !have_wday) + { + if ( !(have_mon && have_mday) && have_yday) + { + /* We don't have tm_mon and/or tm_mday, compute them. */ + int t_mon = 0; + while (__mon_yday[__isleap(1900 + tm->tm_year)][t_mon] <= tm->tm_yday) + t_mon++; + if (!have_mon) + tm->tm_mon = t_mon - 1; + if (!have_mday) + tm->tm_mday = + (tm->tm_yday + - __mon_yday[__isleap(1900 + tm->tm_year)][t_mon - 1] + 1); + } + day_of_the_week (tm); + } + if (want_xday && !have_yday) + day_of_the_year (tm); + + return (char *) rp; +} + + +char * +msn_strptime (buf, format, tm) + const char *buf; + const char *format; + struct tm *tm; +{ + enum locale_status decided; + +#ifdef _NL_CURRENT + decided = not; +#else + decided = raw; +#endif + return strptime_internal (buf, format, tm, &decided, -1); +} +#else +#define msn_strptime strptime +#endif
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/libpurple/protocols/msn/msnutils.h Sun Apr 15 05:01:55 2007 +0000 @@ -0,0 +1,64 @@ +/** + * @file msnutils.h Utility functions + * + * purple + * + * Purple is the legal property of its developers, whose names are too numerous + * to list here. Please refer to the COPYRIGHT file distributed with this + * source distribution. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ +#ifndef _MSN_UTILS_H_ +#define _MSN_UTILS_H_ + +/*encode the str to RFC2047 style*/ +char * msn_encode_mime(const char *str); + +/** + * Generate the Random GUID + */ +char * rand_guid(void); + +/** + * Parses the MSN message formatting into a format compatible with Purple. + * + * @param mime The mime header with the formatting. + * @param pre_ret The returned prefix string. + * @param post_ret The returned postfix string. + * + * @return The new message. + */ +void msn_parse_format(const char *mime, char **pre_ret, char **post_ret); + +/** + * Parses the Purple message formatting (html) into the MSN format. + * + * @param html The html message to format. + * @param attributes The returned attributes string. + * @param message The returned message string. + * + * @return The new message. + */ +void msn_import_html(const char *html, char **attributes, char **message); + +void msn_parse_socket(const char *str, char **ret_host, int *ret_port); +void msn_handle_chl(char *input, char *output); +int isBigEndian(void); +unsigned int swapInt(unsigned int dw); +char * msn_strptime (const char *buf,const char *format,struct tm *tm); + + +#endif /* _MSN_UTILS_H_ */
--- a/libpurple/protocols/msn/notification.c Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/notification.c Sun Apr 15 05:01:55 2007 +0000 @@ -311,17 +311,17 @@ g_snprintf(proto_str, sizeof(proto_str), "MSNP%d", session->protocol_ver); - for (i = 1; i < cmd->param_count; i++) + for (i = 1; i < cmd->param_count -1; i++) { - if (!strcmp(cmd->params[i], proto_str)) + purple_debug_info("MaYuan","%s,proto_str:%s\n",cmd->params[i],proto_str); + if (strcmp(cmd->params[i], proto_str) >= 0) { protocol_supported = TRUE; break; } } - if (!protocol_supported) - { + if (!protocol_supported){ msn_session_set_error(session, MSN_ERROR_UNSUPPORTED_PROTOCOL, NULL); return; @@ -906,7 +906,7 @@ if (session->protocol_ver >= 9 && cmd->param_count == 8) { - msnobj = msn_object_new_from_string(purple_url_decode(cmd->params[5])); + msnobj = msn_object_new_from_string(purple_url_decode(cmd->params[6])); msn_user_set_object(user, msnobj); } @@ -963,8 +963,7 @@ { if (cmd->param_count == 7) { - msnobj = - msn_object_new_from_string(purple_url_decode(cmd->params[4])); + msnobj = msn_object_new_from_string(purple_url_decode(cmd->params[5])); msn_user_set_object(user, msnobj); } else @@ -1443,7 +1442,7 @@ PurpleConnection *gc; MsnUser *user; const char *passport; - char *psm_str; + char *psm_str, *currentmedia_str; /*get the payload content*/ // purple_debug_info("MaYuan","UBX {%s} payload{%s}\n",cmd->params[0], cmd->payload); @@ -1456,10 +1455,15 @@ user = msn_userlist_find_user(session->userlist, passport); psm_str = msn_get_psm(cmd->payload,len); + currentmedia_str = msn_parse_currentmedia( + msn_get_currentmedia(cmd->payload, len)); + msn_user_set_statusline(user, psm_str); + msn_user_set_currentmedia(user, currentmedia_str); msn_user_update(user); g_free(psm_str); + g_free(currentmedia_str); } static void @@ -1496,16 +1500,14 @@ /* This isn't an official message. */ return; - if ((value = msn_message_get_attr(msg, "kv")) != NULL) - { + if ((value = msn_message_get_attr(msg, "kv")) != NULL){ if (session->passport_info.kv != NULL) g_free(session->passport_info.kv); session->passport_info.kv = g_strdup(value); } - if ((value = msn_message_get_attr(msg, "sid")) != NULL) - { + if ((value = msn_message_get_attr(msg, "sid")) != NULL){ if (session->passport_info.sid != NULL) g_free(session->passport_info.sid); @@ -1530,9 +1532,9 @@ } if ((value = msn_message_get_attr(msg, "ClientPort")) != NULL) -{ - session->passport_info.client_port = ntohs(atoi(value)); -} + { + session->passport_info.client_port = ntohs(atoi(value)); + } if ((value = msn_message_get_attr(msg, "LoginTime")) != NULL) session->passport_info.sl = atol(value); @@ -1743,7 +1745,7 @@ { case 1: minutes = atoi(g_hash_table_lookup(table, "Arg1")); - g_snprintf(buf, sizeof(buf), ngettext( + g_snprintf(buf, sizeof(buf), dngettext(PACKAGE, "The MSN server will shut down for maintenance " "in %d minute. You will automatically be " "signed out at that time. Please finish any "
--- a/libpurple/protocols/msn/session.c Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/session.c Sun Apr 15 05:01:55 2007 +0000 @@ -273,10 +273,8 @@ /* The core used to use msn_add_buddy to add all buddies before * being logged in. This no longer happens, so we manually iterate * over the whole buddy list to identify sync issues. */ - - for (gnode = purple_blist_get_root(); gnode; gnode = gnode->next) { - PurpleGroup *group = (PurpleGroup *)gnode; - const char *group_name = group->name; + for (gnode = purple_get_blist()->root; gnode; gnode = gnode->next) + { if(!PURPLE_BLIST_NODE_IS_GROUP(gnode)) continue; for(cnode = gnode->child; cnode; cnode = cnode->next) {
--- a/libpurple/protocols/msn/state.c Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/state.c Sun Apr 15 05:01:55 2007 +0000 @@ -81,6 +81,102 @@ return result; } +/* parse CurrentMedia string */ +char * +msn_parse_currentmedia(const char *cmedia) +{ + char **cmedia_array; + char *buffer=NULL, *inptr, *outptr, *tmpptr; + int length, strings, tmp; + + purple_debug_info("msn", "Parsing currentmedia string: \"%s\"\n", cmedia); + if( (cmedia == NULL) || (!strcmp(cmedia, ""))) { + purple_debug_info("msn", "No currentmedia string\n"); + return NULL; + } + + cmedia_array=g_strsplit(cmedia, "\\0", 0); + + strings=1; /* Skip first empty string */ + length=5; /* Space for '\0' (1 byte) and prefix (4 bytes) */ + while(strcmp(cmedia_array[strings], "")) { + length+= strlen(cmedia_array[strings]); + strings++; + } + + if((strings>3) && (!strcmp(cmedia_array[2], "1"))) { /* Check if enabled */ + + buffer=g_malloc(length); + + inptr=cmedia_array[3]; + outptr=buffer; + + if(!strcmp(cmedia_array[1], "Music")) { + strcpy(outptr, "np. "); + outptr+=4; + }/* else if(!strcmp(cmedia_array[1], "Games")) { + } else if(!strcmp(cmedia_array[1], "Office")) { + }*/ + + while(*inptr!='\0') { + if((*inptr == '{') && (strlen(inptr) > 2) && (*(inptr+2) == '}') ) { + errno = 0; + tmp = strtol(inptr+1,&tmpptr,10); + if( (errno!=0) || (tmpptr == (inptr+1)) || + ((tmp+5)>(strings)) ) { + *outptr = *inptr; /* Conversion not successful */ + outptr++; + } else { + /* Replace {?} tag with appropriate text */ + strcpy(outptr, cmedia_array[tmp+4]); + outptr+=strlen(cmedia_array[tmp+4]); + inptr+=2; + } + } else { + *outptr = *inptr; + outptr++; + } + inptr++; + } + *outptr='\0'; + purple_debug_info("msn", "Parsed currentmedia string, result: \"%s\"\n", + buffer); + } else { + purple_debug_info("msn", "Current media marked disabled, not parsing\n"); + } + + g_strfreev(cmedia_array); + return buffer; +} + +/* get the CurrentMedia info from the XML string */ +char * +msn_get_currentmedia(char *xml_str, gsize len) +{ + xmlnode *payloadNode, *currentmediaNode; + char *currentmedia_str, *currentmedia; + + purple_debug_info("msn","msn get CurrentMedia\n"); + payloadNode = xmlnode_from_str(xml_str, len); + if (!payloadNode){ + purple_debug_error("msn","PSM XML parse Error!\n"); + return NULL; + } + currentmediaNode = xmlnode_get_child(payloadNode, "CurrentMedia"); + if (currentmediaNode == NULL){ + purple_debug_info("msn","No CurrentMedia Node"); + g_free(payloadNode); + return NULL; + } + currentmedia_str = xmlnode_get_data(currentmediaNode); + currentmedia = g_strdup(currentmedia_str); + + g_free(currentmediaNode); + g_free(payloadNode); + + return currentmedia; +} + /*get the PSM info from the XML string*/ char * msn_get_psm(char *xml_str, gsize len)
--- a/libpurple/protocols/msn/state.h Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/state.h Sun Apr 15 05:01:55 2007 +0000 @@ -61,6 +61,12 @@ void msn_set_psm(MsnSession *session); +/* Parse CurrentMedia string */ +char * msn_parse_currentmedia(const char *cmedia); + +/* Get the CurrentMedia info from the XML string */ +char * msn_get_currentmedia(char *xml_str,gsize len); + /*get the PSM info from the XML string*/ char * msn_get_psm(char *xml_str,gsize len);
--- a/libpurple/protocols/msn/user.c Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/user.c Sun Apr 15 05:01:55 2007 +0000 @@ -89,19 +89,31 @@ account = user->userlist->session->account; - if (user->statusline != NULL) { + if (user->statusline != NULL && user->currentmedia != NULL) { + purple_prpl_got_user_status(account, user->passport, user->status, + "message", user->statusline, + "currentmedia", user->currentmedia, NULL); + } else if (user->currentmedia != NULL) { + purple_prpl_got_user_status(account, user->passport, "currentmedia", + user->currentmedia, NULL); + } else if (user->statusline != NULL) { //char *status = g_strdup_printf("%s - %s", user->status, user->statusline); - purple_prpl_got_user_status(account, user->passport, user->status, "message", user->statusline, NULL); - } - else if (user->status != NULL) { - purple_prpl_got_user_status(account, user->passport, user->status, NULL); + purple_prpl_got_user_status(account, user->passport, user->status, + "message", user->statusline, NULL); + } else if (user->status != NULL) { + if (!strcmp(user->status, "offline") && user->mobile) { + purple_prpl_got_user_status(account, user->passport, "available", NULL); + purple_prpl_got_user_status(account, user->passport, "mobile", NULL); + } else { + purple_prpl_got_user_status(account, user->passport, user->status, NULL); + purple_prpl_got_user_status_deactive(account, user->passport, "mobile"); + } } - if (user->idle){ + if (user->idle) purple_prpl_got_user_idle(account, user->passport, TRUE, -1); - }else{ + else purple_prpl_got_user_idle(account, user->passport, FALSE, 0); - } } void @@ -158,6 +170,15 @@ } void +msn_user_set_currentmedia(MsnUser *user, const char *currentmedia) +{ + g_return_if_fail(user != NULL); + + g_free(user->currentmedia); + user->currentmedia = g_strdup(currentmedia); +} + +void msn_user_set_store_name(MsnUser *user, const char *name) { g_return_if_fail(user != NULL);
--- a/libpurple/protocols/msn/user.h Sun Apr 15 04:59:43 2007 +0000 +++ b/libpurple/protocols/msn/user.h Sun Apr 15 05:01:55 2007 +0000 @@ -48,7 +48,8 @@ char * uid; /*< User Id */ const char *status; /**< The state of the user. */ - char *statusline; /**< The state of the user. */ + char *statusline; /**< The state of the user. */ + char *currentmedia; /**< The current media of the user. */ gboolean idle; /**< The idle state of the user. */ @@ -115,6 +116,14 @@ */ void msn_user_set_statusline(MsnUser *user, const char *statusline); + /** + * Sets the current media of user. + * + * @param user The user. + * @param state The statusline string. + */ +void msn_user_set_currentmedia(MsnUser *user, const char *currentmedia); + /** * Sets the new state of user. *