filter.c

   1 /*@ S-nail - a mail user agent derived from Berkeley Mail.
   2  *@ Filter objects.
   3  *
   4  * Copyright (c) 2013 - 2015 Steffen (Daode) Nurpmeso <sdaoden@users.sf.net>.
   5  *
   6  * Permission to use, copy, modify, and/or distribute this software for any
   7  * purpose with or without fee is hereby granted, provided that the above
   8  * copyright notice and this permission notice appear in all copies.
   9  *
  10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  17  */
  18 #undef n_FILE
  19 #define n_FILE filter
  20
  21 #ifndef HAVE_AMALGAMATION
  22 # include "nail.h"
  23 #endif
  24
  25 /*
  26  * Quotation filter
  27  */
  28
  29 /*
  30  * TODO quotation filter: anticipate in future data: don't break if only WS
  31  * TODO or a LF escaping \ follows on the line (simply reuse the latter).
  32  */
  33
  34 #ifdef HAVE_QUOTE_FOLD
  35 CTA(QUOTE_MAX > 3);
  36
  37 enum qf_state {
  38    _QF_CLEAN,
  39    _QF_PREFIX,
  40    _QF_DATA
  41 };
  42
  43 struct qf_vc {
  44    struct quoteflt   *self;
  45    char const        *buf;
  46    size_t            len;
  47 };
  48
  49 /* Print out prefix and current quote */
  50 static ssize_t _qf_dump_prefix(struct quoteflt *self);
  51
  52 /* Add one data character */
  53 static ssize_t _qf_add_data(struct quoteflt *self, wchar_t wc);
  54
  55 /* State machine handlers */
  56 static ssize_t _qf_state_prefix(struct qf_vc *vc);
  57 static ssize_t _qf_state_data(struct qf_vc *vc);
  58
  59 static ssize_t
  60 _qf_dump_prefix(struct quoteflt *self)
  61 {
  62    ssize_t rv;
  63    size_t i;
  64    NYD_ENTER;
  65
  66    if ((i = self->qf_pfix_len) > 0 && i != fwrite(self->qf_pfix, 1, i,
  67          self->qf_os))
  68       goto jerr;
  69    rv = i;
  70
  71    if ((i = self->qf_currq.l) > 0 && i != fwrite(self->qf_currq.s, 1, i,
  72          self->qf_os))
  73       goto jerr;
  74    rv += i;
  75 jleave:
  76    NYD_LEAVE;
  77    return rv;
  78 jerr:
  79    rv = -1;
  80    goto jleave;
  81 }
  82
  83 static ssize_t
  84 _qf_add_data(struct quoteflt *self, wchar_t wc)
  85 {
  86    char *save_b;
  87    ui32_t save_l, save_w;
  88    ssize_t rv = 0;
  89    int w, l;
  90    NYD_ENTER;
  91
  92    save_l = save_w = 0; /* silence cc */
  93    save_b = NULL;
  94    /* <newline> ends state */
  95    if (wc == L'\n')
  96       goto jflush;
  97    if (wc == L'\r') /* TODO CR should be stripped in lower level!! */
  98       goto jleave;
  99
 100    /* Unroll <tab> to spaces */
 101    if (wc == L'\t') {
 102       save_l = self->qf_datw;
 103       save_w = (save_l + QUOTE_TAB_SPACES) & ~(QUOTE_TAB_SPACES - 1);
 104       save_w -= save_l;
 105       while (save_w-- > 0) {
 106          ssize_t j = _qf_add_data(self, L' ');
 107          if (j < 0) {
 108             rv = j;
 109             break;
 110          }
 111          rv += j;
 112       }
 113       goto jleave;
 114    }
 115
 116    w = wcwidth(wc);
 117    if (w == -1) {
 118 jbad:
 119       ++self->qf_datw;
 120       self->qf_dat.s[self->qf_dat.l++] = '?';
 121    } else {
 122       l = wctomb(self->qf_dat.s + self->qf_dat.l, wc);
 123       if (l < 0)
 124          goto jbad;
 125       self->qf_datw += (ui32_t)w;
 126       self->qf_dat.l += (size_t)l;
 127    }
 128
 129    /* TODO The last visual may excess (adjusted!) *qfold-max* if it's a wide;
 130     * TODO place it on the next line, break before */
 131    if (self->qf_datw >= self->qf_qfold_max) {
 132       /* If we have seen a nice breakpoint during traversal, shuffle data
 133        * around a bit so as to restore the trailing part after flushing */
 134       if (self->qf_brkl > 0) {
 135          save_w = self->qf_datw - self->qf_brkw;
 136          save_l = self->qf_dat.l - self->qf_brkl;
 137          save_b = self->qf_dat.s + self->qf_brkl + 2;
 138          memmove(save_b, save_b - 2, save_l);
 139          self->qf_dat.l = self->qf_brkl;
 140       }
 141
 142       self->qf_dat.s[self->qf_dat.l++] = '\\';
 143 jflush:
 144       self->qf_dat.s[self->qf_dat.l++] = '\n';
 145       rv = quoteflt_flush(self);
 146
 147       /* Restore takeovers, if any */
 148       if (save_b != NULL) {
 149          self->qf_brk_isws = FAL0;
 150          self->qf_datw += save_w;
 151          self->qf_dat.l = save_l;
 152          memmove(self->qf_dat.s, save_b, save_l);
 153       }
 154    } else if (self->qf_datw >= self->qf_qfold_min && !self->qf_brk_isws) {
 155       bool_t isws = (iswspace(wc) != 0);
 156
 157       if (isws || !self->qf_brk_isws || self->qf_brkl == 0) {
 158          self->qf_brkl = self->qf_dat.l;
 159          self->qf_brkw = self->qf_datw;
 160          self->qf_brk_isws = isws;
 161       }
 162    }
 163
 164    /* If state changed to prefix, perform full reset (note this implies that
 165     * quoteflt_flush() performs too much work..) */
 166    if (wc == '\n') {
 167       self->qf_state = _QF_PREFIX;
 168       self->qf_wscnt = self->qf_datw = 0;
 169       self->qf_currq.l = 0;
 170    }
 171 jleave:
 172    NYD_LEAVE;
 173    return rv;
 174 }
 175
 176 static ssize_t
 177 _qf_state_prefix(struct qf_vc *vc)
 178 {
 179    struct quoteflt *self;
 180    ssize_t rv;
 181    char const *buf;
 182    size_t len, i;
 183    wchar_t wc;
 184    NYD_ENTER;
 185
 186    self = vc->self;
 187    rv = 0;
 188
 189    for (buf = vc->buf, len = vc->len; len > 0;) {
 190       /* xxx NULL BYTE! */
 191       i = mbrtowc(&wc, buf, len, self->qf_mbps);
 192       if (i == (size_t)-1) {
 193          /* On hard error, don't modify mbstate_t and step one byte */
 194          self->qf_mbps[0] = self->qf_mbps[1];
 195          ++buf;
 196          --len;
 197          self->qf_wscnt = 0;
 198          continue;
 199       }
 200       self->qf_mbps[1] = self->qf_mbps[0];
 201       if (i == (size_t)-2) {
 202          /* Redundant shift sequence, out of buffer */
 203          len = 0;
 204          break;
 205       }
 206       buf += i;
 207       len -= i;
 208
 209       if (wc == L'\n')
 210          goto jfin;
 211       if (iswspace(wc)) {
 212          ++self->qf_wscnt;
 213          continue;
 214       }
 215       if (i == 1 && ISQUOTE(wc)) {
 216          self->qf_wscnt = 0;
 217          if (self->qf_currq.l >= QUOTE_MAX - 3) {
 218             self->qf_currq.s[QUOTE_MAX - 3] = '.';
 219             self->qf_currq.s[QUOTE_MAX - 2] = '.';
 220             self->qf_currq.s[QUOTE_MAX - 1] = '.';
 221             self->qf_currq.l = QUOTE_MAX;
 222          } else
 223             self->qf_currq.s[self->qf_currq.l++] = buf[-1];
 224          continue;
 225       }
 226
 227       /* The quote is parsed and compressed; dump it */
 228 jfin:
 229       self->qf_state = _QF_DATA;
 230       /* Overtake WS to the current quote in order to preserve it for eventual
 231        * necessary follow lines, too */
 232       /* TODO we de-facto "normalize" to ASCII SP here which MESSES tabs!! */
 233       while (self->qf_wscnt-- > 0 && self->qf_currq.l < QUOTE_MAX)
 234          self->qf_currq.s[self->qf_currq.l++] = ' ';
 235       self->qf_datw = self->qf_pfix_len + self->qf_currq.l;
 236       self->qf_wscnt = 0;
 237       rv = _qf_add_data(self, wc);
 238       break;
 239    }
 240
 241    vc->buf = buf;
 242    vc->len = len;
 243    NYD_LEAVE;
 244    return rv;
 245 }
 246
 247 static ssize_t
 248 _qf_state_data(struct qf_vc *vc)
 249 {
 250    struct quoteflt *self;
 251    ssize_t rv;
 252    char const *buf;
 253    size_t len, i;
 254    wchar_t wc;
 255    NYD_ENTER;
 256
 257    self = vc->self;
 258    rv = 0;
 259
 260    for (buf = vc->buf, len = vc->len; len > 0;) {
 261       /* xxx NULL BYTE! */
 262       i = mbrtowc(&wc, buf, len, self->qf_mbps);
 263       if (i == (size_t)-1) {
 264          /* On hard error, don't modify mbstate_t and step one byte */
 265          self->qf_mbps[0] = self->qf_mbps[1];
 266          ++buf;
 267          --len;
 268          continue;
 269       }
 270       self->qf_mbps[1] = self->qf_mbps[0];
 271       if (i == (size_t)-2) {
 272          /* Redundant shift sequence, out of buffer */
 273          len = 0;
 274          break;
 275       }
 276       buf += i;
 277       len -= i;
 278
 279       {  ssize_t j = _qf_add_data(self, wc);
 280          if (j < 0) {
 281             rv = j;
 282             break;
 283          }
 284          rv += j;
 285       }
 286
 287       if (self->qf_state != _QF_DATA)
 288          break;
 289    }
 290
 291    vc->buf = buf;
 292    vc->len = len;
 293    NYD_LEAVE;
 294    return rv;
 295 }
 296 #endif /* HAVE_QUOTE_FOLD */
 297
 298 FL struct quoteflt *
 299 quoteflt_dummy(void) /* TODO LEGACY (until filters are plugged when needed) */
 300 {
 301    static struct quoteflt qf_i;
 302
 303    return &qf_i;
 304 }
 305
 306 FL void
 307 quoteflt_init(struct quoteflt *self, char const *prefix)
 308 {
 309 #ifdef HAVE_QUOTE_FOLD
 310    char *xcp, *cp;
 311 #endif
 312    NYD_ENTER;
 313
 314    memset(self, 0, sizeof *self);
 315
 316    if ((self->qf_pfix = prefix) != NULL)
 317       self->qf_pfix_len = (ui32_t)strlen(prefix);
 318
 319    /* Check wether the user wants the more fancy quoting algorithm */
 320    /* TODO *quote-fold*: QUOTE_MAX may excess it! */
 321 #ifdef HAVE_QUOTE_FOLD
 322    if (self->qf_pfix_len > 0 && (cp = ok_vlook(quote_fold)) != NULL) {
 323       ui32_t qmin, qmax = (ui32_t)strtol(cp, &xcp, 10);
 324       /* These magic values ensure we don't bail :) */
 325       if (qmax < self->qf_pfix_len + 6)
 326          qmax = self->qf_pfix_len + 6;
 327       --qmax; /* The newline escape */
 328       if (cp == xcp || *xcp == '\0')
 329          qmin = (qmax >> 1) + (qmax >> 2) + (qmax >> 5);
 330       else {
 331          qmin = (ui32_t)strtol(xcp + 1, NULL, 10);
 332          if (qmin < qmax >> 1)
 333             qmin = qmax >> 1;
 334          else if (qmin > qmax - 2)
 335             qmin = qmax - 2;
 336       }
 337       self->qf_qfold_min = qmin;
 338       self->qf_qfold_max = qmax;
 339
 340       /* Add pad for takeover copies, backslash and newline */
 341       self->qf_dat.s = salloc((qmax + 3) * mb_cur_max);
 342       self->qf_currq.s = salloc((QUOTE_MAX + 1) * mb_cur_max);
 343    }
 344 #endif
 345    NYD_LEAVE;
 346 }
 347
 348 FL void
 349 quoteflt_destroy(struct quoteflt *self) /* xxx inline */
 350 {
 351    NYD_ENTER;
 352    UNUSED(self);
 353    NYD_LEAVE;
 354 }
 355
 356 FL void
 357 quoteflt_reset(struct quoteflt *self, FILE *f) /* xxx inline */
 358 {
 359    NYD_ENTER;
 360    self->qf_os = f;
 361 #ifdef HAVE_QUOTE_FOLD
 362    self->qf_state = _QF_CLEAN;
 363    self->qf_dat.l =
 364    self->qf_currq.l = 0;
 365    memset(self->qf_mbps, 0, sizeof self->qf_mbps);
 366 #endif
 367    NYD_LEAVE;
 368 }
 369
 370 FL ssize_t
 371 quoteflt_push(struct quoteflt *self, char const *dat, size_t len)
 372 {
 373    /* (xxx Ideally the actual push() [and flush()] would be functions on their
 374     * xxx own, via indirect vtbl call ..) */
 375    ssize_t rv = 0;
 376    NYD_ENTER;
 377
 378    if (len == 0)
 379       goto jleave;
 380
 381    /* Bypass? XXX Finally, this filter simply should not be used, then */
 382    if (self->qf_pfix_len == 0) {
 383       if (len != fwrite(dat, 1, len, self->qf_os))
 384          goto jerr;
 385       rv = len;
 386    }
 387    /* Normal: place *indentprefix* at every BOL */
 388    else
 389 #ifdef HAVE_QUOTE_FOLD
 390       if (self->qf_qfold_max == 0)
 391 #endif
 392    {
 393       void *vp;
 394       size_t ll;
 395       bool_t pxok = (self->qf_qfold_min != 0);
 396
 397       for (;;) {
 398          if (!pxok) {
 399             ll = self->qf_pfix_len;
 400             if (ll != fwrite(self->qf_pfix, 1, ll, self->qf_os))
 401                goto jerr;
 402             rv += ll;
 403             pxok = TRU1;
 404          }
 405
 406          /* xxx Strictly speaking this is invalid, because only `/' and `.' are
 407           * xxx mandated by POSIX.1-2008 as "invariant across all locales
 408           * xxx supported"; though there is no charset known which uses this
 409           * xxx control char as part of a multibyte character; note that S-nail
 410           * XXX (and the Mail codebase as such) do not support EBCDIC */
 411          if ((vp = memchr(dat, '\n', len)) == NULL)
 412             ll = len;
 413          else {
 414             pxok = FAL0;
 415             ll = PTR2SIZE((char*)vp - dat) + 1;
 416          }
 417
 418          if (ll != fwrite(dat, sizeof *dat, ll, self->qf_os))
 419             goto jerr;
 420          rv += ll;
 421          if ((len -= ll) == 0)
 422             break;
 423          dat += ll;
 424       }
 425
 426       self->qf_qfold_min = pxok;
 427    }
 428    /* Overly complicated, though still only line-per-line: *quote-fold*.
 429     * - If .qf_currq.l is 0, then we are in a clean state.  Reset .qf_mbps;
 430     *   TODO note this means we assume that lines start with reset escape seq,
 431     *   TODO but i don't think this is any worse than what we currently do;
 432     *   TODO in 15.0, with the value carrier, we should carry conversion states
 433     *   TODO all along, only resetting on error (or at words for header =???=);
 434     *   TODO this still is weird for error handling, but we need to act more
 435     *   TODO stream-alike (though in practice i don't think cross-line states
 436     *   TODO can be found, because of compatibility reasons; however, being
 437     *   TODO a problem rather than a solution is not a good thing (tm))
 438     * - Lookout for a newline */
 439 #ifdef HAVE_QUOTE_FOLD
 440    else {
 441       struct qf_vc vc;
 442       ssize_t i;
 443
 444       vc.self = self;
 445       vc.buf = dat;
 446       vc.len = len;
 447       while (vc.len > 0) {
 448          switch (self->qf_state) {
 449          case _QF_CLEAN:
 450          case _QF_PREFIX:
 451             i = _qf_state_prefix(&vc);
 452             break;
 453          default: /* silence cc (`i' unused) */
 454          case _QF_DATA:
 455             i = _qf_state_data(&vc);
 456             break;
 457          }
 458          if (i < 0)
 459             goto jerr;
 460          rv += i;
 461       }
 462    }
 463 #endif /* HAVE_QUOTE_FOLD */
 464
 465 jleave:
 466    NYD_LEAVE;
 467    return rv;
 468 jerr:
 469    rv = -1;
 470    goto jleave;
 471 }
 472
 473 FL ssize_t
 474 quoteflt_flush(struct quoteflt *self)
 475 {
 476    ssize_t rv = 0;
 477    NYD_ENTER;
 478    UNUSED(self);
 479
 480 #ifdef HAVE_QUOTE_FOLD
 481    if (self->qf_dat.l > 0) {
 482       rv = _qf_dump_prefix(self);
 483       if (rv >= 0) {
 484          size_t i = self->qf_dat.l;
 485          if (i == fwrite(self->qf_dat.s, 1, i, self->qf_os))
 486             rv += i;
 487          else
 488             rv = -1;
 489          self->qf_dat.l = 0;
 490          self->qf_brk_isws = FAL0;
 491          self->qf_wscnt = self->qf_brkl = self->qf_brkw = 0;
 492          self->qf_datw = self->qf_pfix_len + self->qf_currq.l;
 493       }
 494    }
 495 #endif
 496    NYD_LEAVE;
 497    return rv;
 498 }
 499
 500 /*
 501  * HTML tagsoup filter TODO rewrite wchar_t based (require HAVE_C90AMEND1)
 502  * TODO . Numeric &#NO; entities should also be treated by struct hf_ent
 503  * TODO . Yes, we COULD support CSS based quoting when we'd check type="quote"
 504  * TODO   (nonstandard) and watch out for style="gmail_quote" (or so, VERY
 505  * TODO   nonstandard) and tracking a stack of such elements (to be popped
 506  * TODO   once the closing element is seen).  Then, after writing a newline,
 507  * TODO   place sizeof(stack) ">"s first.  But aren't these HTML mails rude?
 508  * TODO Interlocking and non-well-formed data will break us down
 509  */
 510 #ifdef HAVE_FILTER_HTML_TAGSOUP
 511
 512 enum hf_limits {
 513    _HF_MINLEN  = 10,       /* Minimum line length (can't really be smaller) */
 514    _HF_BRKSUB  = 8         /* Start considering line break MAX - BRKSUB */
 515 };
 516
 517 enum hf_flags {
 518    _HF_UTF8    = 1<<0,     /* Data is in UTF-8 */
 519    _HF_ERROR   = 1<<1,     /* A hard error occurred, bail as soon as possible */
 520    _HF_NOPUT   = 1<<2,     /* (In a tag,) Don't generate output */
 521    _HF_IGN     = 1<<3,     /* Ignore mode on */
 522    _HF_ANY     = 1<<4,     /* Yet seen just any output */
 523    _HF_PRE     = 1<<5,     /* In <pre>formatted mode */
 524    _HF_ENT     = 1<<6,     /* Currently parsing an entity */
 525    _HF_BLANK   = 1<<7,     /* Whitespace last */
 526    _HF_HREF    = 1<<8,     /* External <a href=> was the last href seen */
 527
 528    _HF_NL_1    = 1<<9,     /* One \n seen */
 529    _HF_NL_2    = 2<<9,     /* We have produced an all empty line */
 530    _HF_NL_MASK = _HF_NL_1 | _HF_NL_2
 531 };
 532
 533 enum hf_special_actions {
 534    _HFSA_NEEDSEP  = -1,    /* Need an empty line (paragraph separator) */
 535    _HFSA_NEEDNL   = -2,    /* Need a new line start (table row) */
 536    _HFSA_IGN      = -3,    /* Things like <style>..</style>, <script>.. */
 537    _HFSA_PRE      = -4,    /* <pre>.. */
 538    _HFSA_PRE_END  = -5,
 539    _HFSA_IMG      = -6,    /* <img> */
 540    _HFSA_HREF     = -7,    /* <a>.. */
 541    _HFSA_HREF_END = -8
 542 };
 543
 544 enum hf_entity_flags {
 545    _HFE_HAVE_UNI  = 1<<6,  /* Have a Unicode replacement character */
 546    _HFE_HAVE_CSTR = 1<<7,  /* Have a string replacement */
 547    /* We store the length of the entity name in the flags, too */
 548    _HFE_LENGTH_MASK = (1<<6) - 1
 549 };
 550
 551 struct htmlflt_href {
 552    struct htmlflt_href *hfh_next;
 553    ui32_t      hfh_no;     /* Running sequence */
 554    ui32_t      hfh_len;    /* of .hfh_dat */
 555    char        hfh_dat[VFIELD_SIZE(0)];
 556 };
 557
 558 struct htmlflt_tag {
 559    si32_t      hft_act;    /* char or hf_special_actions */
 560    /* Not NUL: character to inject, with high bit set: place a space
 561     * afterwards.  Note: only recognized with _HFSA_NEEDSEP or _HFSA_NEEDNL */
 562    char        hft_injc;
 563    ui8_t       hft_len;    /* Useful bytes in (NUL terminated) .hft_tag */
 564    char const  hft_tag[10]; /* Tag less < and > surroundings (TR, /TR, ..) */
 565 };
 566 CTA(SIZEOF_FIELD(struct htmlflt_tag, hft_tag) < LINESIZE); /* .hf_ign_tag */
 567
 568 struct hf_ent {
 569    ui8_t       hfe_flags;  /* enum hf_entity_flags plus length of .hfe_ent */
 570    char        hfe_c;      /* Plain replacement character */
 571    ui16_t      hfe_uni;    /* Unicode codepoint if _HFE_HAVE_UNI */
 572    char        hfe_cstr[5]; /* _HFE_HAVE_CSTR (e.g., &hellip; -> ...) */
 573    char const  hfe_ent[7]; /* Entity less & and ; surroundings */
 574 };
 575
 576 /* Tag list; not binary searched :(, so try to take care a bit */
 577 static struct htmlflt_tag const  _hf_tags[] = {
 578 # undef _X
 579 # undef _XC
 580 # define _X(S,A)     {A, '\0', sizeof(S) -1, S "\0"}
 581 # define _XC(S,C,A)  {A, C, sizeof(S) -1, S "\0"}
 582
 583    _X("P", _HFSA_NEEDSEP),       _X("/P", _HFSA_NEEDNL),
 584    _X("DIV", _HFSA_NEEDSEP),     _X("/DIV", _HFSA_NEEDNL),
 585    _X("TR", _HFSA_NEEDNL),
 586                                  _X("/TH", '\t'),
 587                                  _X("/TD", '\t'),
 588    /* Let it stand out; also since we don't support implicit paragraphs after
 589     * block elements, plain running text after a list (seen in Unicode
 590     * announcement via Firefox) */
 591    _X("UL", _HFSA_NEEDSEP),      _X("/UL", _HFSA_NEEDSEP),
 592    _XC("LI", (char)0x80 | '*', _HFSA_NEEDSEP),
 593    _X("DL", _HFSA_NEEDSEP),
 594    _X("DT", _HFSA_NEEDNL),
 595
 596    _X("A", _HFSA_HREF),          _X("/A", _HFSA_HREF_END),
 597    _X("IMG", _HFSA_IMG),
 598    _X("BR", '\n'),
 599    _X("PRE", _HFSA_PRE),         _X("/PRE", _HFSA_PRE_END),
 600    _X("TITLE", _HFSA_NEEDSEP),   /*_X("/TITLE", '\n'),*/
 601    _X("H1", _HFSA_NEEDSEP),      /*_X("/H1", '\n'),*/
 602    _X("H2", _HFSA_NEEDSEP),      /*_X("/H2", '\n'),*/
 603    _X("H3", _HFSA_NEEDSEP),      /*_X("/H3", '\n'),*/
 604    _X("H4", _HFSA_NEEDSEP),      /*_X("/H4", '\n'),*/
 605    _X("H5", _HFSA_NEEDSEP),      /*_X("/H5", '\n'),*/
 606    _X("H6", _HFSA_NEEDSEP),      /*_X("/H6", '\n'),*/
 607
 608    _X("STYLE", _HFSA_IGN),
 609    _X("SCRIPT", _HFSA_IGN),
 610
 611 # undef _X
 612 };
 613
 614 /* Entity list; not binary searched.. */
 615 static struct hf_ent const       _hf_ents[] = {
 616 # undef _X
 617 # undef _XU
 618 # undef _XS
 619 # undef _XUS
 620 # define _X(E,C)     {(sizeof(E) -1), C, 0x0u, "", E "\0"}
 621 # define _XU(E,C,U)  {(sizeof(E) -1) | _HFE_HAVE_UNI, C, U, "", E "\0"}
 622 # define _XS(E,S)    {(sizeof(E) -1) | _HFE_HAVE_CSTR, '\0', 0x0u,S "\0",E "\0"}
 623 # define _XSU(E,S,U) \
 624    {(sizeof(E) -1) | _HFE_HAVE_UNI | _HFE_HAVE_CSTR, '\0', U, S "\0", E "\0"}
 625
 626    _X("quot", '"'),
 627    _X("amp", '&'),
 628    _X("lt", '<'),                _X("gt", '>'),
 629
 630    _XU("nbsp", ' ', 0x0020 /* Note: not 0x00A0 seems to be better for us */),
 631    _XU("middot", '.', 0x00B7),
 632    _XSU("hellip", "...", 0x2026),
 633    _XSU("mdash", "---", 0x2014), _XSU("ndash", "--", 0x2013),
 634    _XSU("laquo", "<<", 0x00AB),  _XSU("raquo", ">>", 0x00BB),
 635    _XSU("lsaquo", "<", 0x2039),  _XSU("rsaquo", ">", 0x203A),
 636    _XSU("lsquo", "'", 0x2018),   _XSU("rsquo", "'", 0x2019),
 637    _XSU("ldquo", "\"", 0x201C),  _XSU("rdquo", "\"", 0x201D),
 638    _XSU("uarr", "^|", 0x2191),   _XSU("darr", "|v", 0x2193),
 639
 640    _XSU("cent", "CENT", 0x00A2),
 641    _XSU("copy", "(C)", 0x00A9),
 642    _XSU("euro", "EUR", 0x20AC),
 643    _XSU("infin", "INFY", 0x221E),
 644    _XSU("pound", "GBP", 0x00A3),
 645    _XSU("reg", "(R)", 0x00AE),
 646    _XSU("sect", "S:", 0x00A7),
 647    _XSU("yen", "JPY", 0x00A5),
 648
 649    /* German umlauts */
 650    _XSU("Auml", "Ae", 0x00C4),   _XSU("auml", "ae", 0x00E4),
 651    _XSU("Ouml", "Oe", 0x00D6),   _XSU("ouml", "oe", 0x00F6),
 652    _XSU("Uuml", "Ue", 0x00DC),   _XSU("uuml", "ue", 0x00FC),
 653    _XSU("szlig", "ss", 0x00DF)
 654
 655 # undef _X
 656 # undef _XU
 657 # undef _XS
 658 # undef _XSU
 659 };
 660
 661 /* Real output */
 662 static struct htmlflt * _hf_dump_hrefs(struct htmlflt *self);
 663 static struct htmlflt * _hf_dump(struct htmlflt *self);
 664 static struct htmlflt * _hf_store(struct htmlflt *self, char c);
 665 # ifdef HAVE_NATCH_CHAR
 666 static struct htmlflt * __hf_sync_mbstuff(struct htmlflt *self);
 667 # endif
 668
 669 /* Virtual output */
 670 static struct htmlflt * _hf_nl(struct htmlflt *self);
 671 static struct htmlflt * _hf_nl_force(struct htmlflt *self);
 672 static struct htmlflt * _hf_putc(struct htmlflt *self, char c);
 673 static struct htmlflt * _hf_putc_premode(struct htmlflt *self, char c);
 674 static struct htmlflt * _hf_puts(struct htmlflt *self, char const *cp);
 675 static struct htmlflt * _hf_putbuf(struct htmlflt *self,
 676                            char const *cp, size_t len);
 677
 678 /* Try to locate a param'eter in >hf_bdat, store it (non-terminated!) or NULL */
 679 static struct htmlflt * _hf_param(struct htmlflt *self, struct str *store,
 680                            char const *param);
 681
 682 /* Expand all entities in the given parameter */
 683 static struct htmlflt * _hf_expand_all_ents(struct htmlflt *self,
 684                            struct str const *param);
 685
 686 /* Completely parsed over a tag / an entity, interpret that */
 687 static struct htmlflt * _hf_check_tag(struct htmlflt *self, char const *s);
 688 static struct htmlflt * _hf_check_ent(struct htmlflt *self, char const *s,
 689                            size_t l);
 690
 691 /* Input handler */
 692 static ssize_t          _hf_add_data(struct htmlflt *self,
 693                            char const *dat, size_t len);
 694
 695 static struct htmlflt *
 696 _hf_dump_hrefs(struct htmlflt *self)
 697 {
 698    struct htmlflt_href *hhp;
 699    NYD2_ENTER;
 700
 701    if (!(self->hf_flags & _HF_NL_2) && putc('\n', self->hf_os) == EOF) {
 702       self->hf_flags |= _HF_ERROR;
 703       goto jleave;
 704    }
 705
 706    /* Reverse the list */
 707    for (hhp = self->hf_hrefs, self->hf_hrefs = NULL; hhp != NULL;) {
 708       struct htmlflt_href *tmp = hhp->hfh_next;
 709       hhp->hfh_next = self->hf_hrefs;
 710       self->hf_hrefs = hhp;
 711       hhp = tmp;
 712    }
 713
 714    /* Then dump it */
 715    while ((hhp = self->hf_hrefs) != NULL) {
 716       self->hf_hrefs = hhp->hfh_next;
 717
 718       if (!(self->hf_flags & _HF_ERROR)) {
 719          int w = fprintf(self->hf_os, "  [%u] %.*s\n",
 720                hhp->hfh_no, (int)hhp->hfh_len, hhp->hfh_dat);
 721          if (w < 0)
 722             self->hf_flags |= _HF_ERROR;
 723       }
 724       free(hhp);
 725    }
 726
 727    self->hf_flags |= (putc('\n', self->hf_os) == EOF)
 728          ?  _HF_ERROR : _HF_NL_1 | _HF_NL_2;
 729    self->hf_href_dist = (ui32_t)realscreenheight >> 1;
 730 jleave:
 731    NYD2_LEAVE;
 732    return self;
 733 }
 734
 735 static struct htmlflt *
 736 _hf_dump(struct htmlflt *self)
 737 {
 738    ui32_t f, l;
 739    char c, *cp;
 740    NYD2_ENTER;
 741
 742    f = self->hf_flags & ~_HF_BLANK;
 743    l = self->hf_len;
 744    cp = self->hf_line;
 745    self->hf_mbwidth = self->hf_mboff = self->hf_last_ws = self->hf_len = 0;
 746
 747    for (c = '\0'; l > 0; --l) {
 748       c = *cp++;
 749 jput:
 750       if (putc(c, self->hf_os) == EOF) {
 751          self->hf_flags = (f |= _HF_ERROR);
 752          goto jleave;
 753       }
 754    }
 755
 756    if (c != '\n') {
 757       f |= (f & _HF_NL_1) ? _HF_NL_2 : _HF_NL_1;
 758       l = 1;
 759       c = '\n';
 760       goto jput;
 761    }
 762    self->hf_flags = f;
 763
 764    /* Check wether there are HREFs to dump; there is so much messy tagsoup out
 765     * there that it seems best not to simply dump HREFs in each _dump(), but
 766     * only with some gap, let's say half the real screen height */
 767    if (--self->hf_href_dist < 0 && (f & _HF_NL_2) && self->hf_hrefs != NULL)
 768       self = _hf_dump_hrefs(self);
 769 jleave:
 770    NYD2_LEAVE;
 771    return self;
 772 }
 773
 774 static struct htmlflt *
 775 _hf_store(struct htmlflt *self, char c)
 776 {
 777    ui32_t f, l, i;
 778    NYD2_ENTER;
 779
 780    assert(c != '\n');
 781
 782    f = self->hf_flags;
 783    l = self->hf_len;
 784    self->hf_line[l] = (c == '\t' ? ' ' : c);
 785    self->hf_len = ++l;
 786    if (blankspacechar(c)) {
 787       if (c == '\t') {
 788          i = 8 - ((l - 1) & 7); /* xxx magic tab width of 8 */
 789          if (i > 0) {
 790             do
 791                self = _hf_store(self, ' ');
 792             while (--i > 0);
 793             goto jleave;
 794          }
 795       }
 796       self->hf_last_ws = l;
 797    } else if (/*c == '.' ||*/ c == ',' || c == ';' || c == '-')
 798       self->hf_last_ws = l;
 799
 800    i = l;
 801 # ifdef HAVE_NATCH_CHAR /* XXX This code is really ridiculous! */
 802    if (mb_cur_max > 1) { /* XXX should mbrtowc() and THEN store, at least.. */
 803       wchar_t wc;
 804       int w, x = mbtowc(&wc, self->hf_line + self->hf_mboff, l - self->hf_mboff);
 805
 806       if (x > 0) {
 807          if ((w = wcwidth(wc)) == -1 ||
 808                /* Actively filter out L-TO-R and R-TO-R marks TODO ctext */
 809                (wc == 0x200E || wc == 0x200F ||
 810                   (wc >= 0x202A && wc <= 0x202E)) ||
 811                /* And some zero-width messes */
 812                wc == 0x00AD || (wc >= 0x200B && wc <= 0x200D) ||
 813                /* Oh about the ISO C wide character interfaces, baby! */
 814                (wc == 0xFEFF)){
 815             self->hf_len -= x;
 816             goto jleave;
 817          } else if (iswspace(wc))
 818             self->hf_last_ws = l;
 819          self->hf_mboff += x;
 820          i = (self->hf_mbwidth += w);
 821       } else {
 822          if (x < 0) {
 823             mbtowc(&wc, NULL, mb_cur_max);
 824             if (UICMP(32, l - self->hf_mboff, >=, mb_cur_max)) { /* XXX */
 825                ++self->hf_mboff;
 826                ++self->hf_mbwidth;
 827             }
 828          }
 829          i = self->hf_mbwidth;
 830       }
 831    }
 832 # endif
 833
 834    /* Do we need to break the line? */
 835    if (i >= self->hf_lmax - _HF_BRKSUB) {
 836       ui32_t lim = self->hf_lmax >> 1;
 837
 838       /* Let's hope we saw a sane place to break this line! */
 839       if (self->hf_last_ws >= lim) {
 840 jput:
 841          i = self->hf_len = self->hf_last_ws;
 842          self = _hf_dump(self);
 843          if ((self->hf_len = (l -= i)) > 0) {
 844             self->hf_flags &= ~_HF_NL_MASK;
 845             memmove(self->hf_line, self->hf_line + i, l);
 846 # ifdef HAVE_NATCH_CHAR
 847             __hf_sync_mbstuff(self);
 848 # endif
 849          }
 850          goto jleave;
 851       }
 852
 853       /* Any 7-bit characters? */
 854       for (i = l; i-- >= lim;)
 855          if (asciichar((c = self->hf_line[i]))) {
 856             self->hf_last_ws = ++i;
 857             goto jput;
 858          } else if ((f & _HF_UTF8) && ((ui8_t)c & 0xC0) != 0x80) {
 859             self->hf_last_ws = i;
 860             goto jput;
 861          }
 862
 863       /* Hard break necessary!  xxx really badly done */
 864       if (l >= self->hf_lmax - 1)
 865          self = _hf_dump(self);
 866    }
 867 jleave:
 868    NYD2_LEAVE;
 869    return self;
 870 }
 871
 872 # ifdef HAVE_NATCH_CHAR
 873 static struct htmlflt *
 874 __hf_sync_mbstuff(struct htmlflt *self)
 875 {
 876    wchar_t wc;
 877    char const *b;
 878    ui32_t o, w, l;
 879    NYD2_ENTER;
 880
 881    b = self->hf_line;
 882    o = w = 0;
 883    l = self->hf_len;
 884    goto jumpin;
 885
 886    while (l > 0) {
 887       int x = mbtowc(&wc, b, l);
 888
 889       if (x == 0)
 890          break;
 891
 892       if (x > 0) {
 893          b += x;
 894          l -= x;
 895          o += x;
 896          if ((x = wcwidth(wc)) == -1)
 897             x = 1;
 898          w += x;
 899          continue;
 900       }
 901
 902       /* Bad, skip over a single character.. XXX very bad indeed */
 903       ++b;
 904       ++o;
 905       ++w;
 906       --l;
 907 jumpin:
 908       mbtowc(&wc, NULL, mb_cur_max);
 909    }
 910
 911    self->hf_mboff = o;
 912    self->hf_mbwidth = w;
 913
 914    NYD2_LEAVE;
 915    return self;
 916 }
 917 # endif /* HAVE_NATCH_CHAR */
 918
 919 static struct htmlflt *
 920 _hf_nl(struct htmlflt *self)
 921 {
 922    ui32_t f;
 923    NYD2_ENTER;
 924
 925    if (!((f = self->hf_flags) & _HF_ERROR)) {
 926       if (f & _HF_ANY) {
 927          if ((f & _HF_NL_MASK) != _HF_NL_MASK)
 928             self = _hf_dump(self);
 929       } else
 930          self->hf_flags = (f |= _HF_NL_MASK);
 931    }
 932    NYD2_LEAVE;
 933    return self;
 934 }
 935
 936 static struct htmlflt *
 937 _hf_nl_force(struct htmlflt *self)
 938 {
 939    NYD2_ENTER;
 940    if (!(self->hf_flags & _HF_ERROR))
 941       self = _hf_dump(self);
 942    NYD2_LEAVE;
 943    return self;
 944 }
 945
 946 static struct htmlflt *
 947 _hf_putc(struct htmlflt *self, char c)
 948 {
 949    ui32_t f;
 950    NYD2_ENTER;
 951
 952    if ((f = self->hf_flags) & _HF_ERROR)
 953       goto jleave;
 954
 955    if (c == '\n') {
 956       self = _hf_nl(self);
 957       goto jleave;
 958    } else if (c == ' ' || c == '\t') {
 959       if ((f & _HF_BLANK) || self->hf_len == 0)
 960          goto jleave;
 961       f |= _HF_BLANK;
 962    } else
 963       f &= ~_HF_BLANK;
 964    f &= ~_HF_NL_MASK;
 965    self->hf_flags = (f |= _HF_ANY);
 966    self = _hf_store(self, c);
 967 jleave:
 968    NYD2_LEAVE;
 969    return self;
 970 }
 971
 972 static struct htmlflt *
 973 _hf_putc_premode(struct htmlflt *self, char c)
 974 {
 975    ui32_t f;
 976    NYD2_ENTER;
 977
 978    if ((f = self->hf_flags) & _HF_ERROR) {
 979       ;
 980    } else if (c == '\n')
 981       self = _hf_nl_force(self);
 982    else {
 983       f &= ~_HF_NL_MASK;
 984       self->hf_flags = (f |= _HF_ANY);
 985       self = _hf_store(self, c);
 986    }
 987    NYD2_LEAVE;
 988    return self;
 989 }
 990
 991 static struct htmlflt *
 992 _hf_puts(struct htmlflt *self, char const *cp)
 993 {
 994    char c;
 995    NYD2_ENTER;
 996
 997    while ((c = *cp++) != '\0')
 998       self = _hf_putc(self, c);
 999    NYD2_LEAVE;
1000    return self;
1001 }
1002
1003 static struct htmlflt *
1004 _hf_putbuf(struct htmlflt *self, char const *cp, size_t len)
1005 {
1006    NYD2_ENTER;
1007
1008    while (len-- > 0)
1009       self = _hf_putc(self, *cp++);
1010    NYD2_LEAVE;
1011    return self;
1012 }
1013
1014 static struct htmlflt *
1015 _hf_param(struct htmlflt *self, struct str *store, char const *param)
1016 {
1017    char const *cp;
1018    char c, x, quote;
1019    size_t i;
1020    bool_t hot;
1021    NYD2_ENTER;
1022
1023    store->s = NULL;
1024    store->l = 0;
1025    cp = self->hf_bdat;
1026
1027    /* Skip over any non-WS first; be aware of soup, if it slipped through */
1028    for(;;){
1029       if((c = *cp++) == '\0' || c == '>')
1030          goto jleave;
1031       if(whitechar(c))
1032          break;
1033    }
1034
1035    /* Search for the parameter, take care of other quoting along the way */
1036    x = *param++;
1037    x = upperconv(x);
1038    quote = '\0';
1039    i = strlen(param);
1040
1041    for(hot = TRU1;;){
1042       if((c = *cp++) == '\0' || c == '>')
1043          goto jleave;
1044       if(whitechar(c)){
1045          hot = TRU1;
1046          continue;
1047       }
1048
1049       /* Could it be a parameter? */
1050       if(hot){
1051          hot = FAL0;
1052
1053          /* Is it the desired one? */
1054          if((c = upperconv(c)) == x && !ascncasecmp(param, cp, i)){
1055             char const *cp2 = cp + i;
1056
1057             if((quote = *cp2++) != '='){
1058                if(quote == '\0' || quote == '>')
1059                   goto jleave;
1060                while(whitechar(quote))
1061                   quote = *cp2++;
1062             }
1063             if(quote == '='){
1064                cp = cp2;
1065                break;
1066             }
1067             continue; /* XXX Optimize: i bytes or even cp2 can't be it! */
1068          }
1069       }
1070
1071       /* Not the desired one; but a parameter? */
1072       if(c != '=')
1073          continue;
1074       /* If so, properly skip over the value */
1075       if(c == '"' || c == '\''){
1076          /* TODO oops i have forgotten wether backslash quoting is allowed in
1077           * TODO quoted HTML parameter values?  not supporting that for now.. */
1078          for(quote = c; (c = *cp) != '\0' && c != quote; ++cp)
1079             ;
1080       }else
1081          while((c = *cp) != '\0' && !whitechar(c) && c != '>')
1082             ++cp;
1083       if(c == '\0')
1084          goto jleave;
1085    }
1086
1087    /* Skip further whitespace */
1088    for(;;){
1089       if((c = *cp++) == '\0' || c == '>')
1090          goto jleave;
1091       if(!whitechar(c))
1092          break;
1093    }
1094
1095    if(c == '"' || c == '\''){
1096       /* TODO oops i have forgotten wether backslash quoting is allowed in
1097        * TODO quoted HTML parameter values?  not supporting that for now.. */
1098       store->s = UNCONST(cp);
1099       for(quote = c; (c = *cp) != '\0' && c != quote; ++cp)
1100          ;
1101       /* XXX ... and we simply ignore a missing trailing " :> */
1102    }else{
1103       store->s = UNCONST(cp - 1);
1104       if(!whitechar(c))
1105          while((c = *cp) != '\0' && !whitechar(c) && c != '>')
1106             ++cp;
1107    }
1108    i = PTR2SIZE(cp - store->s);
1109
1110    /* Terrible tagsoup out there, e.g., groups.google.com produces href=""
1111     * parameter values prefixed and suffixed by newlines!  Therefore trim the
1112     * value content TODO join into the parse step above! */
1113    for (cp = store->s; i > 0 && spacechar(*cp); ++cp, --i)
1114       ;
1115    store->s = UNCONST(cp);
1116    for (cp += i - 1; i > 0 && spacechar(*cp); --cp, --i)
1117       ;
1118    if ((store->l = i) == 0)
1119       store->s = NULL;
1120 jleave:
1121    NYD2_LEAVE;
1122    return self;
1123 }
1124
1125 static struct htmlflt *
1126 _hf_expand_all_ents(struct htmlflt *self, struct str const *param)
1127 {
1128    char const *cp, *maxcp, *ep;
1129    char c;
1130    size_t i;
1131    NYD2_ENTER;
1132
1133    for (cp = param->s, maxcp = cp + param->l; cp < maxcp;)
1134       if ((c = *cp++) != '&')
1135 jputc:
1136          self = _hf_putc(self, c);
1137       else {
1138          for (ep = cp--;;) {
1139             if (ep == maxcp || (c = *ep++) == '\0') {
1140                for (; cp < ep; ++cp)
1141                   self = _hf_putc(self, *cp);
1142                goto jleave;
1143             } else if (c == ';') {
1144                if ((i = PTR2SIZE(ep - cp)) > 1) {
1145                   self = _hf_check_ent(self, cp, i);
1146                   break;
1147                } else {
1148                   c = *cp++;
1149                   goto jputc;
1150                }
1151             }
1152          }
1153          cp = ep;
1154       }
1155 jleave:
1156    NYD2_LEAVE;
1157    return self;
1158 }
1159
1160 static struct htmlflt *
1161 _hf_check_tag(struct htmlflt *self, char const *s)
1162 {
1163    char nobuf[32], c;
1164    struct str param;
1165    size_t i;
1166    struct htmlflt_tag const *hftp;
1167    ui32_t f;
1168    NYD2_ENTER;
1169
1170    /* Extra check only */
1171    assert(s != NULL);
1172    if (*s != '<') {
1173       DBG( n_alert("HTML tagsoup filter _hf_check_tag() called on soup!"); )
1174 jput_as_is:
1175       self = _hf_puts(self, self->hf_bdat);
1176       goto jleave;
1177    }
1178
1179    for (++s, i = 0; (c = s[i]) != '\0' && c != '>' && !whitechar(c); ++i)
1180       /* Special massage for things like <br/>: after the slash only whitespace
1181        * may separate us from the closing right angle! */
1182       if (c == '/') {
1183          size_t j = i + 1;
1184
1185          while ((c = s[j]) != '\0' && c != '>' && whitechar(c))
1186             ++j;
1187          if (c == '>')
1188             break;
1189       }
1190
1191    for (hftp = _hf_tags;;) {
1192       if (i == hftp->hft_len && !ascncasecmp(s, hftp->hft_tag, i)) {
1193          c = s[hftp->hft_len];
1194          if (c == '>' || c == '/' || whitechar(c))
1195             break;
1196       }
1197       if (PTRCMP(++hftp, >=, _hf_tags + NELEM(_hf_tags)))
1198          goto jnotknown;
1199    }
1200
1201    f = self->hf_flags;
1202    switch (hftp->hft_act) {
1203    case _HFSA_PRE_END:
1204       f &= ~_HF_PRE;
1205       if (0) {
1206          /* FALLTHRU */
1207    case _HFSA_PRE:
1208          f |= _HF_PRE;
1209       }
1210       self->hf_flags = f;
1211       /* FALLTHRU */
1212
1213    case _HFSA_NEEDSEP:
1214       if (!(self->hf_flags & _HF_NL_2))
1215          self = _hf_nl(self);
1216       /* FALLTHRU */
1217    case _HFSA_NEEDNL:
1218       if (!(f & _HF_NL_1))
1219          self = _hf_nl(self);
1220       if (hftp->hft_injc != '\0') {
1221          self = _hf_putc(self, hftp->hft_injc & 0x7F);
1222          if ((uc_i)hftp->hft_injc & 0x80)
1223             self = _hf_putc(self, ' ');
1224       }
1225       break;
1226
1227    case _HFSA_IGN:
1228       self->hf_ign_tag = hftp;
1229       self->hf_flags = (f |= _HF_IGN | _HF_NOPUT);
1230       break;
1231
1232    case _HFSA_IMG:
1233       self = _hf_param(self, &param, "alt");
1234       self = _hf_putc(self, '[');
1235       if (param.s == NULL) {
1236          param.s = UNCONST("IMG");
1237          param.l = 3;
1238          goto jimg_put;
1239       } /* else */ if (memchr(param.s, '&', param.l) != NULL)
1240          self = _hf_expand_all_ents(self, &param);
1241       else
1242 jimg_put:
1243          self = _hf_putbuf(self, param.s, param.l);
1244       self = _hf_putc(self, ']');
1245       break;
1246
1247    case _HFSA_HREF:
1248       self = _hf_param(self, &param, "href");
1249       /* Ignore non-external links */
1250       if (param.s != NULL && *param.s != '#') {
1251          struct htmlflt_href *hhp = smalloc(sizeof(*hhp) -
1252                VFIELD_SIZEOF(struct htmlflt_href, hfh_dat) + param.l +1);
1253
1254          hhp->hfh_next = self->hf_hrefs;
1255          hhp->hfh_no = ++self->hf_href_no;
1256          hhp->hfh_len = (ui32_t)param.l;
1257          memcpy(hhp->hfh_dat, param.s, param.l);
1258
1259          snprintf(nobuf, sizeof nobuf, "[%u]", hhp->hfh_no);
1260          self->hf_flags = (f |= _HF_HREF);
1261          self->hf_hrefs = hhp;
1262          self = _hf_puts(self, nobuf);
1263       } else
1264          self->hf_flags = (f &= ~_HF_HREF);
1265       break;
1266    case _HFSA_HREF_END:
1267       if (f & _HF_HREF) {
1268          snprintf(nobuf, sizeof nobuf, "[/%u]", self->hf_href_no);
1269          self = _hf_puts(self, nobuf);
1270       }
1271       break;
1272
1273    default:
1274       c = (char)(hftp->hft_act & 0xFF);
1275       self = _hf_putc(self, c);
1276       break;
1277    case '\0':
1278       break;
1279    }
1280 jleave:
1281    NYD2_LEAVE;
1282    return self;
1283
1284    /* The problem is that even invalid tagsoup is widely used, without real
1285     * searching i have seen e-mail address in <N@H.D> notation, and more.
1286     * To protect us a bit look around and possibly write the content as such */
1287 jnotknown:
1288    switch (*s) {
1289    case '!':
1290    case '?':
1291       /* Ignore <!DOCTYPE, <!-- comments, <? PIs.. */
1292       goto jleave;
1293    case '>':
1294       /* Print out an empty tag as such */
1295       if (s[1] == '\0') {
1296          --s;
1297          goto jput_as_is;
1298       }
1299       break;
1300    case '/':
1301       ++s;
1302       break;
1303    default:
1304       break;
1305    }
1306
1307    /* Also skip over : in order to suppress v:roundrect, w:anchorlock.. */
1308    while ((c = *s++) != '\0' && c != '>' && !whitechar(c) && c != ':')
1309       if (!asciichar(c) || punctchar(c)) {
1310          self = _hf_puts(self, self->hf_bdat);
1311          break;
1312       }
1313    goto jleave;
1314 }
1315
1316 static struct htmlflt *
1317 _hf_check_ent(struct htmlflt *self, char const *s, size_t l)
1318 {
1319    char nobuf[32];
1320    char const *s_save;
1321    size_t l_save;
1322    struct hf_ent const *hfep;
1323    long i;
1324    NYD2_ENTER;
1325
1326    s_save = s;
1327    l_save = l;
1328    assert(*s == '&');
1329    assert(l > 0);
1330    /* False entities seen in the wild assert(s[l - 1] == ';'); */
1331    ++s;
1332    l -= 2;
1333
1334    /* Numeric entity, or try named search */
1335    if (*s == '#') {
1336       i = (*++s == 'x' ? 16 : 10);
1337
1338       if ((i != 16 || (++s, --l) > 0) && l < sizeof(nobuf)) {
1339          memcpy(nobuf, s, l);
1340          nobuf[l] = '\0';
1341          i = strtol(nobuf, NULL, i);
1342          if (i <= 0x7F)
1343             self = _hf_putc(self, (char)i);
1344          else if (self->hf_flags & _HF_UTF8) {
1345 jputuni:
1346             l = n_utf32_to_utf8((ui32_t)i, nobuf);
1347             self = _hf_putbuf(self, nobuf, l);
1348          } else
1349             goto jeent;
1350       } else
1351          goto jeent;
1352    } else {
1353       ui32_t f = self->hf_flags, hf;
1354
1355       for (hfep = _hf_ents; PTRCMP(hfep, <, _hf_ents + NELEM(_hf_ents)); ++hfep)
1356          if (l == ((hf = hfep->hfe_flags) & _HFE_LENGTH_MASK) &&
1357                !strncmp(s, hfep->hfe_ent, l)) {
1358             if ((hf & _HFE_HAVE_UNI) && (f & _HF_UTF8)) {
1359                i = hfep->hfe_uni;
1360                goto jputuni;
1361             } else if (hf & _HFE_HAVE_CSTR)
1362                self = _hf_puts(self, hfep->hfe_cstr);
1363             else
1364                self = _hf_putc(self, hfep->hfe_c);
1365             goto jleave;
1366          }
1367 jeent:
1368       self = _hf_putbuf(self, s_save, l_save);
1369    }
1370 jleave:
1371    NYD2_LEAVE;
1372    return self;
1373 }
1374
1375 static ssize_t
1376 _hf_add_data(struct htmlflt *self, char const *dat, size_t len)
1377 {
1378    char c, *cp, *cp_max;
1379    bool_t hot;
1380    ssize_t rv = 0;
1381    NYD_ENTER;
1382
1383    /* Final put request? */
1384    if (dat == NULL) {
1385       if (self->hf_len > 0 || self->hf_hrefs != NULL) {
1386          self = _hf_dump(self);
1387          if (self->hf_hrefs != NULL)
1388             self = _hf_dump_hrefs(self);
1389          rv = 1;
1390       }
1391       goto jleave;
1392    }
1393
1394    /* Always ensure some initial buffer */
1395    if ((cp = self->hf_curr) != NULL)
1396       cp_max = self->hf_bmax;
1397    else {
1398       cp = self->hf_curr = self->hf_bdat = smalloc(LINESIZE);
1399       cp_max = self->hf_bmax = cp + LINESIZE -1; /* (Always room for NUL!) */
1400    }
1401    hot = (cp != self->hf_bdat);
1402
1403    for (rv = (ssize_t)len; len > 0; --len) {
1404       ui32_t f = self->hf_flags;
1405
1406       if (f & _HF_ERROR)
1407          break;
1408       c = *dat++;
1409
1410       /* Soup is really weird, and scripts may contain almost anything (and
1411        * newer CSS standards are also cryptic): therefore prefix the _HF_IGN
1412        * test and walk until we see the required end tag */
1413       /* TODO For real safety _HF_IGN soup condome would also need to know
1414        * TODO about quoted strings so that 'var i = "</script>";' couldn't
1415        * TODO fool it!   We really want this mode also for _HF_NOPUT to be
1416        * TODO able to *gracefully* detect the tag-closing '>', but then if
1417        * TODO that is a single mechanism we should have made it! */
1418       if (f & _HF_IGN) {
1419          struct htmlflt_tag const *hftp = self->hf_ign_tag;
1420          size_t i;
1421
1422          if (c == '<') {
1423             hot = TRU1;
1424 jcp_reset:
1425             cp = self->hf_bdat;
1426          } else if (c == '>') {
1427             if (hot) {
1428                if ((i = PTR2SIZE(cp - self->hf_bdat)) > 1 &&
1429                      --i == hftp->hft_len &&
1430                      !ascncasecmp(self->hf_bdat + 1, hftp->hft_tag, i))
1431                   self->hf_flags = (f &= ~(_HF_IGN | _HF_NOPUT));
1432                hot = FAL0;
1433                goto jcp_reset;
1434             }
1435          } else if (hot) {
1436             *cp++ = c;
1437             i = PTR2SIZE(cp - self->hf_bdat);
1438             if ((i == 1 && c != '/') || --i > hftp->hft_len) {
1439                hot = FAL0;
1440                goto jcp_reset;
1441             }
1442          }
1443       } else switch (c) {
1444       case '<':
1445          /* People are using & without &amp;ing it, ditto <; be aware */
1446          if (f & (_HF_NOPUT | _HF_ENT)) {
1447             f &= ~_HF_ENT;
1448             /* Special case "<!--" buffer content to deal with really weird
1449              * things that can be done with "<!--[if gte mso 9]>" syntax */
1450             if (PTR2SIZE(cp - self->hf_bdat) != 4 ||
1451                   memcmp(self->hf_bdat, "<!--", 4)) {
1452                self->hf_flags = f;
1453                *cp = '\0';
1454                self = _hf_puts(self, self->hf_bdat);
1455                f = self->hf_flags;
1456             }
1457          }
1458          cp = self->hf_bdat;
1459          *cp++ = c;
1460          self->hf_flags = (f |= _HF_NOPUT);
1461          break;
1462       case '>':
1463          /* Weird tagsoup around, do we actually parse a tag? */
1464          if (!(f & _HF_NOPUT))
1465             goto jdo_c;
1466          cp[0] = c;
1467          cp[1] = '\0';
1468          f &= ~(_HF_NOPUT | _HF_ENT);
1469          self->hf_flags = f;
1470          self = _hf_check_tag(self, self->hf_bdat);
1471          *(cp = self->hf_bdat) = '\0'; /* xxx extra safety */
1472          /* Quick hack to get rid of redundant newline after <pre> XXX */
1473          if (!(f & _HF_PRE) && (self->hf_flags & _HF_PRE) &&
1474                len > 1 && *dat == '\n')
1475             ++dat, --len;
1476          break;
1477
1478       case '\r': /* TODO CR should be stripped in lower level!! (Only B64!?!) */
1479          break;
1480       case '\n':
1481          /* End of line is not considered unless we are in PRE section.
1482           * However, in _HF_NOPUT mode we must be aware of tagsoup which uses
1483           * newlines for separating parameters */
1484          if (f & _HF_NOPUT)
1485             goto jdo_c;
1486          self = (f & _HF_PRE) ? _hf_nl_force(self) : _hf_putc(self, ' ');
1487          break;
1488
1489       case '\t':
1490          if (!(f & _HF_PRE))
1491             c = ' ';
1492          /* FALLTHRU */
1493       default:
1494 jdo_c:
1495          /* If not currently parsing a tag and bypassing normal output.. */
1496          if (!(f & _HF_NOPUT)) {
1497             if (cntrlchar(c))
1498                break;
1499             if (c == '&') {
1500                cp = self->hf_bdat;
1501                *cp++ = c;
1502                self->hf_flags = (f |= _HF_NOPUT | _HF_ENT);
1503             } else if (f & _HF_PRE) {
1504                self = _hf_putc_premode(self, c);
1505                self->hf_flags &= ~_HF_BLANK;
1506             } else
1507               self = _hf_putc(self, c);
1508          } else if ((f & _HF_ENT) && c == ';') {
1509             cp[0] = c;
1510             cp[1] = '\0';
1511             f &= ~(_HF_NOPUT | _HF_ENT);
1512             self->hf_flags = f;
1513            self = _hf_check_ent(self, self->hf_bdat,
1514                PTR2SIZE(cp + 1 - self->hf_bdat));
1515          } else {
1516             /* We may need to grow the buffer */
1517             if (PTRCMP(cp + 42/2, >=, cp_max)) {
1518                size_t i = PTR2SIZE(cp - self->hf_bdat),
1519                   m = PTR2SIZE(self->hf_bmax - self->hf_bdat) + LINESIZE;
1520
1521                cp = self->hf_bdat = srealloc(self->hf_bdat, m);
1522                self->hf_bmax = cp + m -1;
1523                self->hf_curr = (cp += i);
1524             }
1525             *cp++ = c;
1526          }
1527       }
1528    }
1529    self->hf_curr = cp;
1530 jleave:
1531   NYD_LEAVE;
1532   return (self->hf_flags & _HF_ERROR) ? -1 : rv;
1533 }
1534
1535 /*
1536  * TODO Because we don't support filter chains yet this filter will be run
1537  * TODO in a dedicated subprocess, driven via a special Popen() mode
1538  */
1539 static bool_t __hf_hadpipesig;
1540 static void
1541 __hf_onpipe(int signo)
1542 {
1543    NYD_X; /* Signal handler */
1544    UNUSED(signo);
1545    __hf_hadpipesig = TRU1;
1546 }
1547
1548 FL int
1549 htmlflt_process_main(void)
1550 {
1551    char buf[BUFFER_SIZE];
1552    struct htmlflt hf;
1553    size_t i;
1554    int rv;
1555    NYD_ENTER;
1556
1557    __hf_hadpipesig = FAL0;
1558    safe_signal(SIGPIPE, &__hf_onpipe);
1559
1560    htmlflt_init(&hf);
1561    htmlflt_reset(&hf, stdout);
1562
1563    for (;;) {
1564       if ((i = fread(buf, sizeof(buf[0]), NELEM(buf), stdin)) == 0) {
1565          rv = !feof(stdin);
1566          break;
1567       }
1568
1569       if ((rv = __hf_hadpipesig))
1570          break;
1571       /* Just use this directly.. */
1572       if (htmlflt_push(&hf, buf, i) < 0) {
1573          rv = 1;
1574          break;
1575       }
1576    }
1577    if (rv == 0 && htmlflt_flush(&hf) < 0)
1578       rv = 1;
1579
1580    htmlflt_destroy(&hf);
1581
1582    rv |= __hf_hadpipesig;
1583    NYD_LEAVE;
1584    return rv;
1585 }
1586
1587 FL void
1588 htmlflt_init(struct htmlflt *self)
1589 {
1590    NYD_ENTER;
1591    /* (Rather redundant though) */
1592    memset(self, 0, sizeof *self);
1593    NYD_LEAVE;
1594 }
1595
1596 FL void
1597 htmlflt_destroy(struct htmlflt *self)
1598 {
1599    NYD_ENTER;
1600    htmlflt_reset(self, NULL);
1601    NYD_LEAVE;
1602 }
1603
1604 FL void
1605 htmlflt_reset(struct htmlflt *self, FILE *f)
1606 {
1607    struct htmlflt_href *hfhp;
1608    NYD_ENTER;
1609
1610    while ((hfhp = self->hf_hrefs) != NULL) {
1611       self->hf_hrefs = hfhp->hfh_next;
1612       free(hfhp);
1613    }
1614
1615    if (self->hf_bdat != NULL)
1616       free(self->hf_bdat);
1617    if (self->hf_line != NULL)
1618       free(self->hf_line);
1619
1620    memset(self, 0, sizeof *self);
1621
1622    if (f != NULL) {
1623       ui32_t sw = MAX(_HF_MINLEN, (ui32_t)scrnwidth);
1624
1625       self->hf_line = smalloc((size_t)sw * mb_cur_max +1);
1626       self->hf_lmax = sw;
1627
1628       if (options & OPT_UNICODE) /* TODO not truly generic */
1629          self->hf_flags = _HF_UTF8;
1630       self->hf_os = f;
1631    }
1632    NYD_LEAVE;
1633 }
1634
1635 FL ssize_t
1636 htmlflt_push(struct htmlflt *self, char const *dat, size_t len)
1637 {
1638    ssize_t rv;
1639    NYD_ENTER;
1640
1641    rv = _hf_add_data(self, dat, len);
1642    NYD_LEAVE;
1643    return rv;
1644 }
1645
1646 FL ssize_t
1647 htmlflt_flush(struct htmlflt *self)
1648 {
1649    ssize_t rv;
1650    NYD_ENTER;
1651
1652    rv = _hf_add_data(self, NULL, 0);
1653    rv |= !fflush(self->hf_os) ? 0 : -1;
1654    NYD_LEAVE;
1655    return rv;
1656 }
1657 #endif /* HAVE_FILTER_HTML_TAGSOUP */
1658
1659 /* s-it-mode */