1 /*@ S-nail - a mail user agent derived from Berkeley Mail.
2 *@ MIME support functions.
3 *@ TODO Complete rewrite.
5 * Copyright (c) 2000-2004 Gunnar Ritter, Freiburg i. Br., Germany.
6 * Copyright (c) 2012 - 2017 Steffen (Daode) Nurpmeso <steffen@sdaoden.eu>.
10 * Gunnar Ritter. All rights reserved.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. All advertising materials mentioning features or use of this software
21 * must display the following acknowledgement:
22 * This product includes software developed by Gunnar Ritter
23 * and his contributors.
24 * 4. Neither the name of Gunnar Ritter nor the names of his contributors
25 * may be used to endorse or promote products derived from this software
26 * without specific prior written permission.
28 * THIS SOFTWARE IS PROVIDED BY GUNNAR RITTER AND CONTRIBUTORS ``AS IS'' AND
29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31 * ARE DISCLAIMED. IN NO EVENT SHALL GUNNAR RITTER OR CONTRIBUTORS BE LIABLE
32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
43 #ifndef HAVE_AMALGAMATION
47 /* Don't ask, but it keeps body and soul together */
48 enum a_mime_structure_hack
{
54 static char *_cs_iter_base
, *_cs_iter
;
56 # define _CS_ITER_GET() \
57 ((_cs_iter != NULL) ? _cs_iter : ok_vlook(CHARSET_8BIT_OKEY))
59 # define _CS_ITER_GET() ((_cs_iter != NULL) ? _cs_iter : ok_vlook(ttycharset))
61 #define _CS_ITER_STEP() _cs_iter = n_strsep(&_cs_iter_base, ',', TRU1)
63 /* Is 7-bit enough? */
65 static bool_t
_has_highbit(char const *s
);
66 static bool_t
_name_highbit(struct name
*np
);
69 /* fwrite(3) while checking for displayability */
70 static ssize_t
_fwrite_td(struct str
const *input
,
71 bool_t failiconv
, enum tdflags flags
,
72 struct str
*outrest
, struct quoteflt
*qf
);
74 /* Convert header fields to RFC 2047 format and write to the file fo */
75 static ssize_t
mime_write_tohdr(struct str
*in
, FILE *fo
,
76 size_t *colp
, enum a_mime_structure_hack msh
);
78 /* Write len characters of the passed string to the passed file, doing charset
79 * and header conversion */
81 /* Write an address to a header field */
82 static ssize_t
mime_write_tohdr_a(struct str
*in
, FILE *f
,
85 static ssize_t
a_mime__convhdra(struct str
*inp
, FILE *fp
, size_t *colp
,
86 enum a_mime_structure_hack msh
);
88 # define a_mime__convhdra(S,F,C,MSH) mime_write_tohdr(S, F, C, MSH)
91 /* Append to buf, handling resizing */
92 static void _append_str(char **buf
, size_t *sz
, size_t *pos
,
93 char const *str
, size_t len
);
94 static void _append_conv(char **buf
, size_t *sz
, size_t *pos
,
95 char const *str
, size_t len
);
99 _has_highbit(char const *s
)
106 if ((ui8_t
)*s
& 0x80)
108 while (*s
++ != '\0');
117 _name_highbit(struct name
*np
)
123 if (_has_highbit(np
->n_name
) || _has_highbit(np
->n_fullname
))
132 #endif /* HAVE_ICONV */
134 static sigjmp_buf __mimefwtd_actjmp
; /* TODO someday.. */
135 static int __mimefwtd_sig
; /* TODO someday.. */
136 static sighandler_type __mimefwtd_opipe
;
138 __mimefwtd_onsig(int sig
) /* TODO someday, we won't need it no more */
140 NYD_X
; /* Signal handler */
141 __mimefwtd_sig
= sig
;
142 siglongjmp(__mimefwtd_actjmp
, 1);
146 _fwrite_td(struct str
const *input
, bool_t failiconv
, enum tdflags flags
,
147 struct str
*outrest
, struct quoteflt
*qf
)
149 /* TODO note: after send/MIME layer rewrite we will have a string pool
150 * TODO so that memory allocation count drops down massively; for now,
151 * TODO v14.* that is, we pay a lot & heavily depend on the allocator */
152 /* TODO well if we get a broken pipe here, and it happens to
153 * TODO happen pretty easy when sleeping in a full pipe buffer,
154 * TODO then the current codebase performs longjump away;
155 * TODO this leaves memory leaks behind ('think up to 3 per,
156 * TODO dep. upon alloca availability). For this to be fixed
157 * TODO we either need to get rid of the longjmp()s (tm) or
158 * TODO the storage must come from the outside or be tracked
159 * TODO in a carrier struct. Best both. But storage reuse
160 * TODO would be a bigbig win besides */
161 /* *input* _may_ point to non-modifyable buffer; but even then it only
162 * needs to be dup'ed away if we have to transform the content */
174 if ((flags
& TD_ICONV
) && iconvd
!= (iconv_t
)-1) {
180 if (outrest
!= NULL
&& outrest
->l
> 0) {
181 in
.l
= outrest
->l
+ input
->l
;
182 in
.s
= buf
= smalloc(in
.l
+1);
183 memcpy(in
.s
, outrest
->s
, outrest
->l
);
184 memcpy(&in
.s
[outrest
->l
], input
->s
, input
->l
);
189 if((err
= n_iconv_str(iconvd
,
190 (failiconv
? n_ICONV_NONE
: n_ICONV_UNIDEFAULT
),
191 &out
, &in
, &in
)) != 0){
192 if(err
!= n_ERR_INVAL
)
193 n_iconv_reset(iconvd
);
195 if(outrest
!= NULL
&& in
.l
> 0){
196 /* Incomplete multibyte at EOF is special xxx _INVAL? */
197 if (flags
& _TD_EOF
) {
198 out
.s
= srealloc(out
.s
, out
.l
+ sizeof(n_unirepl
));
199 if(n_psonce
& n_PSO_UNICODE
){
200 memcpy(&out
.s
[out
.l
], n_unirepl
, sizeof(n_unirepl
) -1);
201 out
.l
+= sizeof(n_unirepl
) -1;
203 out
.s
[out
.l
++] = '?';
205 n_str_add(outrest
, &in
);
212 flags
&= ~_TD_BUFCOPY
;
219 #endif /* HAVE_ICONV */
220 /* Else, if we will modify the data bytes and thus introduce the potential
221 * of messing up multibyte sequences which become splitted over buffer
222 * boundaries TODO and unless we don't have our filter chain which will
223 * TODO make these hacks go by, buffer data until we see a NL */
224 if((flags
& (TD_ISPR
| TD_DELCTRL
)) && outrest
!= NULL
&&
226 iconvd
== (iconv_t
)-1 &&
228 (!(flags
& _TD_EOF
) || outrest
->l
> 0)
233 for (cp
= &in
.s
[in
.l
]; cp
> in
.s
&& cp
[-1] != '\n'; --cp
)
235 i
= PTR2SIZE(cp
- in
.s
);
239 n_str_assign_buf(outrest
, cp
, in
.l
- i
);
241 memcpy(cp
, in
.s
, in
.l
= i
);
242 (in
.s
= cp
)[in
.l
= i
] = '\0';
243 flags
&= ~_TD_BUFCOPY
;
245 n_str_add_buf(outrest
, input
->s
, input
->l
);
253 makeprint(&in
, &out
);
254 else if (flags
& _TD_BUFCOPY
)
255 n_str_dup(&out
, &in
);
258 if (flags
& TD_DELCTRL
)
259 out
.l
= delctrl(out
.s
, out
.l
);
262 __mimefwtd_opipe
= safe_signal(SIGPIPE
, &__mimefwtd_onsig
);
263 if (sigsetjmp(__mimefwtd_actjmp
, 1)) {
268 rv
= quoteflt_push(qf
, out
.s
, out
.l
);
273 if (in
.s
!= input
->s
)
275 safe_signal(SIGPIPE
, __mimefwtd_opipe
);
276 if (__mimefwtd_sig
!= 0)
277 n_raise(__mimefwtd_sig
);
284 mime_write_tohdr(struct str
*in
, FILE *fo
, size_t *colp
,
285 enum a_mime_structure_hack msh
)
287 /* TODO mime_write_tohdr(): we don't know the name of our header->maxcol..
288 * TODO MIME/send layer rewrite: more available state!!
289 * TODO Because of this we cannot make a difference in between structured
290 * TODO and unstructured headers (RFC 2047, 5. (2))
291 * TODO This means, e.g., that this gets called multiple times for a
292 * TODO structured header and always starts thinking it is at column 0.
293 * TODO I.e., it may get called for only the content of a comment etc.,
294 * TODO not knowing anything of its context.
295 * TODO Instead we should have a list of header body content tokens,
296 * TODO convert them, and then dump the converted tokens, breaking lines.
297 * TODO I.e., get rid of convhdra, mime_write_tohdr_a and such...
298 * TODO Somewhen, the following should produce smooth stuff:
299 * TODO ' "Hallo\"," Dr. Backe "Bl\"ö\"d" (Gell) <ha@llöch.en>
300 * TODO "Nochm\"a\"l"<ta@tu.da>(Dümm)'
301 * TODO NOT MULTIBYTE SAFE IF AN ENCODED WORD HAS TO BE SPLITTED!
302 * TODO To be better we had to mbtowc_l() (non-std! and no locale!!) and
303 * TODO work char-wise! -> S-CText..
304 * TODO The real problem for STD compatibility is however that "in" is
305 * TODO already iconv(3) encoded to the target character set! We could
306 * TODO also solve it (very expensively!) if we would narrow down to an
307 * TODO encoded word and then iconv(3)+MIME encode in one go, in which
308 * TODO case multibyte errors could be catched! */
310 /* Maximum line length */
311 a_MAXCOL_NENC
= MIME_LINELEN
,
312 a_MAXCOL
= MIME_LINELEN_RFC2047
315 struct str cout
, cin
;
317 _FIRST
= 1<<0, /* Nothing written yet, start of string */
318 _MSH_NOTHING
= 1<<1, /* Now, really: nothing at all has been written */
319 a_ANYENC
= 1<<2, /* We have RFC 2047 anything at least once */
320 _NO_QP
= 1<<3, /* No quoted-printable allowed */
321 _NO_B64
= 1<<4, /* Ditto, base64 */
322 _ENC_LAST
= 1<<5, /* Last round generated encoded word */
323 _SHOULD_BEE
= 1<<6, /* Avoid lines longer than SHOULD via encoding */
325 _RND_MASK
= (1<<_RND_SHIFT
) - 1,
326 _SPACE
= 1<<(_RND_SHIFT
+1), /* Leading whitespace */
327 _8BIT
= 1<<(_RND_SHIFT
+2), /* High bit set */
328 _ENCODE
= 1<<(_RND_SHIFT
+3), /* Need encoding */
329 _ENC_B64
= 1<<(_RND_SHIFT
+4), /* - let it be base64 */
330 _OVERLONG
= 1<<(_RND_SHIFT
+5) /* Temporarily rised limit */
332 char const *cset7
, *cset8
, *wbot
, *upper
, *wend
, *wcur
;
333 ui32_t cset7_len
, cset8_len
;
339 cout
.s
= NULL
, cout
.l
= 0;
340 cset7
= ok_vlook(charset_7bit
);
341 cset7_len
= (ui32_t
)strlen(cset7
);
342 cset8
= _CS_ITER_GET(); /* TODO MIME/send layer: iter active? iter! else */
343 cset8_len
= (ui32_t
)strlen(cset8
);
346 if(msh
!= a_MIME_SH_NONE
)
347 flags
|= _MSH_NOTHING
;
349 /* RFC 1468, "MIME Considerations":
350 * ISO-2022-JP may also be used in MIME Part 2 headers. The "B"
351 * encoding should be used with ISO-2022-JP text. */
352 /* TODO of course, our current implementation won't deal properly with
353 * TODO any stateful encoding at all... (the standard says each encoded
354 * TODO word must include all necessary reset sequences..., i.e., each
355 * TODO encoded word must be a self-contained iconv(3) life cycle) */
356 if (!asccasecmp(cset8
, "iso-2022-jp") || mime_enc_target() == MIMEE_B64
)
360 upper
= wbot
+ in
->l
;
363 if(colp
== NULL
|| (col
= *colp
) == 0)
364 col
= sizeof("Mail-Followup-To: ") -1; /* TODO dreadful thing */
366 /* The user may specify empy quoted-strings or comments, keep them! */
368 if(flags
& _MSH_NOTHING
){
369 flags
&= ~_MSH_NOTHING
;
370 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
374 } else for (; wbot
< upper
; flags
&= ~_FIRST
, wbot
= wend
) {
377 while (wcur
< upper
&& whitechar(*wcur
)) {
382 /* Any occurrence of whitespace resets prevention of lines >SHOULD via
383 * enforced encoding (xxx SHOULD, but.. encoding is expensive!!) */
385 flags
&= ~_SHOULD_BEE
;
387 /* Data ends with WS - dump it and done.
388 * Also, if we have seen multiple successive whitespace characters, then
389 * if there was no encoded word last, i.e., if we can simply take them
390 * over to the output as-is, keep one WS for possible later separation
391 * purposes and simply print the others as-is, directly! */
396 if ((flags
& (_ENC_LAST
| _SPACE
)) == _SPACE
&& wcur
- wbot
> 1) {
401 /* Skip over a word to next non-whitespace, keep track along the way
402 * whether our 7-bit charset suffices to represent the data */
403 for (wend
= wcur
; wend
< upper
; ++wend
) {
404 if (whitechar(*wend
))
406 if ((uc_i
)*wend
& 0x80)
410 /* Decide whether the range has to become encoded or not */
411 i
= PTR2SIZE(wend
- wcur
);
412 j
= mime_enc_mustquote(wcur
, i
, MIMEEF_ISHEAD
);
413 /* If it just cannot fit on a SHOULD line length, force encode */
414 if (i
> a_MAXCOL_NENC
) {
415 flags
|= _SHOULD_BEE
; /* (Sigh: SHOULD only, not MUST..) */
418 if ((flags
& _SHOULD_BEE
) || j
> 0) {
421 /* Use base64 if requested or more than 50% -37.5-% of the bytes of
422 * the string need to be encoded */
423 if ((flags
& _NO_QP
) || j
>= i
>> 1)/*(i >> 2) + (i >> 3))*/
426 DBG( if (flags
& _8BIT
) assert(flags
& _ENCODE
); )
428 if (!(flags
& _ENCODE
)) {
429 /* Encoded word produced, but no linear whitespace for necessary RFC
430 * 2047 separation? Generate artificial data (bad standard!) */
431 if ((flags
& (_ENC_LAST
| _SPACE
)) == _ENC_LAST
) {
432 if (col
>= a_MAXCOL
) {
437 if(flags
& _MSH_NOTHING
){
438 flags
&= ~_MSH_NOTHING
;
439 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
451 /* todo No effort here: (1) v15.0 has to bring complete rewrite,
452 * todo (2) the standard is braindead and (3) usually this is one
453 * todo word only, and why be smarter than the standard? */
455 i
= PTR2SIZE(wend
- wbot
);
456 if (i
+ col
+ ((flags
& _MSH_NOTHING
) != 0) <=
457 (flags
& _OVERLONG
? MIME_LINELEN_MAX
458 : (flags
& a_ANYENC
? a_MAXCOL
: a_MAXCOL_NENC
))) {
459 if(flags
& _MSH_NOTHING
){
460 flags
&= ~_MSH_NOTHING
;
461 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
465 i
= fwrite(wbot
, sizeof *wbot
, i
, fo
);
471 /* Doesn't fit, try to break the line first; */
474 if (whitechar(*wbot
)) {
475 putc((uc_i
)*wbot
, fo
);
478 putc(' ', fo
); /* Bad standard: artificial data! */
481 if(flags
& _MSH_NOTHING
){
482 flags
&= ~_MSH_NOTHING
;
483 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
491 /* It is so long that it needs to be broken, effectively causing
492 * artificial spaces to be inserted (bad standard), yuck */
493 /* todo This is not multibyte safe, as above; and completely stupid
494 * todo P.S.: our _SHOULD_BEE prevents these cases in the meanwhile */
495 /* FIXME n_PSO_UNICODE and parse using UTF-8 sync possibility! */
496 wcur
= wbot
+ MIME_LINELEN_MAX
- 8;
501 /* Encoding to encoded word(s); deal with leading whitespace, place
502 * a separator first as necessary: encoded words must always be
503 * separated from text and other encoded words with linear WS.
504 * And if an encoded word was last, intermediate whitespace must
505 * also be encoded, otherwise it would get stripped away! */
506 wcur
= n_UNCONST(n_empty
);
507 if ((flags
& (_ENC_LAST
| _SPACE
)) != _SPACE
) {
508 /* Reinclude whitespace */
510 /* We don't need to place a separator at the very beginning */
511 if (!(flags
& _FIRST
))
512 wcur
= n_UNCONST(" ");
516 flags
|= a_ANYENC
| _ENC_LAST
;
517 n_pstate
|= n_PS_HEADER_NEEDED_MIME
;
520 * An 'encoded-word' may not be more than 75 characters long,
521 * including 'charset', 'encoding', 'encoded-text', and
522 * delimiters. If it is desirable to encode more text than will
523 * fit in an 'encoded-word' of 75 characters, multiple
524 * 'encoded-word's (separated by CRLF SPACE) may be used.
526 * While there is no limit to the length of a multiple-line
527 * header field, each line of a header field that contains one
528 * or more 'encoded-word's is limited to 76 characters */
530 cin
.s
= n_UNCONST(wbot
);
531 cin
.l
= PTR2SIZE(wend
- wbot
);
537 xout
= b64_encode(&cout
, &cin
, B64_ISHEAD
| B64_ISENCWORD
);
539 xout
= qp_encode(&cout
, &cin
, QP_ISHEAD
| QP_ISENCWORD
);
546 /* (Avoid trigraphs in the RFC 2047 placeholder..) */
547 i
= j
+ (flags
& _8BIT
? cset8_len
: cset7_len
) + sizeof("=!!B!!=") -1;
552 /* Unfortunately RFC 2047 explicitly disallows encoded words to be
553 * longer (just like RFC 5322's "a line SHOULD fit in 78 but MAY be
554 * 998 characters long"), so we cannot use the _OVERLONG mechanism,
555 * even though all tested mailers seem to support it */
556 if (i
+ col
<= (/*flags & _OVERLONG ? MIME_LINELEN_MAX :*/ a_MAXCOL
)) {
557 if(flags
& _MSH_NOTHING
){
558 flags
&= ~_MSH_NOTHING
;
559 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
563 fprintf(fo
, "%.1s=?%s?%c?%.*s?=",
564 wcur
, (flags
& _8BIT
? cset8
: cset7
),
565 (flags
& _ENC_B64
? 'B' : 'Q'),
566 (int)cout
.l
, cout
.s
);
572 /* Doesn't fit, try to break the line first */
573 /* TODO I've commented out the _FIRST test since we (1) cannot do
574 * TODO _OVERLONG since (MUAs support but) the standard disallows,
575 * TODO and because of our iconv problem i prefer an empty first line
576 * TODO in favour of a possibly messed up multibytes character. :-( */
577 if (col
> 1 /* TODO && !(flags & _FIRST)*/) {
581 if (!(flags
& _SPACE
)) {
583 wcur
= n_UNCONST(n_empty
);
584 /*flags |= _OVERLONG;*/
585 goto jenc_retry_same
;
587 putc((uc_i
)*wcur
, fo
);
588 if (whitechar(*(wcur
= wbot
)))
592 wcur
= n_UNCONST(n_empty
);
594 /*flags &= ~_OVERLONG;*/
599 /* It is so long that it needs to be broken, effectively causing
600 * artificial data to be inserted (bad standard), yuck */
601 /* todo This is not multibyte safe, as above */
602 /*if (!(flags & _OVERLONG)) { Mechanism explicitly forbidden by 2047
607 /* FIXME n_PSO_UNICODE and parse using UTF-8 sync possibility! */
608 i
= PTR2SIZE(wend
- wbot
) + !!(flags
& _SPACE
);
609 j
= 3 + !(flags
& _ENC_B64
);
613 /* (Note the problem most likely is the transfer-encoding blow,
614 * which is why we test this *after* the decrements.. */
622 if(!(flags
& _MSH_NOTHING
) && msh
!= a_MIME_SH_NONE
){
623 putc((msh
== a_MIME_SH_COMMENT
? ')' : '"'), fo
);
638 mime_write_tohdr_a(struct str
*in
, FILE *f
, size_t *colp
)
642 char const *cp
, *lastcp
;
648 if((cp
= routeaddr(in
->s
)) != NULL
&& cp
> lastcp
) {
650 xin
.l
= PTR2SIZE(cp
- in
->s
);
651 if ((sz
= mime_write_tohdr_a(&xin
, f
, colp
)) < 0)
660 for( ; *cp
!= '\0'; ++cp
){
663 i
= PTR2SIZE(cp
- lastcp
);
665 if(fwrite(lastcp
, 1, i
, f
) != i
)
670 cp
= skip_comment(cp
);
673 /* We want to keep empty comments, too! */
674 xin
.s
= n_UNCONST(lastcp
);
675 xin
.l
= PTR2SIZE(cp
- lastcp
);
676 if ((x
= a_mime__convhdra(&xin
, f
, colp
, a_MIME_SH_COMMENT
)) < 0)
682 i
= PTR2SIZE(cp
- lastcp
);
684 if(fwrite(lastcp
, 1, i
, f
) != i
)
688 for(lastcp
= ++cp
; *cp
!= '\0'; ++cp
){
691 if(*cp
== '\\' && cp
[1] != '\0')
694 /* We want to keep empty quoted-strings, too! */
695 xin
.s
= n_UNCONST(lastcp
);
696 xin
.l
= PTR2SIZE(cp
- lastcp
);
697 if((x
= a_mime__convhdra(&xin
, f
, colp
, a_MIME_SH_QUOTE
)) < 0)
706 i
= PTR2SIZE(cp
- lastcp
);
708 if(fwrite(lastcp
, 1, i
, f
) != i
)
722 a_mime__convhdra(struct str
*inp
, FILE *fp
, size_t *colp
,
723 enum a_mime_structure_hack msh
){
731 if(inp
->l
> 0 && iconvd
!= (iconv_t
)-1){
733 if(n_iconv_str(iconvd
, n_ICONV_NONE
, &ciconv
, inp
, NULL
) != 0){
734 n_iconv_reset(iconvd
);
740 rv
= mime_write_tohdr(inp
, fp
, colp
, msh
);
747 #endif /* HAVE_ICONV */
750 _append_str(char **buf
, size_t *sz
, size_t *pos
, char const *str
, size_t len
)
753 *buf
= srealloc(*buf
, *sz
+= len
);
754 memcpy(&(*buf
)[*pos
], str
, len
);
760 _append_conv(char **buf
, size_t *sz
, size_t *pos
, char const *str
, size_t len
)
765 in
.s
= n_UNCONST(str
);
767 mime_fromhdr(&in
, &out
, TD_ISPR
| TD_ICONV
);
768 _append_str(buf
, sz
, pos
, out
.s
, out
.l
);
774 charset_iter_reset(char const *a_charset_to_try_first
) /* TODO elim. dups! */
777 size_t sarrl
[3], len
;
780 n_UNUSED(a_charset_to_try_first
);
783 sarr
[2] = ok_vlook(CHARSET_8BIT_OKEY
);
785 if(a_charset_to_try_first
!= NULL
&& strcmp(a_charset_to_try_first
, sarr
[2]))
786 sarr
[0] = a_charset_to_try_first
;
790 if((sarr
[1] = ok_vlook(sendcharsets
)) == NULL
&&
791 ok_blook(sendcharsets_else_ttycharset
)){
792 cp
= n_UNCONST(ok_vlook(ttycharset
));
793 if(strcmp(cp
, sarr
[2]) && (sarr
[0] == NULL
|| strcmp(cp
, sarr
[0])))
797 sarr
[2] = ok_vlook(ttycharset
);
800 sarrl
[2] = len
= strlen(sarr
[2]);
802 if ((cp
= n_UNCONST(sarr
[1])) != NULL
)
803 len
+= (sarrl
[1] = strlen(cp
));
806 if ((cp
= n_UNCONST(sarr
[0])) != NULL
)
807 len
+= (sarrl
[0] = strlen(cp
));
812 _cs_iter_base
= cp
= salloc(len
+ 1 + 1 +1);
815 if ((len
= sarrl
[0]) != 0) {
816 memcpy(cp
, sarr
[0], len
);
820 if ((len
= sarrl
[1]) != 0) {
821 memcpy(cp
, sarr
[1], len
);
827 memcpy(cp
, sarr
[2], len
);
832 return (_cs_iter
!= NULL
);
836 charset_iter_next(void)
842 rv
= (_cs_iter
!= NULL
);
848 charset_iter_is_valid(void)
853 rv
= (_cs_iter
!= NULL
);
870 charset_iter_or_fallback(void)
881 charset_iter_recurse(char *outer_storage
[2]) /* TODO LEGACY FUN, REMOVE */
884 outer_storage
[0] = _cs_iter_base
;
885 outer_storage
[1] = _cs_iter
;
890 charset_iter_restore(char *outer_storage
[2]) /* TODO LEGACY FUN, REMOVE */
893 _cs_iter_base
= outer_storage
[0];
894 _cs_iter
= outer_storage
[1];
900 need_hdrconv(struct header
*hp
) /* TODO once only, then iter */
902 struct n_header_field
*hfp
;
908 if((hfp
= hp
->h_user_headers
) != NULL
)
909 do if(_has_highbit(hfp
->hf_dat
+ hfp
->hf_nl
+1))
911 while((hfp
= hfp
->hf_next
) != NULL
);
913 if((hfp
= hp
->h_custom_headers
) != NULL
||
914 (hp
->h_custom_headers
= hfp
= n_customhdr_query()) != NULL
)
915 do if(_has_highbit(hfp
->hf_dat
+ hfp
->hf_nl
+1))
917 while((hfp
= hfp
->hf_next
) != NULL
);
919 if (hp
->h_mft
!= NULL
) {
920 if (_name_highbit(hp
->h_mft
))
923 if (hp
->h_from
!= NULL
) {
924 if (_name_highbit(hp
->h_from
))
926 } else if (_has_highbit(myaddrs(NULL
)))
928 if (hp
->h_reply_to
) {
929 if (_name_highbit(hp
->h_reply_to
))
932 char const *v15compat
;
934 if((v15compat
= ok_vlook(replyto
)) != NULL
)
935 n_OBSOLETE(_("please use *reply-to*, not *replyto*"));
936 if(_has_highbit(v15compat
))
938 if(_has_highbit(ok_vlook(reply_to
)))
942 if (_name_highbit(hp
->h_sender
))
944 } else if (_has_highbit(ok_vlook(sender
)))
947 if (_name_highbit(hp
->h_to
))
949 if (_name_highbit(hp
->h_cc
))
951 if (_name_highbit(hp
->h_bcc
))
953 if (_has_highbit(hp
->h_subject
))
955 rv
= _CS_ITER_GET(); /* TODO MIME/send: iter active? iter! else */
959 #endif /* HAVE_ICONV */
962 mime_fromhdr(struct str
const *in
, struct str
*out
, enum tdflags flags
)
964 /* TODO mime_fromhdr(): is called with strings that contain newlines;
965 * TODO this is the usual newline problem all around the codebase;
966 * TODO i.e., if we strip it, then the display misses it ;>
967 * TODO this is why it is so messy and why S-nail v14.2 plus additional
968 * TODO patch for v14.5.2 (and maybe even v14.5.3 subminor) occurred, and
969 * TODO why our display reflects what is contained in the message: the 1:1
970 * TODO relationship of message content and display!
971 * TODO instead a header line should be decoded to what it is (a single
972 * TODO line that is) and it should be objective to the backend whether
973 * TODO it'll be folded to fit onto the display or not, e.g., for search
974 * TODO purposes etc. then the only condition we have to honour in here
975 * TODO is that whitespace in between multiple adjacent MIME encoded words
976 * TODO á la RFC 2047 is discarded; i.e.: this function should deal with
977 * TODO RFC 2047 and be renamed: mime_fromhdr() -> mime_rfc2047_decode() */
978 struct str cin
, cout
;
979 char *p
, *op
, *upper
;
980 ui32_t convert
, lastenc
, lastoutl
;
984 iconv_t fhicd
= (iconv_t
)-1;
990 *(out
->s
= smalloc(1)) = '\0';
996 tcs
= ok_vlook(ttycharset
);
1000 lastenc
= lastoutl
= 0;
1004 if (*p
== '=' && *(p
+ 1) == '?') {
1009 while (p
< upper
&& *p
!= '?')
1010 ++p
; /* strip charset */
1015 if (flags
& TD_ICONV
) {
1016 size_t i
= PTR2SIZE(p
- cbeg
);
1017 char *ltag
, *cs
= ac_alloc(i
);
1019 memcpy(cs
, cbeg
, --i
);
1021 /* RFC 2231 extends the RFC 2047 character set definition in
1022 * encoded words by language tags - silently strip those off */
1023 if ((ltag
= strchr(cs
, '*')) != NULL
)
1026 if (fhicd
!= (iconv_t
)-1)
1027 n_iconv_close(fhicd
);
1028 fhicd
= asccasecmp(cs
, tcs
) ? n_iconv_open(tcs
, cs
) : (iconv_t
)-1;
1034 convert
= CONV_FROMB64
;
1037 convert
= CONV_FROMQP
;
1039 default: /* invalid, ignore */
1047 if (PTRCMP(p
+ 1, >=, upper
))
1049 if (*p
++ == '?' && *p
== '=')
1058 if (convert
== CONV_FROMB64
) {
1059 if(!b64_decode_header(&cout
, &cin
))
1060 n_str_assign_cp(&cout
, _("[Invalid Base64 encoding]"));
1061 }else if(!qp_decode_header(&cout
, &cin
))
1062 n_str_assign_cp(&cout
, _("[Invalid Quoted-Printable encoding]"));
1066 if ((flags
& TD_ICONV
) && fhicd
!= (iconv_t
)-1) {
1067 cin
.s
= NULL
, cin
.l
= 0; /* XXX string pool ! */
1068 convert
= n_iconv_str(fhicd
, n_ICONV_UNIDEFAULT
, &cin
, &cout
, NULL
);
1069 out
= n_str_add(out
, &cin
);
1070 if (convert
) {/* n_ERR_INVAL at EOS */
1071 n_iconv_reset(fhicd
);
1072 out
= n_str_add_buf(out
, n_qm
, 1); /* TODO unicode replacement */
1077 out
= n_str_add(out
, &cout
);
1078 lastenc
= lastoutl
= out
->l
;
1085 onlyws
= (lastenc
> 0);
1089 if (op
[0] == '=' && (PTRCMP(op
+ 1, ==, upper
) || op
[1] == '?'))
1091 if (onlyws
&& !blankchar(*op
))
1095 out
= n_str_add_buf(out
, p
, PTR2SIZE(op
- p
));
1097 if (!onlyws
|| lastoutl
!= lastenc
)
1102 out
->s
[out
->l
] = '\0';
1104 if (flags
& TD_ISPR
) {
1105 makeprint(out
, &cout
);
1109 if (flags
& TD_DELCTRL
)
1110 out
->l
= delctrl(out
->s
, out
->l
);
1112 if (fhicd
!= (iconv_t
)-1)
1113 n_iconv_close(fhicd
);
1121 mime_fromaddr(char const *name
)
1123 char const *cp
, *lastcp
;
1125 size_t ressz
= 1, rescur
= 0;
1130 if (*name
== '\0') {
1131 res
= savestr(name
);
1135 if ((cp
= routeaddr(name
)) != NULL
&& cp
> name
) {
1136 _append_conv(&res
, &ressz
, &rescur
, name
, PTR2SIZE(cp
- name
));
1141 for ( ; *cp
; ++cp
) {
1144 _append_str(&res
, &ressz
, &rescur
, lastcp
, PTR2SIZE(cp
- lastcp
+ 1));
1146 cp
= skip_comment(cp
);
1148 _append_conv(&res
, &ressz
, &rescur
, lastcp
, PTR2SIZE(cp
- lastcp
));
1155 if (*cp
== '\\' && cp
[1] != '\0')
1162 _append_str(&res
, &ressz
, &rescur
, lastcp
, PTR2SIZE(cp
- lastcp
));
1167 res
= savestrbuf(res
, rescur
);
1177 xmime_write(char const *ptr
, size_t size
, FILE *f
, enum conversion convert
,
1178 enum tdflags dflags
, struct str
* volatile outrest
,
1179 struct str
* volatile inrest
)
1182 struct quoteflt
*qf
;
1185 quoteflt_reset(qf
= quoteflt_dummy(), f
);
1186 rv
= mime_write(ptr
, size
, f
, convert
, dflags
, qf
, outrest
, inrest
);
1192 static sigjmp_buf __mimemw_actjmp
; /* TODO someday.. */
1193 static int __mimemw_sig
; /* TODO someday.. */
1194 static sighandler_type __mimemw_opipe
;
1196 __mimemw_onsig(int sig
) /* TODO someday, we won't need it no more */
1198 NYD_X
; /* Signal handler */
1200 siglongjmp(__mimemw_actjmp
, 1);
1204 mime_write(char const *ptr
, size_t size
, FILE *f
,
1205 enum conversion convert
, enum tdflags
volatile dflags
,
1206 struct quoteflt
*qf
, struct str
* volatile outrest
,
1207 struct str
* volatile inrest
)
1209 /* TODO note: after send/MIME layer rewrite we will have a string pool
1210 * TODO so that memory allocation count drops down massively; for now,
1211 * TODO v14.0 that is, we pay a lot & heavily depend on the allocator.
1212 * TODO P.S.: furthermore all this encapsulated in filter objects instead */
1214 ssize_t
volatile sz
;
1217 dflags
|= _TD_BUFCOPY
;
1218 in
.s
= n_UNCONST(ptr
);
1223 if((sz
= size
) == 0){
1224 if(inrest
!= NULL
&& inrest
->l
!= 0)
1226 if(outrest
!= NULL
&& outrest
->l
!= 0)
1231 /* TODO This crap requires linewise input, then. We need a filter chain
1232 * TODO as in input->iconv->base64 where each filter can have its own
1233 * TODO buffer, with a filter->fflush() call to get rid of those! */
1235 if ((dflags
& TD_ICONV
) && iconvd
!= (iconv_t
)-1 &&
1236 (convert
== CONV_TOQP
|| convert
== CONV_8BIT
||
1237 convert
== CONV_TOB64
|| convert
== CONV_TOHDR
)) {
1238 if (n_iconv_str(iconvd
, n_ICONV_NONE
, &out
, &in
, NULL
) != 0) {
1239 n_iconv_reset(iconvd
);
1240 /* TODO This causes hard-failure. We would need to have an action
1241 * TODO policy FAIL|IGNORE|SETERROR(but continue) */
1247 dflags
&= ~_TD_BUFCOPY
;
1252 if(inrest
!= NULL
&& inrest
->l
> 0){
1258 out
.s
= n_alloc(in
.l
+ inrest
->l
+ 1);
1259 memcpy(out
.s
, inrest
->s
, inrest
->l
);
1261 memcpy(&out
.s
[inrest
->l
], in
.s
, in
.l
);
1264 (in
.s
= out
.s
)[in
.l
+= inrest
->l
] = '\0';
1268 dflags
&= ~_TD_BUFCOPY
;
1273 __mimemw_opipe
= safe_signal(SIGPIPE
, &__mimemw_onsig
);
1274 if (sigsetjmp(__mimemw_actjmp
, 1))
1279 if(!qp_decode_part(&out
, &in
, outrest
, inrest
)){
1280 n_err(_("Invalid Quoted-Printable encoding ignored\n"));
1281 sz
= 0; /* TODO sz = -1 stops outer levels! */
1286 if(qp_encode(&out
, &in
, QP_NONE
) == NULL
){
1287 sz
= 0; /* TODO sz = -1 stops outer levels! */
1292 sz
= quoteflt_push(qf
, in
.s
, in
.l
);
1295 if(!b64_decode_part(&out
, &in
, outrest
, inrest
))
1300 case CONV_FROMB64_T
:
1301 if(!b64_decode_part(&out
, &in
, outrest
, inrest
)){
1303 n_err(_("Invalid Base64 encoding ignored\n"));
1304 sz
= 0; /* TODO sz = -1 stops outer levels! */
1309 if ((sz
= out
.l
) != 0) {
1310 ui32_t opl
= qf
->qf_pfix_len
;
1311 sz
= _fwrite_td(&out
, FAL0
, (dflags
& ~_TD_BUFCOPY
), outrest
, qf
);
1312 qf
->qf_pfix_len
= opl
;
1316 /* TODO hack which is necessary unless this is a filter based approach
1317 * TODO and each filter has its own buffer (as necessary): we must not
1318 * TODO pass through a number of bytes which causes padding, otherwise we
1319 * TODO produce multiple adjacent base64 streams, and that is not treated
1320 * TODO in the same relaxed fashion like completely bogus bytes by at
1321 * TODO least mutt and OpenSSL. So we need an expensive workaround
1322 * TODO unless we have input->iconv->base64 filter chain as such!! :( */
1323 if(size
!= 0 && /* for Coverity, else assert() */ inrest
!= NULL
){
1324 if(in
.l
> B64_ENCODE_INPUT_PER_LINE
){
1327 i
= in
.l
% B64_ENCODE_INPUT_PER_LINE
;
1331 assert(inrest
->l
== 0);
1332 inrest
->s
= n_realloc(inrest
->s
, i
+1);
1333 memcpy(inrest
->s
, &in
.s
[in
.l
], i
);
1334 inrest
->s
[inrest
->l
= i
] = '\0';
1336 }else if(in
.l
< B64_ENCODE_INPUT_PER_LINE
){
1337 inrest
->s
= n_realloc(inrest
->s
, in
.l
+1);
1338 memcpy(inrest
->s
, in
.s
, in
.l
);
1339 inrest
->s
[inrest
->l
= in
.l
] = '\0';
1345 if(b64_encode(&out
, &in
, B64_LF
| B64_MULTILINE
) == NULL
){
1350 sz
= fwrite(out
.s
, sizeof *out
.s
, out
.l
, f
);
1351 if (sz
!= (ssize_t
)out
.l
)
1355 mime_fromhdr(&in
, &out
, TD_ISPR
| TD_ICONV
| (dflags
& TD_DELCTRL
));
1356 sz
= quoteflt_push(qf
, out
.s
, out
.l
);
1359 sz
= mime_write_tohdr(&in
, f
, NULL
, a_MIME_SH_NONE
);
1364 if(dflags
& _TD_BUFCOPY
){
1365 n_str_dup(&out
, &in
);
1368 dflags
&= ~_TD_BUFCOPY
;
1371 sz
= mime_write_tohdr_a(&in
, f
, &col
);
1374 sz
= _fwrite_td(&in
, TRU1
, dflags
, NULL
, qf
);
1383 safe_signal(SIGPIPE
, __mimemw_opipe
);
1384 if (__mimemw_sig
!= 0)
1385 n_raise(__mimemw_sig
);