X-Git-Url: https://diplodocus.org/git/nmh/blobdiff_plain/448debd081b8bfad38ba43b21bea34a9c0eb49c2..17378602e0eed3c39e5f5ba9f3b28274b3dd26b2:/sbr/encode_rfc2047.c?ds=sidebyside diff --git a/sbr/encode_rfc2047.c b/sbr/encode_rfc2047.c index 4cacd837..04a74f1a 100644 --- a/sbr/encode_rfc2047.c +++ b/sbr/encode_rfc2047.c @@ -8,6 +8,7 @@ #include #include +#include #include /* @@ -35,15 +36,25 @@ static char *address_headers[] = { * Macros we use for parsing headers */ -#define is_fws(c) (c == '\t' || c == ' ') +#define is_fws(c) (c == '\t' || c == ' ' || c == '\n') -#define qpspecial(c) (c < ' ' || c == '=' && c == '?' && c == '_') +#define qphrasevalid(c) ((c >= '0' && c <= '9') || (c >= 'A' && c <= 'Z') || \ + (c >= 'a' && c <= 'z') || \ + c == '!' || c == '*' || c == '+' || c == '-' || \ + c == '/' || c == '=' || c == '_') +#define qpspecial(c) (c < ' ' || c == '=' || c == '?' || c == '_') + +#define base64len(n) ((((n) + 2) / 3 ) * 4) /* String len to base64 len */ +#define strbase64(n) ((n) / 4 * 3) /* Chars that fit in base64 */ #define ENCODELINELIMIT 76 static void unfold_header(char **, int); static int field_encode_address(const char *, char **, int, const char *); -static int field_encode_quoted(const char *, char **, const char *, int, int); +static int field_encode_quoted(const char *, char **, const char *, int, + int, int); +static int field_encode_base64(const char *, char **, const char *); +static int scanstring(const char *, int *, int *, int *); static int utf8len(const char *); /* @@ -64,9 +75,9 @@ encode_rfc2047(const char *name, char **value, int encoding, */ for (p = *value; *p != '\0'; p++) { - if (isascii((int) *p)) { + if (isascii((unsigned char) *p)) { asciicount++; - if (qpspecial(*p)) + if (qpspecial((unsigned char) *p)) qpspecialcount++; } else eightbitcount++; @@ -126,14 +137,12 @@ encode_rfc2047(const char *name, char **value, int encoding, switch (encoding) { -#if 0 case CE_BASE64: - return field_encode_base64(name, value, encoding, charset); -#endif + return field_encode_base64(name, value, charset); case CE_QUOTED: return field_encode_quoted(name, value, charset, asciicount, - eightbitcount + qpspecialcount); + eightbitcount + qpspecialcount, 0); default: advise(NULL, "Internal error: unknown RFC-2047 encoding type"); @@ -142,12 +151,12 @@ encode_rfc2047(const char *name, char **value, int encoding, } /* - * Encode our specified header using quoted-printable + * Encode our specified header (or field) using quoted-printable */ static int field_encode_quoted(const char *name, char **value, const char *charset, - int ascii, int encoded) + int ascii, int encoded, int phraserules) { int prefixlen = name ? strlen(name) + 2: 0, outlen = 0, column, newline = 1; int charsetlen = strlen(charset), utf8; @@ -180,6 +189,8 @@ field_encode_quoted(const char *name, char **value, const char *charset, * */ + int tokenlen; + outlen += 9 + charsetlen + ascii + 3 * encoded; /* @@ -209,13 +220,17 @@ field_encode_quoted(const char *name, char **value, const char *charset, *q++ = *p++; } - q += snprintf(q, outlen - (q - output), "=?%s?Q?", charset); - column = prefixlen; + tokenlen = snprintf(q, outlen - (q - output), "=?%s?Q?", charset); + q += tokenlen; + column = prefixlen + tokenlen; newline = 0; } /* * Process each character, encoding if necessary + * + * Note that we have a different set of rules if we're processing + * RFC 5322 'phrase' (something you'd see in an address header). */ column++; @@ -223,13 +238,15 @@ field_encode_quoted(const char *name, char **value, const char *charset, if (*p == ' ') { *q++ = '_'; ascii--; - } else if (!qpspecial(*p)) { + } else if (isascii((unsigned char) *p) && + (phraserules ? qphrasevalid((unsigned char) *p) : + !qpspecial((unsigned char) *p))) { *q++ = *p; ascii--; } else { - snprintf(q, outlen - (q - output), "=%02X", *((unsigned char *) p)); + snprintf(q, outlen - (q - output), "=%02X", (unsigned char) *p); q += 3; - column += 3; + column += 2; /* column already incremented by 1 above */ encoded--; } @@ -268,7 +285,179 @@ field_encode_quoted(const char *name, char **value, const char *charset, } } - strcat(q, "?=\n"); + strcat(q, "?="); + + if (prefixlen) + strcat(q, "\n"); + + free(*value); + + *value = output; + + return 0; +} + +/* + * Encode our specified header (or field) using base64. + * + * This is a little easier since every character gets encoded, we can + * calculate the line wrap up front. + */ + +static int +field_encode_base64(const char *name, char **value, const char *charset) +{ + int prefixlen = name ? strlen(name) + 2 : 0, charsetlen = strlen(charset); + int outlen = 0, numencode, curlen; + char *output = NULL, *p = *value, *q = NULL, *linestart; + + /* + * Skip over any leading white space. + */ + + while (*p == ' ' || *p == '\t') + p++; + + /* + * If we had a zero-length prefix, then just encode the whole field + * as-is, without line wrapping. Note that in addition to the encoding + * + * The added length we need is =? + charset + ?B? ... ?= + * + * That's 7 + strlen(charset) + 2 (for \n NUL). + */ + + while (prefixlen && ((base64len(strlen(p)) + 7 + charsetlen + + prefixlen) > ENCODELINELIMIT)) { + + /* + * Our very first time, don't pad the line in the front + * + * Note ENCODELINELIMIT is + 2 because of \n \0 + */ + + + if (! output) { + outlen += ENCODELINELIMIT + 2; + output = q = mh_xmalloc(outlen); + linestart = q - prefixlen; /* Yes, this is intentional */ + } else { + int curstart = linestart - output; + curlen = q - output; + + outlen += ENCODELINELIMIT + 2; + output = mh_xrealloc(output, outlen); + q = output + curlen; + linestart = output + curstart; + } + + /* + * We should have enough space now, so prepend the encoding markers + * and character set information. The leading space is intentional. + */ + + q += snprintf(q, outlen - (q - output), " =?%s?B?", charset); + + /* + * Find out how much room we have left on the line and see how + * many characters we can stuff in. The start of our line + * is marked by "linestart", so use that to figure out how + * many characters are left out of ENCODELINELIMIT. Reserve + * 2 characters for the end markers and calculate how many + * characters we can fit into that space given the base64 + * encoding expansion. + */ + + numencode = strbase64(ENCODELINELIMIT - (q - linestart) - 2); + + if (numencode <= 0) { + advise(NULL, "Internal error: tried to encode %d characters " + "in base64", numencode); + return 1; + } + + /* + * RFC 2047 prohibits spanning multibyte characters across tokens. + * Right now we only check for UTF-8. + * + * So note the key here ... we want to make sure the character BEYOND + * our last character is not a continuation byte. If it's the start + * of a new multibyte character or a single-byte character, that's ok. + */ + + if (strcasecmp(charset, "UTF-8") == 0) { + /* + * p points to the start of our current buffer, so p + numencode + * is one past the last character to encode + */ + + while (numencode > 0 && ((*(p + numencode) & 0xc0) == 0x80)) + numencode--; + + if (numencode == 0) { + advise(NULL, "Internal error: could not find start of " + "UTF-8 character when base64 encoding header"); + return 1; + } + } + + if (writeBase64raw((unsigned char *) p, numencode, + (unsigned char *) q) != OK) { + advise(NULL, "Internal error: base64 encoding of header failed"); + return 1; + } + + p += numencode; + q += base64len(numencode); + + /* + * This will point us at the beginning of the new line (trust me). + */ + + linestart = q + 3; + + /* + * What's going on here? Well, we know we're continuing to the next + * line, so we want to add continuation padding. We also add the + * trailing marker for the RFC 2047 token at this time as well. + * This uses a trick of snprintf(); we tell it to print a zero-length + * string, but pad it out to prefixlen - 1 characters; that ends + * up always printing out the requested number of spaces. We use + * prefixlen - 1 because we always add a space on the starting + * token marker; this makes things work out correctly for the first + * line, which should have a space between the ':' and the start + * of the token. + * + * It's okay if you don't follow all of that. + */ + + q += snprintf(q, outlen - (q - output), "?=\n%*s", prefixlen - 1, ""); + } + + /* + * We're here if there is either no prefix, or we can fit it in less + * than ENCODELINELIMIT characters. Encode the whole thing. + */ + + outlen += prefixlen + 9 + charsetlen + base64len(strlen(p)); + curlen = q - output; + + output = mh_xrealloc(output, outlen); + q = output + curlen; + + q += snprintf(q, outlen - (q - output), "%s=?%s?B?", + prefixlen ? " " : "", charset); + + if (writeBase64raw((unsigned char *) p, strlen(p), + (unsigned char *) q) != OK) { + advise(NULL, "Internal error: base64 encoding of header failed"); + return 1; + } + + strcat(q, "?="); + + if (prefixlen) + strcat(q, "\n"); free(*value); @@ -292,11 +481,11 @@ utf8len(const char *p) if (*p == '\0') return 0; - if (isascii((int) *p) || (*((unsigned char *) p) & 0xc0) == 0x80) + if (isascii((unsigned char) *p) || (((unsigned char) *p) & 0xc0) == 0x80) return 0; p++; - while ((*((unsigned char *) p++) & 0xc0) == 0x80) + while ((((unsigned char) *p++) & 0xc0) == 0x80) len++; return len; @@ -341,9 +530,68 @@ unfold_header(char **value, int len) *value = str; } +/* + * Decode a header containing addresses. This means we have to parse + * each address and only encode the display-name or comment field. + */ + static int field_encode_address(const char *name, char **value, int encoding, const char *charset) { - return 0; + int prefixlen = strlen(name) + 2, column = prefixlen, groupflag, errflag; + int eightbitchars; + char *mp, *output = NULL; + struct mailname *mn; + + /* + * Because these are addresses, we need to handle them individually. + * + * Break them down and process them one by one. This means we have to + * rewrite the whole header, but that's unavoidable. + */ + + /* + * The output headers always have to start with a space first. + */ + + output = add(" ", output); + + for (groupflag = 0; mp = getname(*value); ) { + if ((mn = getm(mp, NULL, 0, AD_HOST, NULL)) == NULL) { + errflag++; + continue; + } + + /* + * We only care if the phrase (m_pers) or any trailing comment + * (m_note) have 8-bit characters. If doing q-p, we also need + * to encode anything marked as qspecial(). + */ + } +} + +/* + * Scan a string, check for characters that need to be encoded + */ + +static int +scanstring(const char *string, int *asciilen, int *eightbitchars, + int *specialchars) +{ + *asciilen = 0; + *eightbitchars = 0; + *specialchars = 0; + + for (; *string != '\0'; string++) { + if ((isascii((unsigned char) *string))) { + (*asciilen++); + if (!qphrasevalid((unsigned char) *string)) + (*specialchars)++; + } else { + (*eightbitchars)++; + } + } + + return eightbitchars > 0; }