webcit/html2html.c

   1 /*
   2  * $Id$
   3  */
   4 /**
   5  * \defgroup HTML2HTML Output an HTML message, modifying it slightly to make sure it plays nice
   6  * with the rest of our web framework.
   7  * \ingroup WebcitHttpServer
   8  */
   9 /*@{*/
  10 #include "webcit.h"
  11 #include "vcard.h"
  12 #include "webserver.h"
  13
  14
  15 /**
  16  * \brief       Strip surrounding single or double quotes from a string.
  17  *
  18  * \param s     String to be stripped.
  19  */
  20 void stripquotes(char *s)
  21 {
  22         int len;
  23
  24         if (!s) return;
  25
  26         len = strlen(s);
  27         if (len < 2) return;
  28
  29         if ( ( (s[0] == '\"') && (s[len-1] == '\"') ) || ( (s[0] == '\'') && (s[len-1] == '\'') ) ) {
  30                 s[len-1] = 0;
  31                 strcpy(s, &s[1]);
  32         }
  33 }
  34
  35
  36 /**
  37  * \brief Check to see if a META tag has overridden the declared MIME character set.
  38  *
  39  * \param charset               Character set name (left unchanged if we don't do anything)
  40  * \param meta_http_equiv       Content of the "http-equiv" portion of the META tag
  41  * \param meta_content          Content of the "content" portion of the META tag
  42  */
  43 void extract_charset_from_meta(char *charset, char *meta_http_equiv, char *meta_content)
  44 {
  45         char *ptr;
  46         char buf[64];
  47
  48         if (!charset) return;
  49         if (!meta_http_equiv) return;
  50         if (!meta_content) return;
  51
  52
  53         if (strcasecmp(meta_http_equiv, "Content-type")) return;
  54
  55         ptr = strchr(meta_content, ';');
  56         if (!ptr) return;
  57
  58         safestrncpy(buf, ++ptr, sizeof buf);
  59         striplt(buf);
  60         if (!strncasecmp(buf, "charset=", 8)) {
  61                 strcpy(charset, &buf[8]);
  62         }
  63 }
  64
  65
  66
  67 /**
  68  * \brief Sanitize and enhance an HTML message for display.
  69  *        Also convert weird character sets to UTF-8 if necessary.
  70  *
  71  * \param supplied_charset the input charset as declared in the MIME headers
  72  */
  73 void output_html(char *supplied_charset, int treat_as_wiki) {
  74         char buf[SIZ];
  75         char *msg;
  76         char *ptr;
  77         char *msgstart;
  78         char *msgend;
  79         char *converted_msg;
  80         int buffer_length = 1;
  81         int line_length = 0;
  82         int content_length = 0;
  83         int output_length = 0;
  84         char new_window[SIZ];
  85         int brak = 0;
  86         int alevel = 0;
  87         int i;
  88         int linklen;
  89         char charset[128];
  90 #ifdef HAVE_ICONV
  91         iconv_t ic = (iconv_t)(-1) ;
  92         char *ibuf;                   /**< Buffer of characters to be converted */
  93         char *obuf;                   /**< Buffer for converted characters      */
  94         size_t ibuflen;               /**< Length of input buffer               */
  95         size_t obuflen;               /**< Length of output buffer              */
  96         char *osav;                   /**< Saved pointer to output buffer       */
  97 #endif
  98
  99         safestrncpy(charset, supplied_charset, sizeof charset);
 100         msg = strdup("");
 101         sprintf(new_window, "<a target=\"%s\" href=", TARGET);
 102
 103         while (serv_getln(buf, sizeof buf), strcmp(buf, "000")) {
 104                 line_length = strlen(buf);
 105                 buffer_length = content_length + line_length + 2;
 106                 msg = realloc(msg, buffer_length);
 107                 if (msg == NULL) {
 108                         wprintf("<b>");
 109                         wprintf(_("realloc() error! couldn't get %d bytes: %s"),
 110                                 buffer_length + 1,
 111                                 strerror(errno));
 112                         wprintf("</b><br /><br />\n");
 113                         return;
 114                 }
 115                 strcpy(&msg[content_length], buf);
 116                 content_length += line_length;
 117                 strcpy(&msg[content_length], "\n");
 118                 content_length += 1;
 119         }
 120
 121         /** Do a first pass to isolate the message body */
 122         ptr = msg;
 123         msgstart = msg;
 124         msgend = &msg[content_length];
 125
 126         while (ptr < msgend) {
 127
 128                 /** Advance to next tag */
 129                 ptr = strchr(ptr, '<');
 130                 if ((ptr == NULL) || (ptr >= msgend)) break;
 131                 ++ptr;
 132                 if ((ptr == NULL) || (ptr >= msgend)) break;
 133
 134                 /**
 135                  *  Look for META tags.  Some messages (particularly in
 136                  *  Asian locales) illegally declare a message's character
 137                  *  set in the HTML instead of in the MIME headers.  This
 138                  *  is wrong but we have to work around it anyway.
 139                  */
 140                 if (!strncasecmp(ptr, "META", 4)) {
 141
 142                         char *meta_start;
 143                         char *meta_end;
 144                         int meta_length;
 145                         char *meta;
 146                         char *meta_http_equiv;
 147                         char *meta_content;
 148                         char *spaceptr;
 149
 150                         meta_start = &ptr[4];
 151                         meta_end = strchr(ptr, '>');
 152                         if ((meta_end != NULL) && (meta_end <= msgend)) {
 153                                 meta_length = meta_end - meta_start + 1;
 154                                 meta = malloc(meta_length + 1);
 155                                 safestrncpy(meta, meta_start, meta_length);
 156                                 meta[meta_length] = 0;
 157                                 striplt(meta);
 158                                 if (!strncasecmp(meta, "HTTP-EQUIV=", 11)) {
 159                                         meta_http_equiv = strdup(&meta[11]);
 160                                         spaceptr = strchr(meta_http_equiv, ' ');
 161                                         if (spaceptr != NULL) {
 162                                                 *spaceptr = 0;
 163                                                 meta_content = strdup(++spaceptr);
 164                                                 if (!strncasecmp(meta_content, "content=", 8)) {
 165                                                         strcpy(meta_content, &meta_content[8]);
 166                                                         stripquotes(meta_http_equiv);
 167                                                         stripquotes(meta_content);
 168                                                         extract_charset_from_meta(charset,
 169                                                                 meta_http_equiv, meta_content);
 170                                                 }
 171                                                 free(meta_content);
 172                                         }
 173                                         free(meta_http_equiv);
 174                                 }
 175                                 free(meta);
 176                         }
 177                 }
 178
 179                 /**
 180                  * Any of these tags cause everything up to and including
 181                  * the tag to be removed.
 182                  */
 183                 if ( (!strncasecmp(ptr, "HTML", 4))
 184                    ||(!strncasecmp(ptr, "HEAD", 4))
 185                    ||(!strncasecmp(ptr, "/HEAD", 5))
 186                    ||(!strncasecmp(ptr, "BODY", 4)) ) {
 187                         ptr = strchr(ptr, '>');
 188                         if ((ptr == NULL) || (ptr >= msgend)) break;
 189                         ++ptr;
 190                         if ((ptr == NULL) || (ptr >= msgend)) break;
 191                         msgstart = ptr;
 192                 }
 193
 194                 /**
 195                  * Any of these tags cause everything including and following
 196                  * the tag to be removed.
 197                  */
 198                 if ( (!strncasecmp(ptr, "/HTML", 5))
 199                    ||(!strncasecmp(ptr, "/BODY", 5)) ) {
 200                         --ptr;
 201                         msgend = ptr;
 202                         strcpy(ptr, "");
 203
 204                 }
 205
 206                 ++ptr;
 207         }
 208         if (msgstart > msg) {
 209                 strcpy(msg, msgstart);
 210         }
 211
 212         /** Convert foreign character sets to UTF-8 if necessary. */
 213 #ifdef HAVE_ICONV
 214         if ( (strcasecmp(charset, "us-ascii"))
 215            && (strcasecmp(charset, "UTF-8"))
 216            && (strcasecmp(charset, ""))
 217         ) {
 218                 lprintf(9, "Converting %s to UTF-8\n", charset);
 219                 ic = iconv_open("UTF-8", charset);
 220                 if (ic == (iconv_t)(-1) ) {
 221                         lprintf(5, "%s:%d iconv_open() failed: %s\n",
 222                                 __FILE__, __LINE__, strerror(errno));
 223                 }
 224         }
 225         if (ic != (iconv_t)(-1) ) {
 226                 ibuf = msg;
 227                 ibuflen = content_length;
 228                 obuflen = content_length + (content_length / 2) ;
 229                 obuf = (char *) malloc(obuflen);
 230                 osav = obuf;
 231                 iconv(ic, &ibuf, &ibuflen, &obuf, &obuflen);
 232                 content_length = content_length + (content_length / 2) - obuflen;
 233                 osav[content_length] = 0;
 234                 free(msg);
 235                 msg = osav;
 236                 iconv_close(ic);
 237         }
 238 #endif
 239
 240         /** FIXME At this point, shigerugo's messages are still clean.
 241          *        Figure out what is mangling them below.
 242          */
 243
 244         /** Now go through the message, parsing tags as necessary. */
 245         converted_msg = malloc(content_length);
 246         strcpy(converted_msg, "");
 247         ptr = msgstart;
 248         while (ptr < msgend) {
 249                 /**
 250                  * Change mailto: links to WebCit mail, by replacing the
 251                  * link with one that points back to our mail room.  Due to
 252                  * the way we parse URL's, it'll even handle mailto: links
 253                  * that have "?subject=" in them.
 254                  */
 255                 if (!strncasecmp(ptr, "<a href=\"mailto:", 16)) {
 256                         content_length += 64;
 257                         converted_msg = realloc(converted_msg, content_length);
 258                         sprintf(&converted_msg[output_length],
 259                                 "<a href=\"display_enter"
 260                                 "?force_room=_MAIL_&recp=");
 261                         output_length += 47;
 262                         ptr = &ptr[16];
 263                         ++alevel;
 264                 }
 265                 /** Make external links open in a separate window */
 266                 else if (!strncasecmp(ptr, "<a href=", 8)) {
 267                         ++alevel;
 268                         if ( ((strchr(ptr, ':') < strchr(ptr, '/')))
 269                              &&  ((strchr(ptr, '/') < strchr(ptr, '>')))
 270                              ) {
 271                                 /* open external links to new window */
 272                                 content_length += 64;
 273                                 converted_msg = realloc(converted_msg, content_length);
 274                                 sprintf(&converted_msg[output_length], new_window);
 275                                 output_length += strlen(new_window);
 276                                 ptr = &ptr[8];
 277                         }
 278                         else if ( (treat_as_wiki) && (strncasecmp(ptr, "<a href=\"wiki?", 14)) ) {
 279                                 lprintf(9, "converting wiki link\n");
 280                                 content_length += 64;
 281                                 converted_msg = realloc(converted_msg, content_length);
 282                                 sprintf(&converted_msg[output_length], "<a href=\"wiki?page=");
 283                                 output_length += 19;
 284                                 ptr = &ptr[9];
 285                         }
 286                         else {
 287                                 sprintf(&converted_msg[output_length], "<a href=");
 288                                 output_length += 8;
 289                                 ptr = &ptr[8];
 290                         }
 291                 }
 292                 /**
 293                  * Turn anything that looks like a URL into a real link, as long
 294                  * as it's not inside a tag already
 295                  */
 296                 else if ( (brak == 0) && (alevel == 0)
 297                      && (!strncasecmp(ptr, "http://", 7))) {
 298                                 linklen = 0;
 299                                 /** Find the end of the link */
 300                                 for (i=0; i<=strlen(ptr); ++i) {
 301                                         if ((ptr[i]==0)
 302                                            ||(isspace(ptr[i]))
 303                                            ||(ptr[i]==10)
 304                                            ||(ptr[i]==13)
 305                                            ||(ptr[i]=='(')
 306                                            ||(ptr[i]==')')
 307                                            ||(ptr[i]=='<')
 308                                            ||(ptr[i]=='>')
 309                                            ||(ptr[i]=='[')
 310                                            ||(ptr[i]==']')
 311                                         ) linklen = i;
 312                                         if (linklen > 0) break;
 313                                 }
 314                                 if (linklen > 0) {
 315                                         content_length += (32 + linklen);
 316                                         converted_msg = realloc(converted_msg, content_length);
 317                                         sprintf(&converted_msg[output_length], new_window);
 318                                         output_length += strlen(new_window);
 319                                         converted_msg[output_length] = '\"';
 320                                         converted_msg[++output_length] = 0;
 321                                         for (i=0; i<linklen; ++i) {
 322                                                 converted_msg[output_length] = ptr[i];
 323                                                 converted_msg[++output_length] = 0;
 324                                         }
 325                                         sprintf(&converted_msg[output_length], "\">");
 326                                         output_length += 2;
 327                                         for (i=0; i<linklen; ++i) {
 328                                                 converted_msg[output_length] = *ptr++;
 329                                                 converted_msg[++output_length] = 0;
 330                                         }
 331                                         sprintf(&converted_msg[output_length], "</A>");
 332                                         output_length += 4;
 333                                 }
 334                 }
 335                 else {
 336                         /**
 337                          * We need to know when we're inside a tag,
 338                          * so we don't turn things that look like URL's into
 339                          * links, when they're already links - or image sources.
 340                          */
 341                         if (*ptr == '<') ++brak;
 342                         if (*ptr == '>') --brak;
 343                         if (!strncasecmp(ptr, "</A>", 3)) --alevel;
 344                         converted_msg[output_length] = *ptr++;
 345                         converted_msg[++output_length] = 0;
 346                 }
 347         }
 348
 349         /** Output our big pile of markup */
 350         client_write(converted_msg, output_length);
 351
 352         /** A little trailing vertical whitespace... */
 353         wprintf("<br /><br />\n");
 354
 355         /** Now give back the memory */
 356         free(converted_msg);
 357         free(msg);
 358 }
 359
 360 /*@}*/