1 /*@ S-nail - a mail user agent derived from Berkeley Mail.
2 *@ Shell "word", file- and other name expansions, incl. file globbing.
3 *@ TODO v15: peek signal states while opendir/readdir/etc.
5 * Copyright (c) 2000-2004 Gunnar Ritter, Freiburg i. Br., Germany.
6 * Copyright (c) 2012 - 2018 Steffen (Daode) Nurpmeso <steffen@sdaoden.eu>.
9 * Copyright (c) 1980, 1993
10 * The Regents of the University of California. 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. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 #ifndef HAVE_AMALGAMATION
51 * Environment variable names used by the utilities in the Shell and
52 * Utilities volume of POSIX.1-2008 consist solely of uppercase
53 * letters, digits, and the <underscore> ('_') from the characters
54 * defined in Portable Character Set and do not begin with a digit.
55 * Other characters may be permitted by an implementation;
56 * applications shall tolerate the presence of such names.
57 * We do support the hyphen-minus "-" (except in last position for ${x[:]-y}).
58 * We support some special parameter names for one-letter(++) variable names;
59 * these have counterparts in the code that manages internal variables,
60 * and some more special treatment below! */
61 #define a_SHEXP_ISVARC(C) (alnumchar(C) || (C) == '_' || (C) == '-')
62 #define a_SHEXP_ISVARC_BAD1ST(C) (digitchar(C)) /* (Actually assumed below!) */
63 #define a_SHEXP_ISVARC_BADNST(C) ((C) == '-')
65 enum a_shexp_quote_flags
{
67 a_SHEXP_QUOTE_ROUNDTRIP
= 1u<<0, /* Result won't be consumed immediately */
69 a_SHEXP_QUOTE_T_REVSOL
= 1u<<8, /* Type: by reverse solidus */
70 a_SHEXP_QUOTE_T_SINGLE
= 1u<<9, /* Type: single-quotes */
71 a_SHEXP_QUOTE_T_DOUBLE
= 1u<<10, /* Type: double-quotes */
72 a_SHEXP_QUOTE_T_DOLLAR
= 1u<<11, /* Type: dollar-single-quotes */
73 a_SHEXP_QUOTE_T_MASK
= a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
74 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
,
76 a_SHEXP_QUOTE__FREESHIFT
= 16u
80 struct a_shexp_glob_ctx
{
81 char const *sgc_patdat
; /* Remaining pattern (at and below level) */
83 struct n_string
*sgc_outer
; /* Resolved path up to this level */
89 struct a_shexp_quote_ctx
{
90 struct n_string
*sqc_store
; /* Result storage */
91 struct str sqc_input
; /* Input data, topmost level */
93 ui32_t sqc_cnt_single
;
94 ui32_t sqc_cnt_double
;
95 ui32_t sqc_cnt_dollar
;
96 enum a_shexp_quote_flags sqc_flags
;
100 struct a_shexp_quote_lvl
{
101 struct a_shexp_quote_lvl
*sql_link
; /* Outer level */
102 struct str sql_dat
; /* This level (has to) handle(d) */
103 enum a_shexp_quote_flags sql_flags
;
107 /* Locate the user's mailbox file (where new, unread mail is queued) */
108 static char *a_shexp_findmail(char const *user
, bool_t force
);
110 /* Expand ^~/? and ^~USER/? constructs.
111 * Returns the completely resolved (maybe empty or identical to input)
112 * salloc()ed string */
113 static char *a_shexp_tilde(char const *s
);
115 /* Perform fnmatch(3). May return NULL on error */
116 static char *a_shexp_globname(char const *name
, enum fexp_mode fexpm
);
118 static bool_t
a_shexp__glob(struct a_shexp_glob_ctx
*sgcp
,
119 struct n_strlist
**slpp
);
120 static int a_shexp__globsort(void const *cvpa
, void const *cvpb
);
123 /* Parse an input string and create a sh(1)ell-quoted result */
124 static void a_shexp__quote(struct a_shexp_quote_ctx
*sqcp
,
125 struct a_shexp_quote_lvl
*sqlp
);
128 a_shexp_findmail(char const *user
, bool_t force
){
134 if((cp
= ok_vlook(inbox
)) != NULL
&& *cp
!= '\0'){
135 /* _NFOLDER extra introduced to avoid % recursion loops */
136 if((rv
= fexpand(cp
, FEXP_NSPECIAL
| FEXP_NFOLDER
| FEXP_NSHELL
)
139 n_err(_("*inbox* expansion failed, using $MAIL / built-in: %s\n"), cp
);
141 /* Heirloom compatibility: an IMAP *folder* becomes "%" */
143 else if(cp
== NULL
&& !strcmp(user
, ok_vlook(LOGNAME
)) &&
144 which_protocol(cp
= n_folder_query(), FAL0
, FAL0
, NULL
)
146 /* TODO Compat handling of *folder* with IMAP! */
147 n_OBSOLETE("no more expansion of *folder* in \"%\": "
148 "please set *inbox*");
154 if((cp
= ok_vlook(MAIL
)) != NULL
){
163 ul
= strlen(user
) +1;
164 i
= sizeof(VAL_MAIL
) -1 + 1 + ul
;
167 memcpy(rv
, VAL_MAIL
, (i
= sizeof(VAL_MAIL
) -1));
169 memcpy(&rv
[++i
], user
, ul
);
177 a_shexp_tilde(char const *s
){
184 if(*(rp
= &s
[1]) == '/' || *rp
== '\0'){
188 if((rp
= strchr(np
= rp
, '/')) != NULL
){
189 nl
= PTR2SIZE(rp
- np
);
190 np
= savestrbuf(np
, nl
);
195 if((pwp
= getpwnam(np
)) == NULL
){
203 rv
= salloc(nl
+ 1 + rl
+1);
206 memcpy(rv
+ nl
, rp
, rl
);
216 a_shexp_globname(char const *name
, enum fexp_mode fexpm
){
218 struct a_shexp_glob_ctx sgc
;
219 struct n_string outer
;
220 struct n_strlist
*slp
;
224 memset(&sgc
, 0, sizeof sgc
);
225 sgc
.sgc_patlen
= strlen(name
);
226 sgc
.sgc_patdat
= savestrbuf(name
, sgc
.sgc_patlen
);
227 sgc
.sgc_outer
= n_string_reserve(n_string_creat(&outer
), sgc
.sgc_patlen
);
228 sgc
.sgc_flags
= ((fexpm
& FEXP_SILENT
) != 0);
230 if(a_shexp__glob(&sgc
, &slp
))
234 n_string_gut(&outer
);
240 cp
= n_UNCONST(N_("File pattern does not match"));
242 }else if(slp
->sl_next
== NULL
)
243 cp
= savestrbuf(slp
->sl_dat
, slp
->sl_len
);
244 else if(fexpm
& FEXP_MULTIOK
){
245 struct n_strlist
**sorta
, *xslp
;
249 for(xslp
= slp
; xslp
!= NULL
; xslp
= xslp
->sl_next
){
251 l
+= xslp
->sl_len
+ 1;
254 sorta
= smalloc(sizeof(*sorta
) * no
);
256 for(xslp
= slp
; xslp
!= NULL
; xslp
= xslp
->sl_next
)
258 qsort(sorta
, no
, sizeof *sorta
, &a_shexp__globsort
);
262 for(i
= 0; i
< no
; ++i
){
264 memcpy(&cp
[l
], xslp
->sl_dat
, xslp
->sl_len
);
271 n_pstate
|= n_PS_EXPAND_MULTIRESULT
;
273 cp
= n_UNCONST(N_("File pattern matches multiple results"));
279 struct n_strlist
*tmp
= slp
;
288 if(!(fexpm
& FEXP_SILENT
)){
289 name
= n_shexp_quote_cp(name
, FAL0
);
290 n_err("%s: %s\n", V_(cp
), name
);
295 #else /* HAVE_FNMATCH */
298 if(!(fexpm
& FEXP_SILENT
))
299 n_err(_("No filename pattern (fnmatch(3)) support compiled in\n"));
300 return savestr(name
);
306 a_shexp__glob(struct a_shexp_glob_ctx
*sgcp
, struct n_strlist
**slpp
){
307 enum{a_SILENT
= 1<<0, a_DEEP
=1<<1, a_SALLOC
=1<<2};
309 struct a_shexp_glob_ctx nsgc
;
313 char const *ccp
, *myp
;
316 /* We need some special treatment for the outermost level */
317 if(!(sgcp
->sgc_flags
& a_DEEP
)){
318 if(sgcp
->sgc_patlen
> 0 && sgcp
->sgc_patdat
[0] == '/'){
319 myp
= n_string_cp(n_string_push_c(sgcp
->sgc_outer
, '/'));
325 myp
= n_string_cp(sgcp
->sgc_outer
);
326 old_outerlen
= sgcp
->sgc_outer
->s_len
;
328 /* Separate current directory/pattern level from any possible remaining
329 * pattern in order to be able to use it for fnmatch(3) */
330 if((ccp
= memchr(sgcp
->sgc_patdat
, '/', sgcp
->sgc_patlen
)) == NULL
)
334 nsgc
.sgc_flags
|= a_DEEP
;
335 sgcp
->sgc_patlen
= PTR2SIZE((nsgc
.sgc_patdat
= &ccp
[1]) -
336 &sgcp
->sgc_patdat
[0]);
337 nsgc
.sgc_patlen
-= sgcp
->sgc_patlen
;
339 if(sgcp
->sgc_patlen
> 0){
340 assert(sgcp
->sgc_patdat
[sgcp
->sgc_patlen
-1] == '/');
341 ((char*)n_UNCONST(sgcp
->sgc_patdat
))[--sgcp
->sgc_patlen
] = '\0';
345 /* Our current directory level */
346 /* xxx Plenty of room for optimizations, like quickshot lstat(2) which may
347 * xxx be the (sole) result depending on pattern surroundings, etc. */
348 if((dp
= opendir(myp
)) == NULL
){
351 switch((err
= n_err_no
)){
353 ccp
= N_("cannot access paths under non-directory");
356 ccp
= N_("path component of (sub)pattern non-existent");
359 ccp
= N_("file permission for file (sub)pattern denied");
363 ccp
= N_("file descriptor limit reached, cannot open directory");
366 ccp
= N_("cannot open path component as directory");
371 /* As necessary, quote bytes in the current pattern */
377 for(need
= FAL0
, i
= 0, myp
= sgcp
->sgc_patdat
; *myp
!= '\0'; ++myp
)
379 case '\'': case '"': case '\\': case '$':
391 for(i
= 0, myp
= sgcp
->sgc_patdat
; *myp
!= '\0'; ++myp
)
393 case '\'': case '"': case '\\': case '$':
404 myp
= sgcp
->sgc_patdat
;
407 while((dep
= readdir(dp
)) != NULL
){
408 switch(fnmatch(myp
, dep
->d_name
, FNM_PATHNAME
| FNM_PERIOD
)){
410 /* A match expresses the desire to recurse if there is more pattern */
411 if(nsgc
.sgc_patlen
> 0){
414 n_string_push_cp((sgcp
->sgc_outer
->s_len
> 1
415 ? n_string_push_c(sgcp
->sgc_outer
, '/') : sgcp
->sgc_outer
),
419 #ifdef HAVE_DIRENT_TYPE
420 if(dep
->d_type
== DT_DIR
)
422 else if(dep
->d_type
== DT_LNK
|| dep
->d_type
== DT_UNKNOWN
)
427 if(stat(n_string_cp(sgcp
->sgc_outer
), &sb
)){
428 ccp
= N_("I/O error when querying file status");
430 }else if(S_ISDIR(sb
.st_mode
))
434 /* TODO We recurse with current dir FD open, which could E[MN]FILE!
435 * TODO Instead save away a list of such n_string's for later */
436 if(isdir
&& !a_shexp__glob(&nsgc
, slpp
)){
441 n_string_trunc(sgcp
->sgc_outer
, old_outerlen
);
443 struct n_strlist
*slp
;
446 i
= strlen(dep
->d_name
);
447 j
= (old_outerlen
> 0) ? old_outerlen
+ 1 + i
: i
;
448 slp
= n_STRLIST_ALLOC(j
);
450 slpp
= &slp
->sl_next
;
452 if((j
= old_outerlen
) > 0){
453 memcpy(&slp
->sl_dat
[0], sgcp
->sgc_outer
->s_dat
, j
);
454 if(slp
->sl_dat
[j
-1] != '/')
455 slp
->sl_dat
[j
++] = '/';
457 memcpy(&slp
->sl_dat
[j
], dep
->d_name
, i
);
458 slp
->sl_dat
[j
+= i
] = '\0';
465 ccp
= N_("fnmatch(3) cannot handle file (sub)pattern");
475 return (ccp
== NULL
);
478 if(!(sgcp
->sgc_flags
& a_SILENT
)){
481 if(sgcp
->sgc_outer
->s_len
> 0){
482 s2
= n_shexp_quote_cp(n_string_cp(sgcp
->sgc_outer
), FAL0
);
487 n_err("%s: %s%s%s\n", V_(ccp
), s2
, s3
,
488 n_shexp_quote_cp(sgcp
->sgc_patdat
, FAL0
));
494 a_shexp__globsort(void const *cvpa
, void const *cvpb
){
496 struct n_strlist
const * const *slpa
, * const *slpb
;
501 rv
= asccasecmp((*slpa
)->sl_dat
, (*slpb
)->sl_dat
);
505 #endif /* HAVE_FNMATCH */
508 a_shexp__quote(struct a_shexp_quote_ctx
*sqcp
, struct a_shexp_quote_lvl
*sqlp
){
509 /* XXX Because of the problems caused by ISO C multibyte interface we cannot
510 * XXX use the recursive implementation because of stateful encodings.
511 * XXX I.e., if a quoted substring cannot be self-contained - the data after
512 * XXX the quote relies on "the former state", then this doesn't make sense.
513 * XXX Therefore this is not fully programmed out but instead only detects
514 * XXX the "most fancy" quoting necessary, and directly does that.
515 * XXX As a result of this, T_REVSOL and T_DOUBLE are not even considered.
516 * XXX Otherwise we rather have to convert to wide first and act on that,
517 * XXX e.g., call visual_info(n_VISUAL_INFO_WOUT_CREATE) on entire input */
518 #undef a_SHEXP_QUOTE_RECURSE /* XXX (Needs complete revisit, then) */
519 #ifdef a_SHEXP_QUOTE_RECURSE
520 # define jrecurse jrecurse
521 struct a_shexp_quote_lvl sql
;
523 # define jrecurse jstep
525 struct n_visual_info_ctx vic
;
526 union {struct a_shexp_quote_lvl
*head
; struct n_string
*store
;} u
;
529 char const *ib
, *ib_base
;
532 ib_base
= ib
= sqlp
->sql_dat
.s
;
533 il
= sqlp
->sql_dat
.l
;
534 flags
= sqlp
->sql_flags
;
536 /* Iterate over the entire input, classify characters and type of quotes
537 * along the way. Whenever a quote change has to be applied, adjust flags
538 * for the new situation -, setup sql.* and recurse- */
544 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
)
546 if(c
== '\t' && (flags
& (a_SHEXP_QUOTE_T_REVSOL
|
547 a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOUBLE
)))
549 #ifdef a_SHEXP_QUOTE_RECURSE
550 ++sqcp
->sqc_cnt_dollar
;
552 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
554 }else if(blankspacechar(c
) || c
== '|' || c
== '&' || c
== ';' ||
555 /* Whereas we don't support those, quote them for the sh(1)ell */
556 c
== '(' || c
== ')' || c
== '<' || c
== '>' ||
557 c
== '"' || c
== '$'){
558 if(flags
& a_SHEXP_QUOTE_T_MASK
)
560 #ifdef a_SHEXP_QUOTE_RECURSE
561 ++sqcp
->sqc_cnt_single
;
563 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_SINGLE
;
566 if(flags
& (a_SHEXP_QUOTE_T_MASK
& ~a_SHEXP_QUOTE_T_SINGLE
))
568 #ifdef a_SHEXP_QUOTE_RECURSE
569 ++sqcp
->sqc_cnt_dollar
;
571 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
573 }else if(c
== '\\' || (c
== '#' && ib
== ib_base
)){
574 if(flags
& a_SHEXP_QUOTE_T_MASK
)
576 #ifdef a_SHEXP_QUOTE_RECURSE
577 ++sqcp
->sqc_cnt_single
;
579 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_SINGLE
;
581 }else if(!asciichar(c
)){
582 /* Need to keep together multibytes */
583 #ifdef a_SHEXP_QUOTE_RECURSE
584 memset(&vic
, 0, sizeof vic
);
588 n_VISUAL_INFO_ONE_CHAR
| n_VISUAL_INFO_SKIP_ERRORS
);
590 /* xxx check whether resulting \u would be ASCII */
591 if(!(flags
& a_SHEXP_QUOTE_ROUNDTRIP
) ||
592 (flags
& a_SHEXP_QUOTE_T_DOLLAR
)){
593 #ifdef a_SHEXP_QUOTE_RECURSE
601 #ifdef a_SHEXP_QUOTE_RECURSE
602 ++sqcp
->sqc_cnt_dollar
;
604 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
610 sqlp
->sql_flags
= flags
;
612 /* Level made the great and completed processing input. Reverse the list of
613 * levels, detect the "most fancy" quote type needed along this way */
614 /* XXX Due to restriction as above very crude */
615 for(flags
= 0, il
= 0, u
.head
= NULL
; sqlp
!= NULL
;){
616 struct a_shexp_quote_lvl
*tmp
;
618 tmp
= sqlp
->sql_link
;
619 sqlp
->sql_link
= u
.head
;
621 il
+= sqlp
->sql_dat
.l
;
622 if(sqlp
->sql_flags
& a_SHEXP_QUOTE_T_MASK
)
623 il
+= (sqlp
->sql_dat
.l
>> 1);
624 flags
|= sqlp
->sql_flags
;
629 /* Finally work the substrings in the correct order, adjusting quotes along
630 * the way as necessary. Start off with the "most fancy" quote, so that
631 * the user sees an overall boundary she can orientate herself on.
632 * We do it like that to be able to give the user some "encapsulation
633 * experience", to address what strikes me is a problem of sh(1)ell quoting:
634 * different to, e.g., perl(1), where you see at a glance where a string
635 * starts and ends, sh(1) quoting occurs at the "top level", disrupting the
636 * visual appearance of "a string" as such */
637 u
.store
= n_string_reserve(sqcp
->sqc_store
, il
);
639 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
){
640 u
.store
= n_string_push_buf(u
.store
, "$'", sizeof("$'") -1);
641 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOLLAR
;
642 }else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
){
643 u
.store
= n_string_push_c(u
.store
, '"');
644 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_DOUBLE
;
645 }else if(flags
& a_SHEXP_QUOTE_T_SINGLE
){
646 u
.store
= n_string_push_c(u
.store
, '\'');
647 flags
= (flags
& ~a_SHEXP_QUOTE_T_MASK
) | a_SHEXP_QUOTE_T_SINGLE
;
648 }else /*if(flags & a_SHEXP_QUOTE_T_REVSOL)*/
649 flags
&= ~a_SHEXP_QUOTE_T_MASK
;
651 /* Work all the levels */
652 for(; sqlp
!= NULL
; sqlp
= sqlp
->sql_link
){
653 /* As necessary update our mode of quoting */
654 #ifdef a_SHEXP_QUOTE_RECURSE
657 switch(sqlp
->sql_flags
& a_SHEXP_QUOTE_T_MASK
){
658 case a_SHEXP_QUOTE_T_DOLLAR
:
659 if(!(flags
& a_SHEXP_QUOTE_T_DOLLAR
))
660 il
= a_SHEXP_QUOTE_T_DOLLAR
;
662 case a_SHEXP_QUOTE_T_DOUBLE
:
663 if(!(flags
& (a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)))
664 il
= a_SHEXP_QUOTE_T_DOLLAR
;
666 case a_SHEXP_QUOTE_T_SINGLE
:
667 if(!(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
668 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)))
669 il
= a_SHEXP_QUOTE_T_SINGLE
;
672 case a_SHEXP_QUOTE_T_REVSOL
:
673 if(!(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
674 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)))
675 il
= a_SHEXP_QUOTE_T_REVSOL
;
680 if(flags
& (a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOLLAR
))
681 u
.store
= n_string_push_c(u
.store
, '\'');
682 else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
683 u
.store
= n_string_push_c(u
.store
, '"');
684 flags
&= ~a_SHEXP_QUOTE_T_MASK
;
687 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
)
688 u
.store
= n_string_push_buf(u
.store
, "$'", sizeof("$'") -1);
689 else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
690 u
.store
= n_string_push_c(u
.store
, '"');
691 else if(flags
& a_SHEXP_QUOTE_T_SINGLE
)
692 u
.store
= n_string_push_c(u
.store
, '\'');
694 #endif /* a_SHEXP_QUOTE_RECURSE */
696 /* Work the level's substring */
697 ib
= sqlp
->sql_dat
.s
;
698 il
= sqlp
->sql_dat
.l
;
706 assert(c
== '\t' || (flags
& a_SHEXP_QUOTE_T_DOLLAR
));
707 assert((flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_SINGLE
|
708 a_SHEXP_QUOTE_T_DOUBLE
| a_SHEXP_QUOTE_T_DOLLAR
)));
710 case 0x07: c
= 'a'; break;
711 case 0x08: c
= 'b'; break;
712 case 0x0A: c
= 'n'; break;
713 case 0x0B: c
= 'v'; break;
714 case 0x0C: c
= 'f'; break;
715 case 0x0D: c
= 'r'; break;
716 case 0x1B: c
= 'E'; break;
719 if(flags
& a_SHEXP_QUOTE_T_DOLLAR
){
723 if(flags
& a_SHEXP_QUOTE_T_REVSOL
)
724 u
.store
= n_string_push_c(u
.store
, '\\');
727 u
.store
= n_string_push_c(u
.store
, '\\');
729 u
.store
= n_string_push_c(u
.store
, 'c');
733 }else if(blankspacechar(c
) || c
== '|' || c
== '&' || c
== ';' ||
734 /* Whereas we don't support those, quote them for the sh(1)ell */
735 c
== '(' || c
== ')' || c
== '<' || c
== '>' ||
736 c
== '"' || c
== '$'){
737 if(flags
& (a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOLLAR
))
739 assert(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_DOUBLE
));
740 u
.store
= n_string_push_c(u
.store
, '\\');
743 if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
745 assert(!(flags
& a_SHEXP_QUOTE_T_SINGLE
));
746 u
.store
= n_string_push_c(u
.store
, '\\');
748 }else if(c
== '\\' || (c
== '#' && ib
== ib_base
)){
749 if(flags
& a_SHEXP_QUOTE_T_SINGLE
)
751 assert(flags
& (a_SHEXP_QUOTE_T_REVSOL
| a_SHEXP_QUOTE_T_DOUBLE
|
752 a_SHEXP_QUOTE_T_DOLLAR
));
753 u
.store
= n_string_push_c(u
.store
, '\\');
755 }else if(asciichar(c
)){
756 /* Shorthand: we can simply push that thing out */
758 u
.store
= n_string_push_c(u
.store
, c
);
761 /* Not an ASCII character, take care not to split up multibyte
762 * sequences etc. For the sake of compile testing, don't enwrap in
763 * HAVE_ALWAYS_UNICODE_LOCALE || HAVE_NATCH_CHAR */
764 if(n_psonce
& n_PSO_UNICODE
){
771 if((uc
= n_utf8_to_utf32(&ib2
, &il2
)) != UI32_MAX
){
775 il2
= PTR2SIZE(&ib2
[0] - &ib
[0]);
776 if((flags
& a_SHEXP_QUOTE_ROUNDTRIP
) || uc
== 0xFFFDu
){
777 /* Use padding to make ambiguities impossible */
778 il3
= snprintf(itoa
, sizeof itoa
, "\\%c%0*X",
779 (uc
> 0xFFFFu
? 'U' : 'u'),
780 (int)(uc
> 0xFFFFu
? 8 : 4), uc
);
786 u
.store
= n_string_push_buf(u
.store
, cp
, il3
);
787 ib
+= il2
, il
-= il2
;
792 memset(&vic
, 0, sizeof vic
);
796 n_VISUAL_INFO_ONE_CHAR
| n_VISUAL_INFO_SKIP_ERRORS
);
798 /* Work this substring as sensitive as possible */
800 if(!(flags
& a_SHEXP_QUOTE_ROUNDTRIP
))
801 u
.store
= n_string_push_buf(u
.store
, ib
, il
);
803 else if((vic
.vic_indat
= n_iconv_onetime_cp(n_ICONV_NONE
,
804 "utf-8", ok_vlook(ttycharset
), savestrbuf(ib
, il
))) != NULL
){
809 il2
= strlen(ib2
= vic
.vic_indat
);
810 if((uc
= n_utf8_to_utf32(&ib2
, &il2
)) != UI32_MAX
){
813 il2
= PTR2SIZE(&ib2
[0] - &vic
.vic_indat
[0]);
814 /* Use padding to make ambiguities impossible */
815 il3
= snprintf(itoa
, sizeof itoa
, "\\%c%0*X",
816 (uc
> 0xFFFFu
? 'U' : 'u'),
817 (int)(uc
> 0xFFFFu
? 8 : 4), uc
);
818 u
.store
= n_string_push_buf(u
.store
, itoa
, il3
);
828 u
.store
= n_string_push_buf(u
.store
, "\\xFF",
830 n_c_to_hex_base16(&u
.store
->s_dat
[u
.store
->s_len
- 2], *ib
++);
839 /* Close an open quote */
840 if(flags
& (a_SHEXP_QUOTE_T_SINGLE
| a_SHEXP_QUOTE_T_DOLLAR
))
841 u
.store
= n_string_push_c(u
.store
, '\'');
842 else if(flags
& a_SHEXP_QUOTE_T_DOUBLE
)
843 u
.store
= n_string_push_c(u
.store
, '"');
844 #ifdef a_SHEXP_QUOTE_RECURSE
850 #ifdef a_SHEXP_QUOTE_RECURSE
852 sqlp
->sql_dat
.l
-= il
;
855 sql
.sql_dat
.s
= n_UNCONST(ib
);
857 sql
.sql_flags
= flags
;
858 a_shexp__quote(sqcp
, &sql
);
863 #undef a_SHEXP_QUOTE_RECURSE
867 fexpand(char const *name
, enum fexp_mode fexpm
) /* TODO in parts: -> URL::!! */
870 char const *res
, *cp
;
871 bool_t dyn
, haveproto
;
874 n_pstate
&= ~n_PS_EXPAND_MULTIRESULT
;
877 /* The order of evaluation is "%" and "#" expand into constants.
878 * "&" can expand into "+". "+" can expand into shell meta characters.
879 * Shell meta characters expand into constants.
880 * This way, we make no recursive expansion */
881 if((fexpm
& FEXP_NSHORTCUT
) || (res
= shortcut_expand(name
)) == NULL
)
882 res
= n_UNCONST(name
);
885 n_UNINIT(proto
.s
, NULL
), n_UNINIT(proto
.l
, 0);
887 for(cp
= res
; *cp
&& *cp
!= ':'; ++cp
)
890 if(cp
[0] == ':' && cp
[1] == '/' && cp
[2] == '/'){
892 proto
.s
= n_UNCONST(res
);
894 proto
.l
= PTR2SIZE(cp
- res
);
899 if(!(fexpm
& FEXP_NSPECIAL
)){
904 if(res
[1] == ':' && res
[2] != '\0'){
910 force
= (res
[1] != '\0');
911 res
= a_shexp_findmail((force
? &res
[1] : ok_vlook(LOGNAME
)),
920 if (prevfile
[0] == '\0') {
921 n_err(_("No previous file\n"));
929 res
= ok_vlook(MBOX
);
937 if(res
[0] == '@' && which_protocol(mailname
, FAL0
, FAL0
, NULL
)
939 res
= str_concat_csvl(&s
, protbase(mailname
), "/", &res
[1], NULL
)->s
;
944 /* POSIX: if *folder* unset or null, "+" shall be retained */
945 if(!(fexpm
& FEXP_NFOLDER
) && *res
== '+' &&
946 *(cp
= n_folder_query()) != '\0'){
947 res
= str_concat_csvl(&s
, cp
, &res
[1], NULL
)->s
;
951 /* Do some meta expansions */
952 if((fexpm
& (FEXP_NSHELL
| FEXP_NVAR
)) != FEXP_NVAR
&&
953 ((fexpm
& FEXP_NSHELL
) ? (strchr(res
, '$') != NULL
)
954 : n_anyof_cp("{}[]*?$", res
))){
957 if(fexpm
& FEXP_NOPROTO
)
960 cp
= haveproto
? savecat(savestrbuf(proto
.s
, proto
.l
), res
) : res
;
962 switch(which_protocol(cp
, TRU1
, FAL0
, NULL
)){
975 struct n_string shou
, *shoup
;
977 shin
.s
= n_UNCONST(res
);
979 shoup
= n_string_creat_auto(&shou
);
981 enum n_shexp_state shs
;
983 /* TODO shexp: take care to not include backtick eval once avail! */
984 shs
= n_shexp_parse_token((n_SHEXP_PARSE_LOG_D_V
|
985 n_SHEXP_PARSE_QUOTE_AUTO_FIXED
| n_SHEXP_PARSE_QUOTE_AUTO_DQ
|
986 n_SHEXP_PARSE_QUOTE_AUTO_CLOSE
), shoup
, &shin
, NULL
);
987 if(shs
& n_SHEXP_STATE_STOP
)
990 res
= n_string_cp(shoup
);
991 /*shoup = n_string_drop_ownership(shoup);*/
995 res
= a_shexp_tilde(res
);
997 if(!(fexpm
& FEXP_NSHELL
) &&
998 (res
= a_shexp_globname(res
, fexpm
)) == NULL
)
1001 }/* else no tilde */
1002 }else if(res
[0] == '~'){
1003 res
= a_shexp_tilde(res
);
1008 if(res
!= NULL
&& haveproto
){
1009 res
= savecat(savestrbuf(proto
.s
, proto
.l
), res
);
1013 if(fexpm
& FEXP_LOCAL
){
1014 switch (which_protocol(res
, FAL0
, FAL0
, NULL
)) {
1016 case PROTO_MAILDIR
: /* Cannot happen since we don't stat(2), but.. */
1019 n_err(_("Not a local file or directory: %s\n"),
1020 n_shexp_quote_cp(name
, FAL0
));
1027 if(res
!= NULL
&& !dyn
)
1030 return n_UNCONST(res
);
1033 FL
enum n_shexp_state
1034 n_shexp_parse_token(enum n_shexp_parse_flags flags
, struct n_string
*store
,
1035 struct str
*input
, void const **cookie
){
1036 /* TODO shexp_parse_token: WCHAR */
1037 ui32_t last_known_meta_trim_len
;
1038 char c2
, c
, quotec
, utf
[8];
1039 enum n_shexp_state rv
;
1041 char const *ifs
, *ifs_ws
, *ib_save
, *ib
;
1044 a_SKIPQ
= 1u<<0, /* Skip rest of this quote (\u0 ..) */
1045 a_SKIPT
= 1u<<1, /* Skip entire token (\c@) */
1046 a_SKIPMASK
= a_SKIPQ
| a_SKIPT
,
1047 a_SURPLUS
= 1u<<2, /* Extended sequence interpretation */
1048 a_NTOKEN
= 1u<<3, /* "New token": e.g., comments are possible */
1049 a_BRACE
= 1u<<4, /* Variable substitution: brace enclosed */
1050 a_DIGIT1
= 1u<<5, /* ..first character was digit */
1051 a_NONDIGIT
= 1u<<6, /* ..has seen any non-digits */
1052 a_VARSUBST_MASK
= n_BITENUM_MASK(4, 6),
1054 a_ROUND_MASK
= a_SKIPT
| (int)~n_BITENUM_MASK(0, 7),
1057 a_CONSUME
= 1u<<10, /* When done, "consume" remaining input */
1062 assert((flags
& n_SHEXP_PARSE_DRYRUN
) || store
!= NULL
);
1063 assert(input
!= NULL
);
1064 assert(input
->l
== 0 || input
->s
!= NULL
);
1065 assert(!(flags
& n_SHEXP_PARSE_LOG
) || !(flags
& n_SHEXP_PARSE_LOG_D_V
));
1066 assert(!(flags
& n_SHEXP_PARSE_IFS_ADD_COMMA
) ||
1067 !(flags
& n_SHEXP_PARSE_IFS_IS_COMMA
));
1068 assert(!(flags
& n_SHEXP_PARSE_QUOTE_AUTO_FIXED
) ||
1069 (flags
& n__SHEXP_PARSE_QUOTE_AUTO_MASK
));
1071 if((flags
& n_SHEXP_PARSE_LOG_D_V
) && (n_poption
& n_PO_D_V
))
1072 flags
|= n_SHEXP_PARSE_LOG
;
1073 if(flags
& n_SHEXP_PARSE_QUOTE_AUTO_FIXED
)
1074 flags
|= n_SHEXP_PARSE_QUOTE_AUTO_CLOSE
;
1076 if((flags
& n_SHEXP_PARSE_TRUNC
) && store
!= NULL
)
1077 store
= n_string_trunc(store
, 0);
1079 if(flags
& (n_SHEXP_PARSE_IFS_VAR
| n_SHEXP_PARSE_TRIM_IFSSPACE
)){
1080 ifs
= ok_vlook(ifs
);
1081 ifs_ws
= ok_vlook(ifs_ws
);
1083 n_UNINIT(ifs
, n_empty
);
1084 n_UNINIT(ifs_ws
, n_empty
);
1089 if((il
= input
->l
) == UIZ_MAX
)
1090 input
->l
= il
= strlen(ib
);
1093 if(cookie
!= NULL
&& *cookie
!= NULL
){
1094 assert(!(flags
& n_SHEXP_PARSE_DRYRUN
));
1098 rv
= n_SHEXP_STATE_NONE
;
1100 rv
&= n_SHEXP_STATE_WS_LEAD
;
1101 state
&= a_ROUND_MASK
;
1103 /* In cookie mode, the next ARGV entry is the token already, unchanged,
1104 * since it has already been expanded before! */
1105 if(state
& a_COOKIE
){
1106 char const * const *xcookie
, *cp
;
1110 if((store
= n_string_push_cp(store
, *xcookie
))->s_len
> 0)
1111 rv
|= n_SHEXP_STATE_OUTPUT
;
1112 if(*++xcookie
== NULL
){
1115 flags
|= n_SHEXP_PARSE_QUOTE_AUTO_DQ
; /* ..why we are here! */
1117 *cookie
= n_UNCONST(xcookie
);
1119 for(cp
= &n_string_cp(store
)[i
]; (c
= *cp
++) != '\0';)
1121 rv
|= n_SHEXP_STATE_CONTROL
;
1125 /* The last exploded cookie will join with the yielded input token, so
1126 * simply fall through in this case */
1127 if(state
& a_COOKIE
)
1131 if(flags
& n_SHEXP_PARSE_TRIM_SPACE
){
1132 for(; il
> 0; ++ib
, --il
){
1133 if(!blankspacechar(*ib
))
1135 rv
|= n_SHEXP_STATE_WS_LEAD
;
1139 if(flags
& n_SHEXP_PARSE_TRIM_IFSSPACE
){
1140 for(; il
> 0; ++ib
, --il
){
1141 if(strchr(ifs_ws
, *ib
) == NULL
)
1143 rv
|= n_SHEXP_STATE_WS_LEAD
;
1147 input
->s
= n_UNCONST(ib
);
1152 rv
|= n_SHEXP_STATE_STOP
;
1157 store
= n_string_reserve(store
, n_MIN(il
, 32)); /* XXX */
1159 switch(flags
& n__SHEXP_PARSE_QUOTE_AUTO_MASK
){
1160 case n_SHEXP_PARSE_QUOTE_AUTO_SQ
:
1162 rv
|= n_SHEXP_STATE_QUOTE
;
1164 case n_SHEXP_PARSE_QUOTE_AUTO_DQ
:
1167 case n_SHEXP_PARSE_QUOTE_AUTO_DSQ
:
1170 rv
|= n_SHEXP_STATE_QUOTE
;
1179 /* TODO n_SHEXP_PARSE_META_SEMICOLON++, well, hack: we are not the shell,
1180 * TODO we are not a language, and therefore the general *ifs-ws* and normal
1181 * TODO whitespace trimming that input lines undergo (in a_go_evaluate())
1182 * TODO has already happened, our result will be used *as is*, and therefore
1183 * TODO we need to be aware of and remove trailing unquoted WS that would
1184 * TODO otherwise remain, after we have seen a semicolon sequencer.
1185 * By sheer luck we only need to track this in non-quote-mode */
1186 last_known_meta_trim_len
= UI32_MAX
;
1191 /* If no quote-mode active.. */
1193 if(c
== '"' || c
== '\''){
1198 state
&= ~a_SURPLUS
;
1200 last_known_meta_trim_len
= UI32_MAX
;
1201 rv
|= n_SHEXP_STATE_QUOTE
;
1206 last_known_meta_trim_len
= UI32_MAX
;
1211 rv
|= n_SHEXP_STATE_QUOTE
;
1216 }else if(c
== '\\'){
1217 /* Outside of quotes this just escapes any next character, but a sole
1218 * <reverse solidus> at EOS is left unchanged */
1222 last_known_meta_trim_len
= UI32_MAX
;
1224 /* A comment may it be if no token has yet started */
1225 else if(c
== '#' && (state
& a_NTOKEN
)){
1226 rv
|= n_SHEXP_STATE_STOP
;
1227 /*last_known_meta_trim_len = UI32_MAX;*/
1230 /* Metacharacters which separate tokens must be turned on explicitly */
1231 else if(c
== '|' && (flags
& n_SHEXP_PARSE_META_VERTBAR
)){
1232 rv
|= n_SHEXP_STATE_META_VERTBAR
;
1234 /* The parsed sequence may be _the_ output, so ensure we don't
1235 * include the metacharacter, then. */
1236 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1238 /*last_known_meta_trim_len = UI32_MAX;*/
1240 }else if(c
== '&' && (flags
& n_SHEXP_PARSE_META_AMPERSAND
)){
1241 rv
|= n_SHEXP_STATE_META_AMPERSAND
;
1243 /* The parsed sequence may be _the_ output, so ensure we don't
1244 * include the metacharacter, then. */
1245 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1247 /*last_known_meta_trim_len = UI32_MAX;*/
1249 }else if(c
== ';' && (flags
& n_SHEXP_PARSE_META_SEMICOLON
)){
1251 n_go_input_inject(n_GO_INPUT_INJECT_COMMIT
, ib
, il
);
1252 rv
|= n_SHEXP_STATE_META_SEMICOLON
| n_SHEXP_STATE_STOP
;
1254 if(!(flags
& n_SHEXP_PARSE_DRYRUN
) && (rv
& n_SHEXP_STATE_OUTPUT
) &&
1255 last_known_meta_trim_len
!= UI32_MAX
)
1256 store
= n_string_trunc(store
, last_known_meta_trim_len
);
1258 /* The parsed sequence may be _the_ output, so ensure we don't
1259 * include the metacharacter, then. */
1260 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1262 /*last_known_meta_trim_len = UI32_MAX;*/
1264 }else if(c
== ',' && (flags
&
1265 (n_SHEXP_PARSE_IFS_ADD_COMMA
| n_SHEXP_PARSE_IFS_IS_COMMA
))){
1266 /* The parsed sequence may be _the_ output, so ensure we don't
1267 * include the metacharacter, then. */
1268 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1270 /*last_known_meta_trim_len = UI32_MAX;*/
1275 blnk
= blankchar(c
) ? 1 : 0;
1276 blnk
|= ((flags
& (n_SHEXP_PARSE_IFS_VAR
|
1277 n_SHEXP_PARSE_TRIM_IFSSPACE
)) &&
1278 strchr(ifs_ws
, c
) != NULL
) ? 2 : 0;
1280 if((!(flags
& n_SHEXP_PARSE_IFS_VAR
) && (blnk
& 1)) ||
1281 ((flags
& n_SHEXP_PARSE_IFS_VAR
) &&
1282 ((blnk
& 2) || strchr(ifs
, c
) != NULL
))){
1283 if(!(flags
& n_SHEXP_PARSE_IFS_IS_COMMA
)){
1284 /* The parsed sequence may be _the_ output, so ensure we don't
1285 * include the metacharacter, then. */
1286 if(flags
& (n_SHEXP_PARSE_DRYRUN
| n_SHEXP_PARSE_META_KEEP
))
1288 /*last_known_meta_trim_len = UI32_MAX;*/
1295 if(blnk
&& store
!= NULL
){
1296 if(last_known_meta_trim_len
== UI32_MAX
)
1297 last_known_meta_trim_len
= store
->s_len
;
1299 last_known_meta_trim_len
= UI32_MAX
;
1303 assert(!(state
& a_NTOKEN
));
1304 if(c
== quotec
&& !(flags
& n_SHEXP_PARSE_QUOTE_AUTO_FIXED
)){
1305 state
&= a_ROUND_MASK
;
1307 /* Users may need to recognize the presence of empty quotes */
1308 rv
|= n_SHEXP_STATE_OUTPUT
;
1310 }else if(c
== '\\' && (state
& a_SURPLUS
)){
1312 /* A sole <reverse solidus> at EOS is treated as-is! This is ok
1313 * since the "closing quote" error will occur next, anyway */
1316 else if((c2
= *ib
) == quotec
){
1319 }else if(quotec
== '"'){
1320 /* Double quotes, POSIX says:
1321 * The <backslash> shall retain its special meaning as an
1322 * escape character (see Section 2.2.1) only when followed
1323 * by one of the following characters when considered
1324 * special: $ ` " \ <newline> */
1328 /* case '"': already handled via c2 == quotec */
1337 /* Dollar-single-quote */
1341 /* case '\'': already handled via c2 == quotec */
1346 case 'b': c
= '\b'; break;
1347 case 'f': c
= '\f'; break;
1348 case 'n': c
= '\n'; break;
1349 case 'r': c
= '\r'; break;
1350 case 't': c
= '\t'; break;
1351 case 'v': c
= '\v'; break;
1354 case 'e': c
= '\033'; break;
1356 /* Control character */
1359 goto j_dollar_ungetc
;
1361 if(state
& a_SKIPMASK
)
1363 /* ASCII C0: 0..1F, 7F <- @.._ (+ a-z -> A-Z), ? */
1364 c
= upperconv(c2
) ^ 0x40;
1365 if((ui8_t
)c
> 0x1F && c
!= 0x7F){
1366 if(flags
& n_SHEXP_PARSE_LOG
)
1367 n_err(_("Invalid \\c notation: %.*s: %.*s\n"),
1368 (int)input
->l
, input
->s
,
1369 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1370 rv
|= n_SHEXP_STATE_ERR_CONTROL
;
1372 /* As an implementation-defined extension, support \c@
1373 * EQ printf(1) alike \c */
1380 /* Octal sequence: 1 to 3 octal bytes */
1382 /* As an extension (dependent on where you look, echo(1), or
1383 * awk(1)/tr(1) etc.), allow leading "0" octal indicator */
1384 if(il
> 0 && (c
= *ib
) >= '0' && c
<= '7'){
1389 case '1': case '2': case '3':
1390 case '4': case '5': case '6': case '7':
1392 if(il
> 0 && (c
= *ib
) >= '0' && c
<= '7'){
1393 c2
= (c2
<< 3) | (c
- '0');
1396 if(il
> 0 && (c
= *ib
) >= '0' && c
<= '7'){
1397 if(!(state
& a_SKIPMASK
) && (ui8_t
)c2
> 0x1F){
1398 rv
|= n_SHEXP_STATE_ERR_NUMBER
;
1400 if(flags
& n_SHEXP_PARSE_LOG
)
1401 n_err(_("\\0 argument exceeds a byte: %.*s: %.*s\n"),
1402 (int)input
->l
, input
->s
,
1403 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1404 /* Write unchanged */
1406 rv
|= n_SHEXP_STATE_OUTPUT
;
1407 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1408 store
= n_string_push_buf(store
, ib_save
,
1409 PTR2SIZE(ib
- ib_save
));
1412 c2
= (c2
<< 3) | (c
-= '0');
1415 if(state
& a_SKIPMASK
)
1417 if((c
= c2
) == '\0'){
1423 /* ISO 10646 / Unicode sequence, 8 or 4 hexadecimal bytes */
1432 goto j_dollar_ungetc
;
1436 /* Hexadecimal sequence, 1 or 2 hexadecimal bytes */
1440 goto j_dollar_ungetc
;
1444 static ui8_t
const hexatoi
[] = { /* XXX uses ASCII */
1445 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
1450 for(no
= j
= 0; i
-- > 0; --il
, ++ib
, ++j
){
1454 no
+= hexatoi
[(ui8_t
)((c
) - ((c
) <= '9' ? 48
1455 : ((c
) <= 'F' ? 55 : 87)))];
1457 if(state
& a_SKIPMASK
)
1459 c2
= (c2
== 'U' || c2
== 'u') ? 'u' : 'x';
1460 if(flags
& n_SHEXP_PARSE_LOG
)
1461 n_err(_("Invalid \\%c notation: %.*s: %.*s\n"),
1462 c2
, (int)input
->l
, input
->s
,
1463 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1464 rv
|= n_SHEXP_STATE_ERR_NUMBER
;
1470 /* Unicode massage */
1471 if((c2
!= 'U' && c2
!= 'u') || n_uasciichar(no
)){
1472 if((c
= (char)no
) == '\0')
1476 else if(!(state
& a_SKIPMASK
)){
1477 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1478 store
= n_string_reserve(store
, n_MAX(j
, 4));
1480 if(no
> 0x10FFFF){ /* XXX magic; CText */
1481 if(flags
& n_SHEXP_PARSE_LOG
)
1482 n_err(_("\\U argument exceeds 0x10FFFF: %.*s: "
1484 (int)input
->l
, input
->s
,
1485 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1486 rv
|= n_SHEXP_STATE_ERR_NUMBER
;
1487 /* But normalize the output anyway */
1491 j
= n_utf32_to_utf8(no
, utf
);
1493 if(n_psonce
& n_PSO_UNICODE
){
1494 rv
|= n_SHEXP_STATE_OUTPUT
| n_SHEXP_STATE_UNICODE
;
1495 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1496 store
= n_string_push_buf(store
, utf
, j
);
1503 icp
= n_iconv_onetime_cp(n_ICONV_NONE
,
1506 rv
|= n_SHEXP_STATE_OUTPUT
;
1507 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1508 store
= n_string_push_cp(store
, icp
);
1513 if(!(flags
& n_SHEXP_PARSE_DRYRUN
)) Jerr_uni_norm
:{
1516 rv
|= n_SHEXP_STATE_OUTPUT
|
1517 n_SHEXP_STATE_ERR_UNICODE
;
1518 i
= snprintf(itoa
, sizeof itoa
, "\\%c%0*X",
1519 (no
> 0xFFFFu
? 'U' : 'u'),
1520 (int)(no
> 0xFFFFu
? 8 : 4), (ui32_t
)no
);
1521 store
= n_string_push_buf(store
, itoa
, i
);
1525 if(state
& a_SKIPMASK
)
1530 /* Extension: \$ can be used to expand a variable.
1531 * B(ug|ad) effect: if conversion fails, not written "as-is" */
1534 goto j_dollar_ungetc
;
1539 /* Follow bash(1) behaviour, print sequence unchanged */
1544 }else if(c
== '$' && quotec
== '"' && il
> 0) J_var_expand
:{
1545 state
&= ~a_VARSUBST_MASK
;
1549 /* Scan variable name */
1550 if(!(state
& a_BRACE
) || il
> 1){
1551 char const *cp
, *vp
;
1557 state
&= ~a_EXPLODE
;
1559 for(i
= 0; il
> 0; --il
, ++ib
, ++i
){
1560 /* We have some special cases regarding special parameters,
1561 * so ensure these don't cause failure. This code has
1562 * counterparts in code that manages internal variables! */
1564 if(!a_SHEXP_ISVARC(c
)){
1566 /* Simply skip over multiplexer */
1569 if(c
== '*' || c
== '@' || c
== '#' || c
== '?' ||
1580 }else if(a_SHEXP_ISVARC_BAD1ST(c
)){
1584 state
|= a_NONDIGIT
;
1587 /* In skip mode, be easy and.. skip over */
1588 if(state
& a_SKIPMASK
){
1589 if((state
& a_BRACE
) && il
> 0 && *ib
== '}')
1594 /* Handle the scan error cases */
1595 if((state
& (a_DIGIT1
| a_NONDIGIT
)) == (a_DIGIT1
| a_NONDIGIT
)){
1596 if(state
& a_BRACE
){
1597 if(il
> 0 && *ib
== '}')
1600 rv
|= n_SHEXP_STATE_ERR_GROUPOPEN
;
1602 if(flags
& n_SHEXP_PARSE_LOG
)
1603 n_err(_("Invalid identifier for ${}: %.*s: %.*s\n"),
1604 (int)input
->l
, input
->s
,
1605 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1606 rv
|= n_SHEXP_STATE_ERR_IDENTIFIER
;
1609 if(state
& a_BRACE
){
1610 if(il
== 0 || *ib
!= '}'){
1611 if(flags
& n_SHEXP_PARSE_LOG
)
1612 n_err(_("No closing brace for ${}: %.*s: %.*s\n"),
1613 (int)input
->l
, input
->s
,
1614 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1615 rv
|= n_SHEXP_STATE_ERR_GROUPOPEN
;
1621 if(flags
& n_SHEXP_PARSE_LOG
)
1622 n_err(_("Bad substitution for ${}: %.*s: %.*s\n"),
1623 (int)input
->l
, input
->s
,
1624 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1625 rv
|= n_SHEXP_STATE_ERR_BADSUB
;
1629 /* Simply write dollar as-is? */
1632 if(state
& a_BRACE
){
1633 if(il
== 0 || *ib
!= '}'){
1634 if(flags
& n_SHEXP_PARSE_LOG
)
1635 n_err(_("No closing brace for ${}: %.*s: %.*s\n"),
1636 (int)input
->l
, input
->s
,
1637 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1638 rv
|= n_SHEXP_STATE_ERR_GROUPOPEN
;
1644 if(flags
& n_SHEXP_PARSE_LOG
)
1645 n_err(_("Bad substitution for ${}: %.*s: %.*s\n"),
1646 (int)input
->l
, input
->s
,
1647 (int)PTR2SIZE(ib
- ib_save
), ib_save
);
1648 rv
|= n_SHEXP_STATE_ERR_BADSUB
;
1653 if(flags
& n_SHEXP_PARSE_DRYRUN
)
1656 /* We may shall explode "${@}" to a series of successive,
1657 * properly quoted tokens (instead). The first exploded
1658 * cookie will join with the current token */
1659 if(n_UNLIKELY(state
& a_EXPLODE
) &&
1660 !(flags
& n_SHEXP_PARSE_DRYRUN
) && cookie
!= NULL
){
1661 if(n_var_vexplode(cookie
))
1663 /* On the other hand, if $@ expands to nothing and is the
1664 * sole content of this quote then act like the shell does
1665 * and throw away the entire atxplode construct */
1666 else if(!(rv
& n_SHEXP_STATE_OUTPUT
) &&
1667 il
== 1 && *ib
== '"' &&
1668 ib_save
== &input
->s
[1] && ib_save
[-1] == '"')
1672 input
->s
= n_UNCONST(ib
);
1674 goto jrestart_empty
;
1677 /* Check getenv(3) shall no internal variable exist!
1678 * XXX We have some common idioms, avoid memory for them
1679 * XXX Even better would be var_vlook_buf()! */
1682 case '?': vp
= n_qm
; break;
1683 case '!': vp
= n_em
; break;
1684 case '*': vp
= n_star
; break;
1685 case '@': vp
= n_at
; break;
1686 case '#': vp
= n_ns
; break;
1687 default: goto j_var_look_buf
;
1691 vp
= savestrbuf(vp
, i
);
1693 if((cp
= n_var_vlook(vp
, TRU1
)) != NULL
){
1694 rv
|= n_SHEXP_STATE_OUTPUT
;
1695 store
= n_string_push_cp(store
, cp
);
1696 for(; (c
= *cp
) != '\0'; ++cp
)
1698 rv
|= n_SHEXP_STATE_CONTROL
;
1705 }else if(c
== '`' && quotec
== '"' && il
> 0){ /* TODO shell command */
1710 if(!(state
& a_SKIPMASK
)){
1711 rv
|= n_SHEXP_STATE_OUTPUT
;
1713 rv
|= n_SHEXP_STATE_CONTROL
;
1714 if(!(flags
& n_SHEXP_PARSE_DRYRUN
))
1715 store
= n_string_push_c(store
, c
);
1719 if(quotec
!= '\0' && !(flags
& n_SHEXP_PARSE_QUOTE_AUTO_CLOSE
)){
1720 if(flags
& n_SHEXP_PARSE_LOG
)
1721 n_err(_("No closing quote: %.*s\n"), (int)input
->l
, input
->s
);
1722 rv
|= n_SHEXP_STATE_ERR_QUOTEOPEN
;
1726 assert(!(state
& a_COOKIE
));
1727 if((flags
& n_SHEXP_PARSE_DRYRUN
) && store
!= NULL
){
1728 store
= n_string_push_buf(store
, input
->s
, PTR2SIZE(ib
- input
->s
));
1729 rv
|= n_SHEXP_STATE_OUTPUT
;
1732 if(state
& a_CONSUME
){
1733 input
->s
= n_UNCONST(&ib
[il
]);
1736 if(flags
& n_SHEXP_PARSE_TRIM_SPACE
){
1737 for(; il
> 0; ++ib
, --il
){
1738 if(!blankspacechar(*ib
))
1740 rv
|= n_SHEXP_STATE_WS_TRAIL
;
1744 if(flags
& n_SHEXP_PARSE_TRIM_IFSSPACE
){
1745 for(; il
> 0; ++ib
, --il
){
1746 if(strchr(ifs_ws
, *ib
) == NULL
)
1748 rv
|= n_SHEXP_STATE_WS_TRAIL
;
1753 input
->s
= n_UNCONST(ib
);
1756 if(!(rv
& n_SHEXP_STATE_STOP
)){
1757 if(!(rv
& (n_SHEXP_STATE_OUTPUT
| n_SHEXP_STATE_META_MASK
)) &&
1758 (flags
& n_SHEXP_PARSE_IGNORE_EMPTY
) && il
> 0)
1759 goto jrestart_empty
;
1760 if(/*!(rv & n_SHEXP_STATE_OUTPUT) &&*/ il
== 0)
1761 rv
|= n_SHEXP_STATE_STOP
;
1764 if((state
& a_SKIPT
) && !(rv
& n_SHEXP_STATE_STOP
) &&
1765 (flags
& n_SHEXP_PARSE_META_MASK
))
1768 assert((rv
& n_SHEXP_STATE_OUTPUT
) || !(rv
& n_SHEXP_STATE_UNICODE
));
1769 assert((rv
& n_SHEXP_STATE_OUTPUT
) || !(rv
& n_SHEXP_STATE_CONTROL
));
1775 n_shexp_parse_token_cp(enum n_shexp_parse_flags flags
, char const **cp
){
1777 struct n_string sou
, *soup
;
1779 enum n_shexp_state shs
;
1784 input
.s
= n_UNCONST(*cp
);
1786 soup
= n_string_creat_auto(&sou
);
1788 shs
= n_shexp_parse_token(flags
, soup
, &input
, NULL
);
1789 if(shs
& n_SHEXP_STATE_ERR_MASK
){
1790 soup
= n_string_assign_cp(soup
, *cp
);
1795 rv
= n_string_cp(soup
);
1796 /*n_string_gut(n_string_drop_ownership(soup));*/
1801 FL
struct n_string
*
1802 n_shexp_quote(struct n_string
*store
, struct str
const *input
, bool_t rndtrip
){
1803 struct a_shexp_quote_lvl sql
;
1804 struct a_shexp_quote_ctx sqc
;
1807 assert(store
!= NULL
);
1808 assert(input
!= NULL
);
1809 assert(input
->l
== 0 || input
->s
!= NULL
);
1811 memset(&sqc
, 0, sizeof sqc
);
1812 sqc
.sqc_store
= store
;
1813 sqc
.sqc_input
.s
= input
->s
;
1814 if((sqc
.sqc_input
.l
= input
->l
) == UIZ_MAX
)
1815 sqc
.sqc_input
.l
= strlen(input
->s
);
1816 sqc
.sqc_flags
= rndtrip
? a_SHEXP_QUOTE_ROUNDTRIP
: a_SHEXP_QUOTE_NONE
;
1818 if(sqc
.sqc_input
.l
== 0)
1819 store
= n_string_push_buf(store
, "''", sizeof("''") -1);
1821 memset(&sql
, 0, sizeof sql
);
1822 sql
.sql_dat
= sqc
.sqc_input
;
1823 sql
.sql_flags
= sqc
.sqc_flags
;
1824 a_shexp__quote(&sqc
, &sql
);
1831 n_shexp_quote_cp(char const *cp
, bool_t rndtrip
){
1832 struct n_string store
;
1839 input
.s
= n_UNCONST(cp
);
1841 rv
= n_string_cp(n_shexp_quote(n_string_creat_auto(&store
), &input
,
1843 n_string_gut(n_string_drop_ownership(&store
));
1849 n_shexp_is_valid_varname(char const *name
){
1856 for(lc
= '\0'; (c
= *name
++) != '\0'; lc
= c
)
1857 if(!a_SHEXP_ISVARC(c
))
1859 else if(lc
== '\0' && a_SHEXP_ISVARC_BAD1ST(c
))
1861 if(a_SHEXP_ISVARC_BADNST(lc
))
1871 c_shcodec(void *vp
){
1873 struct n_string sou_b
, *soup
;
1877 char const **argv
, *varname
, *act
, *cp
;
1879 soup
= n_string_creat_auto(&sou_b
);
1881 varname
= (n_pstate
& n_PS_ARGMOD_VPUT
) ? *argv
++ : NULL
;
1884 for(cp
= act
; *cp
!= '\0' && !blankspacechar(*cp
); ++cp
)
1886 if((norndtrip
= (*act
== '+')))
1890 alen
= PTR2SIZE(cp
- act
);
1894 in
.l
= strlen(in
.s
= n_UNCONST(cp
));
1897 if(is_ascncaseprefix(act
, "encode", alen
))
1898 soup
= n_shexp_quote(soup
, &in
, !norndtrip
);
1899 else if(!norndtrip
&& is_ascncaseprefix(act
, "decode", alen
)){
1901 enum n_shexp_state shs
;
1903 shs
= n_shexp_parse_token((n_SHEXP_PARSE_LOG
|
1904 n_SHEXP_PARSE_IGNORE_EMPTY
), soup
, &in
, NULL
);
1905 if(shs
& n_SHEXP_STATE_ERR_MASK
){
1906 soup
= n_string_assign_cp(soup
, cp
);
1907 nerrn
= n_ERR_CANCELED
;
1911 if(shs
& n_SHEXP_STATE_STOP
)
1917 if(varname
!= NULL
){
1918 cp
= n_string_cp(soup
);
1919 if(!n_var_vset(varname
, (uintptr_t)cp
)){
1920 nerrn
= n_ERR_NOTSUP
;
1926 in
.s
= n_string_cp(soup
);
1928 makeprint(&in
, &out
);
1929 if(fprintf(n_stdout
, "%s\n", out
.s
) < 0){
1937 n_pstate_err_no
= nerrn
;
1939 return (vp
!= NULL
? 0 : 1);
1941 n_err(_("Synopsis: shcodec: <[+]e[ncode]|d[ecode]> <rest-of-line>\n"));
1942 nerrn
= n_ERR_INVAL
;