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 - 2018 Steffen (Daode) Nurpmeso <steffen@sdaoden.eu>.
7 * SPDX-License-Identifier: BSD-4-Clause
11 * Gunnar Ritter. All rights reserved.
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. All advertising materials mentioning features or use of this software
22 * must display the following acknowledgement:
23 * This product includes software developed by Gunnar Ritter
24 * and his contributors.
25 * 4. Neither the name of Gunnar Ritter nor the names of his contributors
26 * may be used to endorse or promote products derived from this software
27 * without specific prior written permission.
29 * THIS SOFTWARE IS PROVIDED BY GUNNAR RITTER AND CONTRIBUTORS ``AS IS'' AND
30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32 * ARE DISCLAIMED. IN NO EVENT SHALL GUNNAR RITTER OR CONTRIBUTORS BE LIABLE
33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
44 #ifndef HAVE_AMALGAMATION
48 /* Don't ask, but it keeps body and soul together */
49 enum a_mime_structure_hack
{
55 static char *_cs_iter_base
, *_cs_iter
;
57 # define _CS_ITER_GET() \
58 ((_cs_iter != NULL) ? _cs_iter : ok_vlook(CHARSET_8BIT_OKEY))
60 # define _CS_ITER_GET() ((_cs_iter != NULL) ? _cs_iter : ok_vlook(ttycharset))
62 #define _CS_ITER_STEP() _cs_iter = n_strsep(&_cs_iter_base, ',', TRU1)
64 /* Is 7-bit enough? */
66 static bool_t
_has_highbit(char const *s
);
67 static bool_t
_name_highbit(struct name
*np
);
70 /* fwrite(3) while checking for displayability */
71 static ssize_t
_fwrite_td(struct str
const *input
,
72 bool_t failiconv
, enum tdflags flags
,
73 struct str
*outrest
, struct quoteflt
*qf
);
75 /* Convert header fields to RFC 2047 format and write to the file fo */
76 static ssize_t
mime_write_tohdr(struct str
*in
, FILE *fo
,
77 size_t *colp
, enum a_mime_structure_hack msh
);
80 static ssize_t
a_mime__convhdra(struct str
*inp
, FILE *fp
, size_t *colp
,
81 enum a_mime_structure_hack msh
);
83 # define a_mime__convhdra(S,F,C,MSH) mime_write_tohdr(S, F, C, MSH)
86 /* Write an address to a header field */
87 static ssize_t
mime_write_tohdr_a(struct str
*in
, FILE *f
,
88 size_t *colp
, enum a_mime_structure_hack msh
);
90 /* Append to buf, handling resizing */
91 static void _append_str(char **buf
, size_t *sz
, size_t *pos
,
92 char const *str
, size_t len
);
93 static void _append_conv(char **buf
, size_t *sz
, size_t *pos
,
94 char const *str
, size_t len
);
98 _has_highbit(char const *s
)
105 if ((ui8_t
)*s
& 0x80)
107 while (*s
++ != '\0');
116 _name_highbit(struct name
*np
)
122 if (_has_highbit(np
->n_name
) || _has_highbit(np
->n_fullname
))
131 #endif /* HAVE_ICONV */
133 static sigjmp_buf __mimefwtd_actjmp
; /* TODO someday.. */
134 static int __mimefwtd_sig
; /* TODO someday.. */
135 static sighandler_type __mimefwtd_opipe
;
137 __mimefwtd_onsig(int sig
) /* TODO someday, we won't need it no more */
139 NYD_X
; /* Signal handler */
140 __mimefwtd_sig
= sig
;
141 siglongjmp(__mimefwtd_actjmp
, 1);
145 _fwrite_td(struct str
const *input
, bool_t failiconv
, enum tdflags flags
,
146 struct str
*outrest
, struct quoteflt
*qf
)
148 /* TODO note: after send/MIME layer rewrite we will have a string pool
149 * TODO so that memory allocation count drops down massively; for now,
150 * TODO v14.* that is, we pay a lot & heavily depend on the allocator */
151 /* TODO well if we get a broken pipe here, and it happens to
152 * TODO happen pretty easy when sleeping in a full pipe buffer,
153 * TODO then the current codebase performs longjump away;
154 * TODO this leaves memory leaks behind ('think up to 3 per,
155 * TODO dep. upon alloca availability). For this to be fixed
156 * TODO we either need to get rid of the longjmp()s (tm) or
157 * TODO the storage must come from the outside or be tracked
158 * TODO in a carrier struct. Best both. But storage reuse
159 * TODO would be a bigbig win besides */
160 /* *input* _may_ point to non-modifyable buffer; but even then it only
161 * needs to be dup'ed away if we have to transform the content */
173 if ((flags
& TD_ICONV
) && iconvd
!= (iconv_t
)-1) {
179 if (outrest
!= NULL
&& outrest
->l
> 0) {
180 in
.l
= outrest
->l
+ input
->l
;
181 in
.s
= buf
= n_alloc(in
.l
+1);
182 memcpy(in
.s
, outrest
->s
, outrest
->l
);
183 memcpy(&in
.s
[outrest
->l
], input
->s
, input
->l
);
189 /* TODO Sigh, no problem if we have a filter that has a buffer (or
190 * TODO become fed with entire lines, whatever), but for now we need
191 * TODO to ensure we pass entire lines from in here to iconv(3), because
192 * TODO the Citrus iconv(3) will fail tests with stateful encodings
193 * TODO if we do not (only seen on FreeBSD) */
194 #if 0 /* TODO actually not needed indeed, it was known iswprint() error! */
195 if(!(flags
& _TD_EOF
) && outrest
!= NULL
){
199 if((cp
= memchr(in
.s
, '\n', j
= in
.l
)) != NULL
){
200 i
= PTR2SIZE(cp
- in
.s
);
202 while(j
> 0 && *cp
== '\n') /* XXX one iteration too much */
205 n_str_assign_buf(outrest
, cp
, j
);
208 n_str_assign(outrest
, &in
);
214 if((err
= n_iconv_str(iconvd
,
215 (failiconv
? n_ICONV_NONE
: n_ICONV_UNIDEFAULT
),
216 &out
, &in
, &in
)) != 0){
217 if(err
!= n_ERR_INVAL
)
218 n_iconv_reset(iconvd
);
220 if(outrest
!= NULL
&& in
.l
> 0){
221 /* Incomplete multibyte at EOF is special xxx _INVAL? */
222 if (flags
& _TD_EOF
) {
223 out
.s
= n_realloc(out
.s
, out
.l
+ sizeof(n_unirepl
));
224 if(n_psonce
& n_PSO_UNICODE
){
225 memcpy(&out
.s
[out
.l
], n_unirepl
, sizeof(n_unirepl
) -1);
226 out
.l
+= sizeof(n_unirepl
) -1;
228 out
.s
[out
.l
++] = '?';
230 n_str_add(outrest
, &in
);
237 flags
&= ~_TD_BUFCOPY
;
244 #endif /* HAVE_ICONV */
245 /* Else, if we will modify the data bytes and thus introduce the potential
246 * of messing up multibyte sequences which become splitted over buffer
247 * boundaries TODO and unless we don't have our filter chain which will
248 * TODO make these hacks go by, buffer data until we see a NL */
249 if((flags
& (TD_ISPR
| TD_DELCTRL
)) && outrest
!= NULL
&&
251 iconvd
== (iconv_t
)-1 &&
253 (!(flags
& _TD_EOF
) || outrest
->l
> 0)
258 for (cp
= &in
.s
[in
.l
]; cp
> in
.s
&& cp
[-1] != '\n'; --cp
)
260 i
= PTR2SIZE(cp
- in
.s
);
264 n_str_assign_buf(outrest
, cp
, in
.l
- i
);
266 memcpy(cp
, in
.s
, in
.l
= i
);
267 (in
.s
= cp
)[in
.l
= i
] = '\0';
268 flags
&= ~_TD_BUFCOPY
;
270 n_str_add_buf(outrest
, input
->s
, input
->l
);
278 makeprint(&in
, &out
);
279 else if (flags
& _TD_BUFCOPY
)
280 n_str_dup(&out
, &in
);
283 if (flags
& TD_DELCTRL
)
284 out
.l
= delctrl(out
.s
, out
.l
);
287 __mimefwtd_opipe
= safe_signal(SIGPIPE
, &__mimefwtd_onsig
);
288 if (sigsetjmp(__mimefwtd_actjmp
, 1)) {
293 rv
= quoteflt_push(qf
, out
.s
, out
.l
);
298 if (in
.s
!= input
->s
)
300 safe_signal(SIGPIPE
, __mimefwtd_opipe
);
301 if (__mimefwtd_sig
!= 0)
302 n_raise(__mimefwtd_sig
);
309 mime_write_tohdr(struct str
*in
, FILE *fo
, size_t *colp
,
310 enum a_mime_structure_hack msh
)
312 /* TODO mime_write_tohdr(): we don't know the name of our header->maxcol..
313 * TODO MIME/send layer rewrite: more available state!!
314 * TODO Because of this we cannot make a difference in between structured
315 * TODO and unstructured headers (RFC 2047, 5. (2))
316 * TODO This means, e.g., that this gets called multiple times for a
317 * TODO structured header and always starts thinking it is at column 0.
318 * TODO I.e., it may get called for only the content of a comment etc.,
319 * TODO not knowing anything of its context.
320 * TODO Instead we should have a list of header body content tokens,
321 * TODO convert them, and then dump the converted tokens, breaking lines.
322 * TODO I.e., get rid of convhdra, mime_write_tohdr_a and such...
323 * TODO Somewhen, the following should produce smooth stuff:
324 * TODO ' "Hallo\"," Dr. Backe "Bl\"ö\"d" (Gell) <ha@llöch.en>
325 * TODO "Nochm\"a\"l"<ta@tu.da>(Dümm)'
326 * TODO NOT MULTIBYTE SAFE IF AN ENCODED WORD HAS TO BE SPLITTED!
327 * TODO To be better we had to mbtowc_l() (non-std! and no locale!!) and
328 * TODO work char-wise! -> S-CText..
329 * TODO The real problem for STD compatibility is however that "in" is
330 * TODO already iconv(3) encoded to the target character set! We could
331 * TODO also solve it (very expensively!) if we would narrow down to an
332 * TODO encoded word and then iconv(3)+MIME encode in one go, in which
333 * TODO case multibyte errors could be caught! */
335 /* Maximum line length */
336 a_MAXCOL_NENC
= MIME_LINELEN
,
337 a_MAXCOL
= MIME_LINELEN_RFC2047
340 struct str cout
, cin
;
342 _FIRST
= 1<<0, /* Nothing written yet, start of string */
343 _MSH_NOTHING
= 1<<1, /* Now, really: nothing at all has been written */
344 a_ANYENC
= 1<<2, /* We have RFC 2047 anything at least once */
345 _NO_QP
= 1<<3, /* No quoted-printable allowed */
346 _NO_B64
= 1<<4, /* Ditto, base64 */
347 _ENC_LAST
= 1<<5, /* Last round generated encoded word */
348 _SHOULD_BEE
= 1<<6, /* Avoid lines longer than SHOULD via encoding */
350 _RND_MASK
= (1<<_RND_SHIFT
) - 1,
351 _SPACE
= 1<<(_RND_SHIFT
+1), /* Leading whitespace */
352 _8BIT
= 1<<(_RND_SHIFT
+2), /* High bit set */
353 _ENCODE
= 1<<(_RND_SHIFT
+3), /* Need encoding */
354 _ENC_B64
= 1<<(_RND_SHIFT
+4), /* - let it be base64 */
355 _OVERLONG
= 1<<(_RND_SHIFT
+5) /* Temporarily rised limit */
357 char const *cset7
, *cset8
, *wbot
, *upper
, *wend
, *wcur
;
358 ui32_t cset7_len
, cset8_len
;
364 cout
.s
= NULL
, cout
.l
= 0;
365 cset7
= ok_vlook(charset_7bit
);
366 cset7_len
= (ui32_t
)strlen(cset7
);
367 cset8
= _CS_ITER_GET(); /* TODO MIME/send layer: iter active? iter! else */
368 cset8_len
= (ui32_t
)strlen(cset8
);
371 if(msh
!= a_MIME_SH_NONE
)
372 flags
|= _MSH_NOTHING
;
374 /* RFC 1468, "MIME Considerations":
375 * ISO-2022-JP may also be used in MIME Part 2 headers. The "B"
376 * encoding should be used with ISO-2022-JP text. */
377 /* TODO of course, our current implementation won't deal properly with
378 * TODO any stateful encoding at all... (the standard says each encoded
379 * TODO word must include all necessary reset sequences..., i.e., each
380 * TODO encoded word must be a self-contained iconv(3) life cycle) */
381 if (!asccasecmp(cset8
, "iso-2022-jp") || mime_enc_target() == MIMEE_B64
)
385 upper
= wbot
+ in
->l
;
388 if(colp
== NULL
|| (col
= *colp
) == 0)
389 col
= sizeof("Mail-Followup-To: ") -1; /* TODO dreadful thing */
391 /* The user may specify empy quoted-strings or comments, keep them! */
393 if(flags
& _MSH_NOTHING
){
394 flags
&= ~_MSH_NOTHING
;
395 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
399 } else for (; wbot
< upper
; flags
&= ~_FIRST
, wbot
= wend
) {
402 while (wcur
< upper
&& whitechar(*wcur
)) {
407 /* Any occurrence of whitespace resets prevention of lines >SHOULD via
408 * enforced encoding (xxx SHOULD, but.. encoding is expensive!!) */
410 flags
&= ~_SHOULD_BEE
;
412 /* Data ends with WS - dump it and done.
413 * Also, if we have seen multiple successive whitespace characters, then
414 * if there was no encoded word last, i.e., if we can simply take them
415 * over to the output as-is, keep one WS for possible later separation
416 * purposes and simply print the others as-is, directly! */
421 if ((flags
& (_ENC_LAST
| _SPACE
)) == _SPACE
&& wcur
- wbot
> 1) {
426 /* Skip over a word to next non-whitespace, keep track along the way
427 * whether our 7-bit charset suffices to represent the data */
428 for (wend
= wcur
; wend
< upper
; ++wend
) {
429 if (whitechar(*wend
))
431 if ((uc_i
)*wend
& 0x80)
435 /* Decide whether the range has to become encoded or not */
436 i
= PTR2SIZE(wend
- wcur
);
437 j
= mime_enc_mustquote(wcur
, i
, MIMEEF_ISHEAD
);
438 /* If it just cannot fit on a SHOULD line length, force encode */
439 if (i
> a_MAXCOL_NENC
) {
440 flags
|= _SHOULD_BEE
; /* (Sigh: SHOULD only, not MUST..) */
443 if ((flags
& _SHOULD_BEE
) || j
> 0) {
446 /* Use base64 if requested or more than 50% -37.5-% of the bytes of
447 * the string need to be encoded */
448 if ((flags
& _NO_QP
) || j
>= i
>> 1)/*(i >> 2) + (i >> 3))*/
451 DBG( if (flags
& _8BIT
) assert(flags
& _ENCODE
); )
453 if (!(flags
& _ENCODE
)) {
454 /* Encoded word produced, but no linear whitespace for necessary RFC
455 * 2047 separation? Generate artificial data (bad standard!) */
456 if ((flags
& (_ENC_LAST
| _SPACE
)) == _ENC_LAST
) {
457 if (col
>= a_MAXCOL
) {
462 if(flags
& _MSH_NOTHING
){
463 flags
&= ~_MSH_NOTHING
;
464 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
476 /* todo No effort here: (1) v15.0 has to bring complete rewrite,
477 * todo (2) the standard is braindead and (3) usually this is one
478 * todo word only, and why be smarter than the standard? */
480 i
= PTR2SIZE(wend
- wbot
);
481 if (i
+ col
+ ((flags
& _MSH_NOTHING
) != 0) <=
482 (flags
& _OVERLONG
? MIME_LINELEN_MAX
483 : (flags
& a_ANYENC
? a_MAXCOL
: a_MAXCOL_NENC
))) {
484 if(flags
& _MSH_NOTHING
){
485 flags
&= ~_MSH_NOTHING
;
486 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
490 i
= fwrite(wbot
, sizeof *wbot
, i
, fo
);
496 /* Doesn't fit, try to break the line first; */
499 if (whitechar(*wbot
)) {
500 putc((uc_i
)*wbot
, fo
);
503 putc(' ', fo
); /* Bad standard: artificial data! */
506 if(flags
& _MSH_NOTHING
){
507 flags
&= ~_MSH_NOTHING
;
508 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
516 /* It is so long that it needs to be broken, effectively causing
517 * artificial spaces to be inserted (bad standard), yuck */
518 /* todo This is not multibyte safe, as above; and completely stupid
519 * todo P.S.: our _SHOULD_BEE prevents these cases in the meanwhile */
520 /* FIXME n_PSO_UNICODE and parse using UTF-8 sync possibility! */
521 wcur
= wbot
+ MIME_LINELEN_MAX
- 8;
526 /* Encoding to encoded word(s); deal with leading whitespace, place
527 * a separator first as necessary: encoded words must always be
528 * separated from text and other encoded words with linear WS.
529 * And if an encoded word was last, intermediate whitespace must
530 * also be encoded, otherwise it would get stripped away! */
531 wcur
= n_UNCONST(n_empty
);
532 if ((flags
& (_ENC_LAST
| _SPACE
)) != _SPACE
) {
533 /* Reinclude whitespace */
535 /* We don't need to place a separator at the very beginning */
536 if (!(flags
& _FIRST
))
537 wcur
= n_UNCONST(" ");
541 flags
|= a_ANYENC
| _ENC_LAST
;
542 n_pstate
|= n_PS_HEADER_NEEDED_MIME
;
545 * An 'encoded-word' may not be more than 75 characters long,
546 * including 'charset', 'encoding', 'encoded-text', and
547 * delimiters. If it is desirable to encode more text than will
548 * fit in an 'encoded-word' of 75 characters, multiple
549 * 'encoded-word's (separated by CRLF SPACE) may be used.
551 * While there is no limit to the length of a multiple-line
552 * header field, each line of a header field that contains one
553 * or more 'encoded-word's is limited to 76 characters */
555 cin
.s
= n_UNCONST(wbot
);
556 cin
.l
= PTR2SIZE(wend
- wbot
);
562 xout
= b64_encode(&cout
, &cin
, B64_ISHEAD
| B64_ISENCWORD
);
564 xout
= qp_encode(&cout
, &cin
, QP_ISHEAD
| QP_ISENCWORD
);
571 /* (Avoid trigraphs in the RFC 2047 placeholder..) */
572 i
= j
+ (flags
& _8BIT
? cset8_len
: cset7_len
) + sizeof("=!!B!!=") -1;
577 /* Unfortunately RFC 2047 explicitly disallows encoded words to be
578 * longer (just like RFC 5322's "a line SHOULD fit in 78 but MAY be
579 * 998 characters long"), so we cannot use the _OVERLONG mechanism,
580 * even though all tested mailers seem to support it */
581 if (i
+ col
<= (/*flags & _OVERLONG ? MIME_LINELEN_MAX :*/ a_MAXCOL
)) {
582 if(flags
& _MSH_NOTHING
){
583 flags
&= ~_MSH_NOTHING
;
584 putc((msh
== a_MIME_SH_COMMENT
? '(' : '"'), fo
);
588 fprintf(fo
, "%.1s=?%s?%c?%.*s?=",
589 wcur
, (flags
& _8BIT
? cset8
: cset7
),
590 (flags
& _ENC_B64
? 'B' : 'Q'),
591 (int)cout
.l
, cout
.s
);
597 /* Doesn't fit, try to break the line first */
598 /* TODO I've commented out the _FIRST test since we (1) cannot do
599 * TODO _OVERLONG since (MUAs support but) the standard disallows,
600 * TODO and because of our iconv problem i prefer an empty first line
601 * TODO in favour of a possibly messed up multibytes character. :-( */
602 if (col
> 1 /* TODO && !(flags & _FIRST)*/) {
606 if (!(flags
& _SPACE
)) {
608 wcur
= n_UNCONST(n_empty
);
609 /*flags |= _OVERLONG;*/
610 goto jenc_retry_same
;
612 putc((uc_i
)*wcur
, fo
);
613 if (whitechar(*(wcur
= wbot
)))
617 wcur
= n_UNCONST(n_empty
);
619 /*flags &= ~_OVERLONG;*/
624 /* It is so long that it needs to be broken, effectively causing
625 * artificial data to be inserted (bad standard), yuck */
626 /* todo This is not multibyte safe, as above */
627 /*if (!(flags & _OVERLONG)) { Mechanism explicitly forbidden by 2047
632 /* FIXME n_PSO_UNICODE and parse using UTF-8 sync possibility! */
633 i
= PTR2SIZE(wend
- wbot
) + !!(flags
& _SPACE
);
634 j
= 3 + !(flags
& _ENC_B64
);
638 /* (Note the problem most likely is the transfer-encoding blow,
639 * which is why we test this *after* the decrements.. */
647 if(!(flags
& _MSH_NOTHING
) && msh
!= a_MIME_SH_NONE
){
648 putc((msh
== a_MIME_SH_COMMENT
? ')' : '"'), fo
);
664 a_mime__convhdra(struct str
*inp
, FILE *fp
, size_t *colp
,
665 enum a_mime_structure_hack msh
){
673 if(inp
->l
> 0 && iconvd
!= (iconv_t
)-1){
675 if(n_iconv_str(iconvd
, n_ICONV_NONE
, &ciconv
, inp
, NULL
) != 0){
676 n_iconv_reset(iconvd
);
682 rv
= mime_write_tohdr(inp
, fp
, colp
, msh
);
689 #endif /* HAVE_ICONV */
692 mime_write_tohdr_a(struct str
*in
, FILE *f
, size_t *colp
,
693 enum a_mime_structure_hack msh
)
697 char const *cp
, *lastcp
;
703 if((cp
= routeaddr(lastcp
= in
->s
)) != NULL
&& cp
> lastcp
) {
704 xin
.s
= n_UNCONST(lastcp
);
705 xin
.l
= PTR2SIZE(cp
- lastcp
);
706 if ((sz
= a_mime__convhdra(&xin
, f
, colp
, msh
)) < 0)
714 for( ; *cp
!= '\0'; ++cp
){
717 i
= PTR2SIZE(cp
- lastcp
);
719 if(fwrite(lastcp
, 1, i
, f
) != i
)
724 cp
= skip_comment(cp
);
727 /* We want to keep empty comments, too! */
728 xin
.s
= n_UNCONST(lastcp
);
729 xin
.l
= PTR2SIZE(cp
- lastcp
);
730 if ((x
= a_mime__convhdra(&xin
, f
, colp
, a_MIME_SH_COMMENT
)) < 0)
736 i
= PTR2SIZE(cp
- lastcp
);
738 if(fwrite(lastcp
, 1, i
, f
) != i
)
742 for(lastcp
= ++cp
; *cp
!= '\0'; ++cp
){
745 if(*cp
== '\\' && cp
[1] != '\0')
748 /* We want to keep empty quoted-strings, too! */
749 xin
.s
= n_UNCONST(lastcp
);
750 xin
.l
= PTR2SIZE(cp
- lastcp
);
751 if((x
= a_mime__convhdra(&xin
, f
, colp
, a_MIME_SH_QUOTE
)) < 0)
760 i
= PTR2SIZE(cp
- lastcp
);
762 if(fwrite(lastcp
, 1, i
, f
) != i
)
775 _append_str(char **buf
, size_t *sz
, size_t *pos
, char const *str
, size_t len
)
778 *buf
= n_realloc(*buf
, *sz
+= len
);
779 memcpy(&(*buf
)[*pos
], str
, len
);
785 _append_conv(char **buf
, size_t *sz
, size_t *pos
, char const *str
, size_t len
)
790 in
.s
= n_UNCONST(str
);
792 mime_fromhdr(&in
, &out
, TD_ISPR
| TD_ICONV
);
793 _append_str(buf
, sz
, pos
, out
.s
, out
.l
);
799 charset_iter_reset(char const *a_charset_to_try_first
) /* TODO elim. dups! */
802 size_t sarrl
[3], len
;
805 n_UNUSED(a_charset_to_try_first
);
808 sarr
[2] = ok_vlook(CHARSET_8BIT_OKEY
);
810 if(a_charset_to_try_first
!= NULL
&& strcmp(a_charset_to_try_first
, sarr
[2]))
811 sarr
[0] = a_charset_to_try_first
;
815 if((sarr
[1] = ok_vlook(sendcharsets
)) == NULL
&&
816 ok_blook(sendcharsets_else_ttycharset
)){
817 cp
= n_UNCONST(ok_vlook(ttycharset
));
818 if(strcmp(cp
, sarr
[2]) && (sarr
[0] == NULL
|| strcmp(cp
, sarr
[0])))
822 sarr
[2] = ok_vlook(ttycharset
);
825 sarrl
[2] = len
= strlen(sarr
[2]);
827 if ((cp
= n_UNCONST(sarr
[1])) != NULL
)
828 len
+= (sarrl
[1] = strlen(cp
));
831 if ((cp
= n_UNCONST(sarr
[0])) != NULL
)
832 len
+= (sarrl
[0] = strlen(cp
));
837 _cs_iter_base
= cp
= n_autorec_alloc(len
+ 1 + 1 +1);
840 if ((len
= sarrl
[0]) != 0) {
841 memcpy(cp
, sarr
[0], len
);
845 if ((len
= sarrl
[1]) != 0) {
846 memcpy(cp
, sarr
[1], len
);
852 memcpy(cp
, sarr
[2], len
);
857 return (_cs_iter
!= NULL
);
861 charset_iter_next(void)
867 rv
= (_cs_iter
!= NULL
);
873 charset_iter_is_valid(void)
878 rv
= (_cs_iter
!= NULL
);
895 charset_iter_or_fallback(void)
906 charset_iter_recurse(char *outer_storage
[2]) /* TODO LEGACY FUN, REMOVE */
909 outer_storage
[0] = _cs_iter_base
;
910 outer_storage
[1] = _cs_iter
;
915 charset_iter_restore(char *outer_storage
[2]) /* TODO LEGACY FUN, REMOVE */
918 _cs_iter_base
= outer_storage
[0];
919 _cs_iter
= outer_storage
[1];
925 need_hdrconv(struct header
*hp
) /* TODO once only, then iter */
927 struct n_header_field
*hfp
;
934 struct n_header_field
*chlp
[3]; /* TODO JOINED AFTER COMPOSE! */
937 chlp
[0] = n_poption_arg_C
;
938 chlp
[1] = n_customhdr_list
;
939 chlp
[2] = hp
->h_user_headers
;
941 for(i
= 0; i
< n_NELEM(chlp
); ++i
)
942 if((hfp
= chlp
[i
]) != NULL
)
943 do if(_has_highbit(hfp
->hf_dat
+ hfp
->hf_nl
+1))
945 while((hfp
= hfp
->hf_next
) != NULL
);
948 if (hp
->h_mft
!= NULL
) {
949 if (_name_highbit(hp
->h_mft
))
952 if (hp
->h_from
!= NULL
) {
953 if (_name_highbit(hp
->h_from
))
955 } else if (_has_highbit(myaddrs(NULL
)))
957 if (hp
->h_reply_to
) {
958 if (_name_highbit(hp
->h_reply_to
))
961 char const *v15compat
;
963 if((v15compat
= ok_vlook(replyto
)) != NULL
)
964 n_OBSOLETE(_("please use *reply-to*, not *replyto*"));
965 if(_has_highbit(v15compat
))
967 if(_has_highbit(ok_vlook(reply_to
)))
971 if (_name_highbit(hp
->h_sender
))
973 } else if (_has_highbit(ok_vlook(sender
)))
976 if (_name_highbit(hp
->h_to
))
978 if (_name_highbit(hp
->h_cc
))
980 if (_name_highbit(hp
->h_bcc
))
982 if (_has_highbit(hp
->h_subject
))
984 rv
= _CS_ITER_GET(); /* TODO MIME/send: iter active? iter! else */
988 #endif /* HAVE_ICONV */
991 mime_fromhdr(struct str
const *in
, struct str
*out
, enum tdflags flags
)
993 /* TODO mime_fromhdr(): is called with strings that contain newlines;
994 * TODO this is the usual newline problem all around the codebase;
995 * TODO i.e., if we strip it, then the display misses it ;>
996 * TODO this is why it is so messy and why S-nail v14.2 plus additional
997 * TODO patch for v14.5.2 (and maybe even v14.5.3 subminor) occurred, and
998 * TODO why our display reflects what is contained in the message: the 1:1
999 * TODO relationship of message content and display!
1000 * TODO instead a header line should be decoded to what it is (a single
1001 * TODO line that is) and it should be objective to the backend whether
1002 * TODO it'll be folded to fit onto the display or not, e.g., for search
1003 * TODO purposes etc. then the only condition we have to honour in here
1004 * TODO is that whitespace in between multiple adjacent MIME encoded words
1005 * TODO á la RFC 2047 is discarded; i.e.: this function should deal with
1006 * TODO RFC 2047 and be renamed: mime_fromhdr() -> mime_rfc2047_decode() */
1007 struct str cin
, cout
;
1008 char *p
, *op
, *upper
;
1009 ui32_t convert
, lastenc
, lastoutl
;
1013 iconv_t fhicd
= (iconv_t
)-1;
1019 *(out
->s
= n_alloc(1)) = '\0';
1025 tcs
= ok_vlook(ttycharset
);
1029 lastenc
= lastoutl
= 0;
1033 if (*p
== '=' && *(p
+ 1) == '?') {
1038 while (p
< upper
&& *p
!= '?')
1039 ++p
; /* strip charset */
1044 if (flags
& TD_ICONV
) {
1045 size_t i
= PTR2SIZE(p
- cbeg
);
1046 char *ltag
, *cs
= n_lofi_alloc(i
);
1048 memcpy(cs
, cbeg
, --i
);
1050 /* RFC 2231 extends the RFC 2047 character set definition in
1051 * encoded words by language tags - silently strip those off */
1052 if ((ltag
= strchr(cs
, '*')) != NULL
)
1055 if (fhicd
!= (iconv_t
)-1)
1056 n_iconv_close(fhicd
);
1057 fhicd
= asccasecmp(cs
, tcs
) ? n_iconv_open(tcs
, cs
) : (iconv_t
)-1;
1063 convert
= CONV_FROMB64
;
1066 convert
= CONV_FROMQP
;
1068 default: /* invalid, ignore */
1076 if (PTRCMP(p
+ 1, >=, upper
))
1078 if (*p
++ == '?' && *p
== '=')
1087 if (convert
== CONV_FROMB64
) {
1088 if(!b64_decode_header(&cout
, &cin
))
1089 n_str_assign_cp(&cout
, _("[Invalid Base64 encoding]"));
1090 }else if(!qp_decode_header(&cout
, &cin
))
1091 n_str_assign_cp(&cout
, _("[Invalid Quoted-Printable encoding]"));
1092 /* Normalize all decoded newlines to spaces XXX only \0/\n yet */
1098 for(any
= FAL0
, i
= cout
.l
; i
-- != 0;)
1111 /* I18N: must be non-empty, last must be closing bracket/xy */
1112 xcp
= _("[Content normalized: ]");
1115 n_str_add_buf(&cout
, xcp
, i
);
1116 memmove(&cout
.s
[i
- 1], cout
.s
, j
);
1117 memcpy(&cout
.s
[0], xcp
, i
- 1);
1118 cout
.s
[cout
.l
- 1] = xcp
[i
- 1];
1125 /* TODO Does not really work if we have assigned some ASCII or even
1126 * TODO translated strings because of errors! */
1127 if ((flags
& TD_ICONV
) && fhicd
!= (iconv_t
)-1) {
1128 cin
.s
= NULL
, cin
.l
= 0; /* XXX string pool ! */
1129 convert
= n_iconv_str(fhicd
, n_ICONV_UNIDEFAULT
, &cin
, &cout
, NULL
);
1130 out
= n_str_add(out
, &cin
);
1131 if (convert
) {/* n_ERR_INVAL at EOS */
1132 n_iconv_reset(fhicd
);
1133 out
= n_str_add_buf(out
, n_qm
, 1); /* TODO unicode replacement */
1138 out
= n_str_add(out
, &cout
);
1139 lastenc
= lastoutl
= out
->l
;
1146 onlyws
= (lastenc
> 0);
1150 if (op
[0] == '=' && (PTRCMP(op
+ 1, ==, upper
) || op
[1] == '?'))
1152 if (onlyws
&& !blankchar(*op
))
1156 out
= n_str_add_buf(out
, p
, PTR2SIZE(op
- p
));
1158 if (!onlyws
|| lastoutl
!= lastenc
)
1163 out
->s
[out
->l
] = '\0';
1165 if (flags
& TD_ISPR
) {
1166 makeprint(out
, &cout
);
1170 if (flags
& TD_DELCTRL
)
1171 out
->l
= delctrl(out
->s
, out
->l
);
1173 if (fhicd
!= (iconv_t
)-1)
1174 n_iconv_close(fhicd
);
1182 mime_fromaddr(char const *name
)
1184 char const *cp
, *lastcp
;
1186 size_t ressz
= 1, rescur
= 0;
1191 if (*name
== '\0') {
1192 res
= savestr(name
);
1196 if ((cp
= routeaddr(name
)) != NULL
&& cp
> name
) {
1197 _append_conv(&res
, &ressz
, &rescur
, name
, PTR2SIZE(cp
- name
));
1202 for ( ; *cp
; ++cp
) {
1205 _append_str(&res
, &ressz
, &rescur
, lastcp
, PTR2SIZE(cp
- lastcp
+ 1));
1207 cp
= skip_comment(cp
);
1209 _append_conv(&res
, &ressz
, &rescur
, lastcp
, PTR2SIZE(cp
- lastcp
));
1216 if (*cp
== '\\' && cp
[1] != '\0')
1223 _append_str(&res
, &ressz
, &rescur
, lastcp
, PTR2SIZE(cp
- lastcp
));
1228 res
= savestrbuf(res
, rescur
);
1238 xmime_write(char const *ptr
, size_t size
, FILE *f
, enum conversion convert
,
1239 enum tdflags dflags
, struct str
* volatile outrest
,
1240 struct str
* volatile inrest
)
1243 struct quoteflt
*qf
;
1246 quoteflt_reset(qf
= quoteflt_dummy(), f
);
1247 rv
= mime_write(ptr
, size
, f
, convert
, dflags
, qf
, outrest
, inrest
);
1253 static sigjmp_buf __mimemw_actjmp
; /* TODO someday.. */
1254 static int __mimemw_sig
; /* TODO someday.. */
1255 static sighandler_type __mimemw_opipe
;
1257 __mimemw_onsig(int sig
) /* TODO someday, we won't need it no more */
1259 NYD_X
; /* Signal handler */
1261 siglongjmp(__mimemw_actjmp
, 1);
1265 mime_write(char const *ptr
, size_t size
, FILE *f
,
1266 enum conversion convert
, enum tdflags
volatile dflags
,
1267 struct quoteflt
*qf
, struct str
* volatile outrest
,
1268 struct str
* volatile inrest
)
1270 /* TODO note: after send/MIME layer rewrite we will have a string pool
1271 * TODO so that memory allocation count drops down massively; for now,
1272 * TODO v14.0 that is, we pay a lot & heavily depend on the allocator.
1273 * TODO P.S.: furthermore all this encapsulated in filter objects instead */
1275 ssize_t
volatile sz
;
1278 dflags
|= _TD_BUFCOPY
;
1279 in
.s
= n_UNCONST(ptr
);
1284 if((sz
= size
) == 0){
1285 if(inrest
!= NULL
&& inrest
->l
!= 0)
1287 if(outrest
!= NULL
&& outrest
->l
!= 0)
1292 /* TODO This crap requires linewise input, then. We need a filter chain
1293 * TODO as in input->iconv->base64 where each filter can have its own
1294 * TODO buffer, with a filter->fflush() call to get rid of those! */
1296 if ((dflags
& TD_ICONV
) && iconvd
!= (iconv_t
)-1 &&
1297 (convert
== CONV_TOQP
|| convert
== CONV_8BIT
||
1298 convert
== CONV_TOB64
|| convert
== CONV_TOHDR
)) {
1299 if (n_iconv_str(iconvd
, n_ICONV_NONE
, &out
, &in
, NULL
) != 0) {
1300 n_iconv_reset(iconvd
);
1301 /* TODO This causes hard-failure. We would need to have an action
1302 * TODO policy FAIL|IGNORE|SETERROR(but continue) */
1308 dflags
&= ~_TD_BUFCOPY
;
1313 if(inrest
!= NULL
&& inrest
->l
> 0){
1319 out
.s
= n_alloc(in
.l
+ inrest
->l
+ 1);
1320 memcpy(out
.s
, inrest
->s
, inrest
->l
);
1322 memcpy(&out
.s
[inrest
->l
], in
.s
, in
.l
);
1325 (in
.s
= out
.s
)[in
.l
+= inrest
->l
] = '\0';
1329 dflags
&= ~_TD_BUFCOPY
;
1334 __mimemw_opipe
= safe_signal(SIGPIPE
, &__mimemw_onsig
);
1335 if (sigsetjmp(__mimemw_actjmp
, 1))
1340 if(!qp_decode_part(&out
, &in
, outrest
, inrest
)){
1341 n_err(_("Invalid Quoted-Printable encoding ignored\n"));
1342 sz
= 0; /* TODO sz = -1 stops outer levels! */
1347 if(qp_encode(&out
, &in
, QP_NONE
) == NULL
){
1348 sz
= 0; /* TODO sz = -1 stops outer levels! */
1353 sz
= quoteflt_push(qf
, in
.s
, in
.l
);
1356 if(!b64_decode_part(&out
, &in
, outrest
, inrest
))
1361 case CONV_FROMB64_T
:
1362 if(!b64_decode_part(&out
, &in
, outrest
, inrest
)){
1364 n_err(_("Invalid Base64 encoding ignored\n"));
1365 sz
= 0; /* TODO sz = -1 stops outer levels! */
1370 if ((sz
= out
.l
) != 0)
1371 sz
= _fwrite_td(&out
, FAL0
, (dflags
& ~_TD_BUFCOPY
), outrest
, qf
);
1374 /* TODO hack which is necessary unless this is a filter based approach
1375 * TODO and each filter has its own buffer (as necessary): we must not
1376 * TODO pass through a number of bytes which causes padding, otherwise we
1377 * TODO produce multiple adjacent base64 streams, and that is not treated
1378 * TODO in the same relaxed fashion like completely bogus bytes by at
1379 * TODO least mutt and OpenSSL. So we need an expensive workaround
1380 * TODO unless we have input->iconv->base64 filter chain as such!! :( */
1381 if(size
!= 0 && /* for Coverity, else assert() */ inrest
!= NULL
){
1382 if(in
.l
> B64_ENCODE_INPUT_PER_LINE
){
1385 i
= in
.l
% B64_ENCODE_INPUT_PER_LINE
;
1389 assert(inrest
->l
== 0);
1390 inrest
->s
= n_realloc(inrest
->s
, i
+1);
1391 memcpy(inrest
->s
, &in
.s
[in
.l
], i
);
1392 inrest
->s
[inrest
->l
= i
] = '\0';
1394 }else if(in
.l
< B64_ENCODE_INPUT_PER_LINE
){
1395 inrest
->s
= n_realloc(inrest
->s
, in
.l
+1);
1396 memcpy(inrest
->s
, in
.s
, in
.l
);
1397 inrest
->s
[inrest
->l
= in
.l
] = '\0';
1403 if(b64_encode(&out
, &in
, B64_LF
| B64_MULTILINE
) == NULL
){
1408 sz
= fwrite(out
.s
, sizeof *out
.s
, out
.l
, f
);
1409 if (sz
!= (ssize_t
)out
.l
)
1413 mime_fromhdr(&in
, &out
, TD_ISPR
| TD_ICONV
| (dflags
& TD_DELCTRL
));
1414 sz
= quoteflt_push(qf
, out
.s
, out
.l
);
1417 sz
= mime_write_tohdr(&in
, f
, NULL
, a_MIME_SH_NONE
);
1422 if(dflags
& _TD_BUFCOPY
){
1423 n_str_dup(&out
, &in
);
1426 dflags
&= ~_TD_BUFCOPY
;
1429 sz
= mime_write_tohdr_a(&in
, f
, &col
, a_MIME_SH_NONE
);
1432 sz
= _fwrite_td(&in
, TRU1
, dflags
, NULL
, qf
);
1441 safe_signal(SIGPIPE
, __mimemw_opipe
);
1442 if (__mimemw_sig
!= 0)
1443 n_raise(__mimemw_sig
);