mime_enc.c

   1 /*@ S-nail - a mail user agent derived from Berkeley Mail.
   2  *@ Content-Transfer-Encodings as defined in RFC 2045 (and RFC 2047):
   3  *@ - Quoted-Printable, section 6.7
   4  *@ - Base64, section 6.8
   5  *
   6  * Copyright (c) 2000-2004 Gunnar Ritter, Freiburg i. Br., Germany.
   7  * Copyright (c) 2012 - 2015 Steffen (Daode) Nurpmeso <sdaoden@users.sf.net>.
   8  */
   9 /* QP quoting idea, _b64_decode(), b64_encode() taken from NetBSDs mailx(1): */
  10 /* $NetBSD: mime_codecs.c,v 1.9 2009/04/10 13:08:25 christos Exp $ */
  11 /*
  12  * Copyright (c) 2006 The NetBSD Foundation, Inc.
  13  * All rights reserved.
  14  *
  15  * This code is derived from software contributed to The NetBSD Foundation
  16  * by Anon Ymous.
  17  *
  18  * Redistribution and use in source and binary forms, with or without
  19  * modification, are permitted provided that the following conditions
  20  * are met:
  21  * 1. Redistributions of source code must retain the above copyright
  22  *    notice, this list of conditions and the following disclaimer.
  23  * 2. Redistributions in binary form must reproduce the above copyright
  24  *    notice, this list of conditions and the following disclaimer in the
  25  *    documentation and/or other materials provided with the distribution.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
  28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
  31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  37  * POSSIBILITY OF SUCH DAMAGE.
  38  */
  39 #undef n_FILE
  40 #define n_FILE mime_enc
  41
  42 #ifndef HAVE_AMALGAMATION
  43 # include "nail.h"
  44 #endif
  45
  46 enum _qact {
  47     N =   0,   /* Do not quote */
  48     Q =   1,   /* Must quote */
  49    SP =   2,   /* sp */
  50    XF =   3,   /* Special character 'F' - maybe quoted */
  51    XD =   4,   /* Special character '.' - maybe quoted */
  52    UU =   5,   /* In header, _ must be quoted in encoded word */
  53    US = '_',   /* In header, ' ' must be quoted as _ in encoded word */
  54    QM = '?',   /* In header, special character ? not always quoted */
  55    EQ = '=',   /* In header, '=' must be quoted in encoded word */
  56    HT ='\t',   /* In body HT=SP, in head HT=HT, but quote in encoded word */
  57    NL =   N,   /* Don't quote '\n' (NL) */
  58    CR =   Q    /* Always quote a '\r' (CR) */
  59 };
  60
  61 /* Lookup tables to decide wether a character must be encoded or not.
  62  * Email header differences according to RFC 2047, section 4.2:
  63  * - also quote SP (as the underscore _), TAB, ?, _, CR, LF
  64  * - don't care about the special ^F[rom] and ^.$ */
  65 static ui8_t const         _qtab_body[] = {
  66     Q, Q, Q, Q,  Q, Q, Q, Q,  Q,SP,NL, Q,  Q,CR, Q, Q,
  67     Q, Q, Q, Q,  Q, Q, Q, Q,  Q, Q, Q, Q,  Q, Q, Q, Q,
  68    SP, N, N, N,  N, N, N, N,  N, N, N, N,  N, N,XD, N,
  69     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, Q, N, N,
  70
  71     N, N, N, N,  N, N,XF, N,  N, N, N, N,  N, N, N, N,
  72     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N, N,
  73     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N, N,
  74     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N, Q,
  75 },
  76                            _qtab_head[] = {
  77     Q, Q, Q, Q,  Q, Q, Q, Q,  Q,HT, Q, Q,  Q, Q, Q, Q,
  78     Q, Q, Q, Q,  Q, Q, Q, Q,  Q, Q, Q, Q,  Q, Q, Q, Q,
  79    US, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N, N,
  80     N, N, N, N,  N, N, N, N,  N, N, N, N,  N,EQ, N,QM,
  81
  82     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N, N,
  83     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N,UU,
  84     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N, N,
  85     N, N, N, N,  N, N, N, N,  N, N, N, N,  N, N, N, Q,
  86 };
  87
  88 /* For decoding be robust and allow lowercase letters, too */
  89 static char const          _qp_itoa16[] = "0123456789ABCDEF";
  90 static ui8_t const         _qp_atoi16[] = {
  91    0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, /* 0x30-0x37 */
  92    0x08, 0x09, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, /* 0x38-0x3F */
  93    0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0xFF, /* 0x40-0x47 */
  94    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, /* 0x48-0x4f */
  95    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, /* 0x50-0x57 */
  96    0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, /* 0x58-0x5f */
  97    0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0xFF  /* 0x60-0x67 */
  98 };
  99
 100 /* The decoding table is only accessed via _B64_DECUI8() */
 101 static char const          _b64_enctbl[] =
 102     "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
 103 static signed char const   _b64__dectbl[] = {
 104    -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1,
 105    -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1,
 106    -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,62, -1,-1,-1,63,
 107    52,53,54,55, 56,57,58,59, 60,61,-1,-1, -1,-2,-1,-1,
 108    -1, 0, 1, 2,  3, 4, 5, 6,  7, 8, 9,10, 11,12,13,14,
 109    15,16,17,18, 19,20,21,22, 23,24,25,-1, -1,-1,-1,-1,
 110    -1,26,27,28, 29,30,31,32, 33,34,35,36, 37,38,39,40,
 111    41,42,43,44, 45,46,47,48, 49,50,51,-1, -1,-1,-1,-1
 112 };
 113 #define _B64_EQU           (ui32_t)-2
 114 #define _B64_BAD           (ui32_t)-1
 115 #define _B64_DECUI8(C)     \
 116    ((C) >= sizeof(_b64__dectbl) ? _B64_BAD : (ui32_t)_b64__dectbl[(ui8_t)(C)])
 117
 118 /* ASCII case-insensitive check wether Content-Transfer-Encoding: header body
 119  * hbody defined this encoding type */
 120 static bool_t        _is_ct_enc(char const *hbody, char const *encoding);
 121
 122 /* Check wether *s must be quoted according to flags, else body rules;
 123  * sol indicates wether we are at the first character of a line/field */
 124 SINLINE enum _qact   _mustquote(char const *s, char const *e, bool_t sol,
 125                         enum mime_enc_flags flags);
 126
 127 /* Convert c to/from a hexadecimal character string */
 128 SINLINE char *       _qp_ctohex(char *store, char c);
 129 SINLINE si32_t       _qp_cfromhex(char const *hex);
 130
 131 /* Trim WS and make work point to the decodable range of in*
 132  * Return the amount of bytes a b64_decode operation on that buffer requires */
 133 static size_t        _b64_decode_prepare(struct str *work,
 134                         struct str const *in);
 135
 136 /* Perform b64_decode on sufficiently spaced & multiple-of-4 base in(put).
 137  * Return number of useful bytes in out or -1 on error */
 138 static ssize_t       _b64_decode(struct str *out, struct str *in);
 139
 140 static bool_t
 141 _is_ct_enc(char const *hbody, char const *encoding)
 142 {
 143    bool_t quoted, rv;
 144    int c;
 145    NYD2_ENTER;
 146
 147    if (*hbody == '"')
 148       quoted = TRU1, ++hbody;
 149    else
 150       quoted = FAL0;
 151    rv = FAL0;
 152
 153    while (*hbody != '\0' && *encoding != '\0')
 154       if ((c = *hbody++, lowerconv(c) != *encoding++))
 155          goto jleave;
 156    rv = TRU1;
 157
 158    if (quoted && *hbody == '"')
 159       goto jleave;
 160    if (*hbody == '\0' || whitechar(*hbody))
 161       goto jleave;
 162    rv = FAL0;
 163 jleave:
 164    NYD2_LEAVE;
 165    return rv;
 166 }
 167
 168 SINLINE enum _qact
 169 _mustquote(char const *s, char const *e, bool_t sol, enum mime_enc_flags flags)
 170 {
 171    ui8_t const *qtab;
 172    enum _qact a, r;
 173    NYD2_ENTER;
 174
 175    qtab = (flags & (MIMEEF_ISHEAD | MIMEEF_ISENCWORD))
 176          ? _qtab_head : _qtab_body;
 177    a = ((ui8_t)*s > 0x7F) ? Q : qtab[(ui8_t)*s];
 178
 179    if ((r = a) == N || (r = a) == Q)
 180       goto jleave;
 181    r = Q;
 182
 183    /* Special header fields */
 184    if (flags & (MIMEEF_ISHEAD | MIMEEF_ISENCWORD)) {
 185       /* Special massage for encoded words */
 186       if (flags & MIMEEF_ISENCWORD) {
 187          switch (a) {
 188          case HT:
 189          case US:
 190          case EQ:
 191             r = a;
 192             /* FALLTHRU */
 193          case UU:
 194             goto jleave;
 195          default:
 196             break;
 197          }
 198       }
 199
 200       /* Treat '?' only special if part of '=?' .. '?=' (still too much quoting
 201        * since it's '=?CHARSET?CTE?stuff?=', and especially the trailing ?=
 202        * should be hard too match */
 203       if (a == QM && ((!sol && s[-1] == '=') || (s < e && s[1] == '=')))
 204          goto jleave;
 205       goto jnquote;
 206    }
 207
 208    /* Body-only */
 209
 210    if (a == SP) {
 211       /* WS only if trailing white space */
 212       if (PTRCMP(s + 1, ==, e) || s[1] == '\n')
 213          goto jleave;
 214       goto jnquote;
 215    }
 216
 217    /* Rest are special begin-of-line cases */
 218    if (!sol)
 219       goto jnquote;
 220
 221    /* ^From */
 222    if (a == XF) {
 223       if (PTRCMP(s + 4, <, e) && s[1] == 'r' && s[2] == 'o' && s[3] == 'm')
 224          goto jleave;
 225       goto jnquote;
 226    }
 227    /* ^.$ */
 228    if (a == XD && (PTRCMP(s + 1, ==, e) || s[1] == '\n'))
 229       goto jleave;
 230 jnquote:
 231    r = N;
 232 jleave:
 233    NYD2_LEAVE;
 234    return r;
 235 }
 236
 237 SINLINE char *
 238 _qp_ctohex(char *store, char c)
 239 {
 240    NYD2_ENTER;
 241    store[2] = '\0';
 242    store[1] = _qp_itoa16[(ui8_t)c & 0x0F];
 243    c = ((ui8_t)c >> 4) & 0x0F;
 244    store[0] = _qp_itoa16[(ui8_t)c];
 245    NYD2_LEAVE;
 246    return store;
 247 }
 248
 249 SINLINE si32_t
 250 _qp_cfromhex(char const *hex)
 251 {
 252    ui8_t i1, i2;
 253    si32_t rv;
 254    NYD2_ENTER;
 255
 256    if ((i1 = (ui8_t)hex[0] - '0') >= NELEM(_qp_atoi16) ||
 257          (i2 = (ui8_t)hex[1] - '0') >= NELEM(_qp_atoi16))
 258       goto jerr;
 259    i1 = _qp_atoi16[i1];
 260    i2 = _qp_atoi16[i2];
 261    if ((i1 | i2) & 0xF0u)
 262       goto jerr;
 263    rv = i1;
 264    rv <<= 4;
 265    rv += i2;
 266 jleave:
 267    NYD2_LEAVE;
 268    return rv;
 269 jerr:
 270    rv = -1;
 271    goto jleave;
 272 }
 273
 274 static size_t
 275 _b64_decode_prepare(struct str *work, struct str const *in)
 276 {
 277    char *cp;
 278    size_t cp_len;
 279    NYD2_ENTER;
 280
 281    cp = in->s;
 282    cp_len = in->l;
 283
 284    while (cp_len > 0 && spacechar(*cp))
 285       ++cp, --cp_len;
 286    work->s = cp;
 287
 288    for (cp += cp_len; cp_len > 0; --cp_len) {
 289       char c = *--cp;
 290       if (!spacechar(c))
 291          break;
 292    }
 293    work->l = cp_len;
 294
 295    if (cp_len > 16)
 296       cp_len = ((cp_len * 3) >> 2) + (cp_len >> 3);
 297    NYD2_LEAVE;
 298    return cp_len;
 299 }
 300
 301 static ssize_t
 302 _b64_decode(struct str *out, struct str *in)
 303 {
 304    ssize_t rv = -1;
 305    ui8_t *p;
 306    ui8_t const *q, *end;
 307    NYD2_ENTER;
 308
 309    p = (ui8_t*)out->s + out->l;
 310    q = (ui8_t const*)in->s;
 311
 312    for (end = q + in->l; PTRCMP(q + 4, <=, end);) {
 313       ui32_t a = _B64_DECUI8(q[0]), b = _B64_DECUI8(q[1]),
 314          c = _B64_DECUI8(q[2]), d = _B64_DECUI8(q[3]);
 315       q += 4;
 316
 317       if (a >= _B64_EQU || b >= _B64_EQU || c == _B64_BAD || d == _B64_BAD)
 318          goto jleave;
 319
 320       *p++ = ((a << 2) | ((b & 0x30) >> 4));
 321       if (c == _B64_EQU)  { /* got '=' */
 322          if (d != _B64_EQU)
 323             goto jleave;
 324          break;
 325       }
 326       *p++ = (((b & 0x0F) << 4) | ((c & 0x3C) >> 2));
 327       if (d == _B64_EQU) /* got '=' */
 328          break;
 329       *p++ = (((c & 0x03) << 6) | d);
 330    }
 331    rv ^= rv;
 332
 333 jleave: {
 334       size_t i = PTR2SIZE((char*)p - out->s);
 335       out->l = i;
 336       if (rv == 0)
 337          rv = (ssize_t)i;
 338    }
 339    in->l -= PTR2SIZE((char*)UNCONST(q) - in->s);
 340    in->s = UNCONST(q);
 341    NYD2_LEAVE;
 342    return rv;
 343 }
 344
 345 FL char *
 346 mime_char_to_hexseq(char store[3], char c)
 347 {
 348    char *rv;
 349    NYD2_ENTER;
 350
 351    rv = _qp_ctohex(store, c);
 352    NYD2_LEAVE;
 353    return rv;
 354 }
 355
 356 FL si32_t
 357 mime_hexseq_to_char(char const *hex)
 358 {
 359    si32_t rv;
 360    NYD2_ENTER;
 361
 362    rv = _qp_cfromhex(hex);
 363    NYD2_LEAVE;
 364    return rv;
 365 }
 366
 367 FL enum mime_enc
 368 mime_enc_target(void)
 369 {
 370    char const *cp;
 371    enum mime_enc rv;
 372    NYD2_ENTER;
 373
 374    if ((cp = ok_vlook(encoding)) == NULL)
 375       rv = MIME_DEFAULT_ENCODING;
 376    else if (!asccasecmp(cp, "quoted-printable"))
 377       rv = MIMEE_QP;
 378    else if (!asccasecmp(cp, "8bit"))
 379       rv = MIMEE_8B;
 380    else if (!asccasecmp(cp, "base64"))
 381       rv = MIMEE_B64;
 382    else {
 383       n_err(_("Warning: invalid *encoding*, using Base64: \"%s\"\n"), cp);
 384       rv = MIMEE_B64;
 385    }
 386    NYD2_LEAVE;
 387    return rv;
 388 }
 389
 390 FL enum mime_enc
 391 mime_enc_from_ctehead(char const *hbody)
 392 {
 393    enum mime_enc rv;
 394    NYD2_ENTER;
 395
 396    if (hbody == NULL || _is_ct_enc(hbody, "7bit"))
 397       rv = MIMEE_7B;
 398    else if (_is_ct_enc(hbody, "8bit"))
 399       rv = MIMEE_8B;
 400    else if (_is_ct_enc(hbody, "base64"))
 401       rv = MIMEE_B64;
 402    else if (_is_ct_enc(hbody, "binary"))
 403       rv = MIMEE_BIN;
 404    else if (_is_ct_enc(hbody, "quoted-printable"))
 405       rv = MIMEE_QP;
 406    else
 407       rv = MIMEE_NONE;
 408    NYD2_LEAVE;
 409    return rv;
 410 }
 411
 412 FL char const *
 413 mime_enc_from_conversion(enum conversion const convert) /* TODO booom */
 414 {
 415    char const *rv;
 416    NYD_ENTER;
 417
 418    switch (convert) {
 419    case CONV_7BIT:   rv = "7bit"; break;
 420    case CONV_8BIT:   rv = "8bit"; break;
 421    case CONV_TOQP:   rv = "quoted-printable"; break;
 422    case CONV_TOB64:  rv = "base64"; break;
 423    default:          rv = ""; break;
 424    }
 425    NYD_LEAVE;
 426    return rv;
 427 }
 428
 429 FL size_t
 430 mime_enc_mustquote(char const *ln, size_t lnlen, enum mime_enc_flags flags)
 431 {
 432    size_t rv;
 433    bool_t sol;
 434    NYD_ENTER;
 435
 436    for (rv = 0, sol = TRU1; lnlen > 0; sol = FAL0, ++ln, --lnlen)
 437       switch (_mustquote(ln, ln + lnlen, sol, flags)) {
 438       case US:
 439       case EQ:
 440       case HT:
 441          assert(flags & MIMEEF_ISENCWORD);
 442          /* FALLTHRU */
 443       case N:
 444          continue;
 445       default:
 446          ++rv;
 447       }
 448    NYD_LEAVE;
 449    return rv;
 450 }
 451
 452 FL size_t
 453 qp_encode_calc_size(size_t len)
 454 {
 455    size_t bytes, lines;
 456    NYD_ENTER;
 457
 458    /* The worst case sequence is 'CRLF' -> '=0D=0A=\n\0'.
 459     * However, we must be aware that (a) the output may span multiple lines
 460     * and (b) the input does not end with a newline itself (nonetheless):
 461     *    LC_ALL=C awk 'BEGIN{
 462     *       for (i = 1; i < 100000; ++i) printf "\xC3\xBC"
 463     *    }' |
 464     *    MAILRC=/dev/null LC_ALL=en_US.UTF-8 s-nail -nvvd \
 465     *       -Ssendcharsets=utf8 -s testsub ./LETTER */
 466    bytes = len * 3;
 467    lines = bytes / QP_LINESIZE;
 468    len += lines;
 469
 470    bytes = len * 3;
 471    /* Trailing hard NL may be missing, so there may be two lines.
 472     * Thus add soft + hard NL per line and a trailing NUL */
 473    lines = (bytes / QP_LINESIZE) + 1;
 474    lines <<= 1;
 475    bytes += lines;
 476    len = ++bytes;
 477
 478    NYD_LEAVE;
 479    return len;
 480 }
 481
 482 #ifdef notyet
 483 FL struct str *
 484 qp_encode_cp(struct str *out, char const *cp, enum qpflags flags)
 485 {
 486    struct str in;
 487    NYD_ENTER;
 488
 489    in.s = UNCONST(cp);
 490    in.l = strlen(cp);
 491    out = qp_encode(out, &in, flags);
 492    NYD_LEAVE;
 493    return out;
 494 }
 495
 496 FL struct str *
 497 qp_encode_buf(struct str *out, void const *vp, size_t vp_len,
 498    enum qpflags flags)
 499 {
 500    struct str in;
 501    NYD_ENTER;
 502
 503    in.s = UNCONST(vp);
 504    in.l = vp_len;
 505    out = qp_encode(out, &in, flags);
 506    NYD_LEAVE;
 507    return out;
 508 }
 509 #endif /* notyet */
 510
 511 FL struct str *
 512 qp_encode(struct str *out, struct str const *in, enum qpflags flags)
 513 {
 514    bool_t sol = (flags & QP_ISHEAD ? FAL0 : TRU1), seenx;
 515    ssize_t lnlen;
 516    char *qp;
 517    char const *is, *ie;
 518    NYD_ENTER;
 519
 520    if (!(flags & QP_BUF)) {
 521       lnlen = qp_encode_calc_size(in->l);
 522       out->s = (flags & QP_SALLOC) ? salloc(lnlen) : srealloc(out->s, lnlen);
 523    }
 524    qp = out->s;
 525    is = in->s;
 526    ie = is + in->l;
 527
 528    /* QP_ISHEAD? */
 529    if (!sol) {
 530       enum mime_enc_flags ef = MIMEEF_ISHEAD |
 531             (flags & QP_ISENCWORD ? MIMEEF_ISENCWORD : 0);
 532
 533       for (seenx = FAL0, sol = TRU1; is < ie; sol = FAL0, ++qp) {
 534          enum _qact mq = _mustquote(is, ie, sol, ef);
 535          char c = *is++;
 536
 537          if (mq == N) {
 538             /* We convert into a single *encoded-word*, that'll end up in
 539              * =?C?Q??=; quote '?' from when we're inside there on */
 540             if (seenx && c == '?')
 541                goto jheadq;
 542             *qp = c;
 543          } else if (mq == US)
 544             *qp = US;
 545          else {
 546             seenx = TRU1;
 547 jheadq:
 548             *qp++ = '=';
 549             qp = _qp_ctohex(qp, c) + 1;
 550          }
 551       }
 552       goto jleave;
 553    }
 554
 555    /* The body needs to take care for soft line breaks etc. */
 556    for (lnlen = 0, seenx = FAL0; is < ie; sol = FAL0) {
 557       enum _qact mq = _mustquote(is, ie, sol, MIMEEF_NONE);
 558       char c = *is++;
 559
 560       if (mq == N && (c != '\n' || !seenx)) {
 561          *qp++ = c;
 562          if (++lnlen < QP_LINESIZE - 1)
 563             continue;
 564          /* Don't write a soft line break when we're in the last possible
 565           * column and either an LF has been written or only an LF follows, as
 566           * that'll end the line anyway */
 567          /* XXX but - ensure is+1>=ie, then??
 568           * xxx and/or - what about resetting lnlen; that contra
 569           * xxx dicts input==1 input line assertion, though */
 570          if (c == '\n' || is == ie || *is == '\n')
 571             continue;
 572 jsoftnl:
 573          qp[0] = '=';
 574          qp[1] = '\n';
 575          qp += 2;
 576          lnlen = 0;
 577          continue;
 578       }
 579
 580       if (lnlen > QP_LINESIZE - 3 - 1) {
 581          qp[0] = '=';
 582          qp[1] = '\n';
 583          qp += 2;
 584          lnlen = 0;
 585       }
 586       *qp++ = '=';
 587       qp = _qp_ctohex(qp, c);
 588       qp += 2;
 589       lnlen += 3;
 590       if (c != '\n' || !seenx)
 591          seenx = (c == '\r');
 592       else {
 593          seenx = FAL0;
 594          goto jsoftnl;
 595       }
 596    }
 597
 598    /* Enforce soft line break if we haven't seen LF */
 599    if (in->l > 0 && *--is != '\n') {
 600       qp[0] = '=';
 601       qp[1] = '\n';
 602       qp += 2;
 603    }
 604 jleave:
 605    out->l = PTR2SIZE(qp - out->s);
 606    out->s[out->l] = '\0';
 607    NYD_LEAVE;
 608    return out;
 609 }
 610
 611 FL int
 612 qp_decode(struct str *out, struct str const *in, struct str *rest)
 613 {
 614    int rv = STOP;
 615    char *os, *oc;
 616    char const *is, *ie;
 617    NYD_ENTER;
 618
 619    if (rest != NULL && rest->l != 0) {
 620       os = out->s;
 621       *out = *rest;
 622       rest->s = os;
 623       rest->l = 0;
 624    }
 625
 626    oc = os =
 627    out->s = srealloc(out->s, out->l + in->l + 3);
 628    oc += out->l;
 629    is = in->s;
 630    ie = is + in->l;
 631
 632    /* Decoding encoded-word (RFC 2049) in a header field? */
 633    if (rest == NULL) {
 634       while (is < ie) {
 635          si32_t c = *is++;
 636          if (c == '=') {
 637             if (PTRCMP(is + 1, >=, ie)) {
 638                ++is;
 639                goto jehead;
 640             }
 641             c = _qp_cfromhex(is);
 642             is += 2;
 643             if (c >= 0)
 644                *oc++ = (char)c;
 645             else {
 646                /* Invalid according to RFC 2045, section 6.7. Almost follow */
 647 jehead:
 648                /* TODO 0xFFFD
 649                *oc[0] = '['; oc[1] = '?'; oc[2] = ']';
 650                *oc += 3; 0xFFFD TODO
 651                */ *oc++ = '?';
 652             }
 653          } else
 654             *oc++ = (c == '_' /* US */) ? ' ' : (char)c;
 655       }
 656       goto jleave; /* XXX QP decode, header: errors not reported */
 657    }
 658
 659    /* Decoding a complete message/mimepart body line */
 660    while (is < ie) {
 661       si32_t c = *is++;
 662       if (c != '=') {
 663          *oc++ = (char)c;
 664          continue;
 665       }
 666
 667       /* RFC 2045, 6.7:
 668        *   Therefore, when decoding a Quoted-Printable body, any
 669        *   trailing white space on a line must be deleted, as it will
 670        *   necessarily have been added by intermediate transport
 671        *   agents */
 672       for (; is < ie && blankchar(*is); ++is)
 673          ;
 674       if (PTRCMP(is + 1, >=, ie)) {
 675          /* Soft line break? */
 676          if (*is == '\n')
 677             goto jsoftnl;
 678          ++is;
 679          goto jebody;
 680       }
 681
 682       /* Not a soft line break? */
 683       if (*is != '\n') {
 684          c = _qp_cfromhex(is);
 685          is += 2;
 686          if (c >= 0)
 687             *oc++ = (char)c;
 688          else {
 689             /* Invalid according to RFC 2045, section 6.7.
 690              * Almost follow it and include the = and the follow char */
 691 jebody:
 692             /* TODO 0xFFFD
 693             *oc[0] = '['; oc[1] = '?'; oc[2] = ']';
 694             *oc += 3; 0xFFFD TODO
 695             */ *oc++ = '?';
 696          }
 697          continue;
 698       }
 699
 700       /* CRLF line endings are encoded as QP, followed by a soft line break, so
 701        * check for this special case, and simply forget we have seen one, so as
 702        * not to end up with the entire DOS file in a contiguous buffer */
 703 jsoftnl:
 704       if (oc > os && oc[-1] == '\n') {
 705 #if 0       /* TODO qp_decode() we do not normalize CRLF
 706           * TODO to LF because for that we would need
 707           * TODO to know if we are about to write to
 708           * TODO the display or do save the file!
 709           * TODO 'hope the MIME/send layer rewrite will
 710           * TODO offer the possibility to DTRT */
 711          if (oc - 1 > os && oc[-2] == '\r') {
 712             --oc;
 713             oc[-1] = '\n';
 714          }
 715 #endif
 716          break;
 717       }
 718       out->l = PTR2SIZE(oc - os);
 719       rest->s = srealloc(rest->s, rest->l + out->l);
 720       memcpy(rest->s + rest->l, out->s, out->l);
 721       rest->l += out->l;
 722       oc = os;
 723       break;
 724    }
 725    /* XXX RFC: QP decode should check no trailing WS on line */
 726 jleave:
 727    out->l = PTR2SIZE(oc - os);
 728    rv = OKAY;
 729    NYD_LEAVE;
 730    return rv;
 731 }
 732
 733 FL size_t
 734 b64_encode_calc_size(size_t len)
 735 {
 736    NYD_ENTER;
 737    len = (len * 4) / 3;
 738    len += (((len / B64_ENCODE_INPUT_PER_LINE) + 1) * 3);
 739    len += 2 + 1; /* CRLF, \0 */
 740    NYD_LEAVE;
 741    return len;
 742 }
 743
 744 FL struct str *
 745 b64_encode(struct str *out, struct str const *in, enum b64flags flags)
 746 {
 747    ui8_t const *p;
 748    ssize_t i, lnlen;
 749    char *b64;
 750    NYD_ENTER;
 751
 752    p = (ui8_t const*)in->s;
 753
 754    if (!(flags & B64_BUF)) {
 755       i = b64_encode_calc_size(in->l);
 756       out->s = (flags & B64_SALLOC) ? salloc(i) : srealloc(out->s, i);
 757    }
 758    b64 = out->s;
 759
 760    if (!(flags & (B64_CRLF | B64_LF)))
 761       flags &= ~B64_MULTILINE;
 762
 763    for (lnlen = 0, i = (ssize_t)in->l; i > 0; p += 3, i -= 3) {
 764       ui32_t a = p[0], b, c;
 765
 766       b64[0] = _b64_enctbl[a >> 2];
 767       switch (i) {
 768       case 1:
 769          b64[1] = _b64_enctbl[((a & 0x3) << 4)];
 770          b64[2] =
 771          b64[3] = '=';
 772          break;
 773       case 2:
 774          b = p[1];
 775          b64[1] = _b64_enctbl[((a & 0x03) << 4) | ((b & 0xF0u) >> 4)];
 776          b64[2] = _b64_enctbl[((b & 0x0F) << 2)];
 777          b64[3] = '=';
 778          break;
 779       default:
 780          b = p[1];
 781          c = p[2];
 782          b64[1] = _b64_enctbl[((a & 0x03) << 4) | ((b & 0xF0u) >> 4)];
 783          b64[2] = _b64_enctbl[((b & 0x0F) << 2) | ((c & 0xC0u) >> 6)];
 784          b64[3] = _b64_enctbl[c & 0x3F];
 785          break;
 786       }
 787
 788       b64 += 4;
 789       if (!(flags & B64_MULTILINE))
 790          continue;
 791       lnlen += 4;
 792       if (lnlen < B64_LINESIZE)
 793          continue;
 794
 795       lnlen = 0;
 796       if (flags & B64_CRLF)
 797          *b64++ = '\r';
 798       if (flags & (B64_CRLF | B64_LF))
 799          *b64++ = '\n';
 800    }
 801
 802    if ((flags & (B64_CRLF | B64_LF)) &&
 803          (!(flags & B64_MULTILINE) || lnlen != 0)) {
 804       if (flags & B64_CRLF)
 805          *b64++ = '\r';
 806       if (flags & (B64_CRLF | B64_LF))
 807          *b64++ = '\n';
 808    }
 809    out->l = PTR2SIZE(b64 - out->s);
 810    out->s[out->l] = '\0';
 811
 812    /* Base64 includes + and /, replace them with _ and -.
 813     * This is base64url according to RFC 4648, then.  Since we only support
 814     * that for encoding and it is only used for boundary strings, this is
 815     * yet a primitive implementation; xxx use tables; support decoding */
 816    if (flags & B64_RFC4648URL) {
 817       char c;
 818
 819       for (b64 = out->s; (c = *b64) != '\0'; ++b64)
 820          if (c == '+')
 821             *b64 = '-';
 822          else if (c == '/')
 823             *b64 = '_';
 824    }
 825    NYD_LEAVE;
 826    return out;
 827 }
 828
 829 FL struct str *
 830 b64_encode_buf(struct str *out, void const *vp, size_t vp_len,
 831    enum b64flags flags)
 832 {
 833    struct str in;
 834    NYD_ENTER;
 835
 836    in.s = UNCONST(vp);
 837    in.l = vp_len;
 838    out = b64_encode(out, &in, flags);
 839    NYD_LEAVE;
 840    return out;
 841 }
 842
 843 #ifdef HAVE_SMTP
 844 FL struct str *
 845 b64_encode_cp(struct str *out, char const *cp, enum b64flags flags)
 846 {
 847    struct str in;
 848    NYD_ENTER;
 849
 850    in.s = UNCONST(cp);
 851    in.l = strlen(cp);
 852    out = b64_encode(out, &in, flags);
 853    NYD_LEAVE;
 854    return out;
 855 }
 856 #endif
 857
 858 FL int
 859 b64_decode(struct str *out, struct str const *in, struct str *rest)
 860 {
 861    struct str work;
 862    char *x;
 863    size_t len;
 864    int rv; /* XXX -> bool_t */
 865    NYD_ENTER;
 866
 867    len = _b64_decode_prepare(&work, in);
 868    out->l = 0;
 869
 870    /* TODO B64_T is different since we must not fail for errors; in v15.0 this
 871     * TODO will be filter based and B64_T will have a different one than B64,
 872     * TODO for now special treat this all-horror */
 873    if (rest != NULL) {
 874       /* With B64_T there may be leftover decoded data for iconv(3), even if
 875        * that means it's incomplete multibyte character we have to copy over */
 876       /* TODO strictly speaking this should not be handled in here,
 877        * TODO since its leftover decoded data from an iconv(3);
 878        * TODO In v15.0 this path will be filter based, each filter having its
 879        * TODO own buffer for such purpose; for now we are BUSTED since for
 880        * TODO Base64 rest is owned by iconv(3) */
 881       if (rest->l > 0) {
 882          x = out->s;
 883          *out = *rest;
 884          rest->s = x; /* Just for ownership reasons (all TODO in here..) */
 885          rest->l = 0;
 886          len += out->l;
 887       }
 888
 889       out->s = srealloc(out->s, len +1);
 890
 891       for (;;) {
 892          if (_b64_decode(out, &work) >= 0) {
 893             if (work.l == 0)
 894                break;
 895          }
 896          x = out->s + out->l;
 897
 898          /* Partial/False last sequence.  TODO not solvable for non-EOF;
 899           * TODO yes, invalid, but seen in the wild and should be handled,
 900           * TODO but for that we had to have our v15.0 filter which doesn't
 901           * TODO work line based but content buffer based */
 902          if ((len = work.l) <= 4) {
 903             switch (len) {
 904             case 4:  /* FALLTHRU */
 905             case 3:  x[2] = '?'; /* FALLTHRU */
 906             case 2:  x[1] = '?'; /* FALLTHRU */
 907             default: x[0] = '?'; break;
 908             }
 909             out->l += len;
 910             break;
 911          }
 912
 913          /* TODO Bad content: this problem is not solvable!  I've seen
 914           * TODO messages which broke lines in the middle of a Base64
 915           * TODO tuple, followed by an invalid character ("!"), the follow
 916           * TODO line starting with whitespace and the remaining sequence.
 917           * TODO OpenSSL bailed, mutt(1) got it right (silently..).
 918           * TODO Since "rest" is not usable by us, we cannot continue
 919           * TODO sequences.  We will be able to do so with the v15.0 filter
 920           * TODO approach, if we */
 921          /* Bad content: skip over a single sequence */
 922          for (;;) {
 923             *x++ = '?';
 924             ++out->l;
 925             if (--work.l == 0)
 926                break;
 927             else {
 928                ui8_t bc = (ui8_t)*++work.s;
 929                ui32_t state = _B64_DECUI8(bc);
 930
 931                if (state != _B64_EQU && state != _B64_BAD)
 932                   break;
 933             }
 934          }
 935       }
 936       rv = OKAY;
 937       goto jleave;
 938    }
 939
 940    /* Ignore an empty input, as may happen for an empty final line */
 941    if (work.l == 0) {
 942       out->s = srealloc(out->s, 1);
 943       rv = OKAY;
 944    } else if (work.l >= 4 && !(work.l & 3)) {
 945       out->s = srealloc(out->s, len +1);
 946       if ((ssize_t)(len = _b64_decode(out, &work)) < 0)
 947          goto jerr;
 948       rv = OKAY;
 949    } else
 950       goto jerr;
 951
 952 jleave:
 953    out->s[out->l] = '\0';
 954    NYD_LEAVE;
 955    return rv;
 956
 957 jerr: {
 958    char const *err = _("[Invalid Base64 encoding]\n");
 959    out->l = len = strlen(err);
 960    out->s = srealloc(out->s, len +1);
 961    memcpy(out->s, err, len);
 962    rv = STOP;
 963    goto jleave;
 964    }
 965 }
 966
 967 /* s-it-mode */