/* parse a decimal number: returns 0 if string isn't a number, 1 otherwise */
int
getn(const char *s, int *ai)
{
int i, c, rv = 0;
bool neg = false;
do {
c = *s++;
} while (ksh_isspace(c));
if (c == '-') {
neg = true;
c = *s++;
} else if (c == '+')
c = *s++;
*ai = i = 0;
do {
if (!ksh_isdigit(c))
goto getn_out;
i *= 10;
if (i < *ai)
/* overflow */
goto getn_out;
i += c - '0';
*ai = i;
} while ((c = *s++));
rv = 1;
getn_out:
if (neg)
*ai = -*ai;
return (rv);
}
/*
* Called by iosetup() (deals with 2>&4, etc.), c_read, c_print to turn
* a string (the X in 2>&X, read -uX, print -uX) into a file descriptor.
*/
int
check_fd(const char *name, int mode, const char **emsgp)
{
int fd = 0, fl;
if (name[0] == 'p' && !name[1])
return (coproc_getfd(mode, emsgp));
while (ksh_isdigit(*name)) {
fd = (fd * 10) + *name - '0';
if (fd >= FDBASE) {
if (emsgp)
*emsgp = "file descriptor too large";
return (-1);
}
++name;
}
if (*name) {
if (emsgp)
*emsgp = "illegal file descriptor name";
return (-1);
}
if ((fl = fcntl(fd, F_GETFL, 0)) < 0) {
if (emsgp)
*emsgp = "bad file descriptor";
return (-1);
}
fl &= O_ACCMODE;
/*
* X_OK is a kludge to disable this check for dups (x<&1):
* historical shells never did this check (XXX don't know what
* POSIX has to say).
*/
if (!(mode & X_OK) && fl != O_RDWR && (
((mode & R_OK) && fl != O_RDONLY) ||
((mode & W_OK) && fl != O_WRONLY))) {
if (emsgp)
*emsgp = (fl == O_WRONLY) ?
"fd not open for reading" :
"fd not open for writing";
return (-1);
}
return (fd);
}
/* parse a decimal number: returns 0 if string isn't a number, 1 otherwise */
int
getn(const char *s, int *ai)
{
char c;
mksh_ari_u num;
bool neg = false;
num.u = 0;
do {
c = *s++;
} while (ksh_isspace(c));
switch (c) {
case '-':
neg = true;
/* FALLTHROUGH */
case '+':
c = *s++;
break;
}
do {
if (!ksh_isdigit(c))
/* not numeric */
return (0);
if (num.u > 214748364U)
/* overflow on multiplication */
return (0);
num.u = num.u * 10U + (unsigned int)ksh_numdig(c);
/* now: num.u <= 2147483649U */
} while ((c = *s++));
if (num.u > (neg ? 2147483648U : 2147483647U))
/* overflow for signed 32-bit int */
return (0);
if (neg)
num.u = -num.u;
*ai = num.i;
return (1);
}
/* getopt() used for shell built-in commands, the getopts command, and
* command line options.
* A leading ':' in options means don't print errors, instead return '?'
* or ':' and set go->optarg to the offending option character.
* If GF_ERROR is set (and option doesn't start with :), errors result in
* a call to bi_errorf().
*
* Non-standard features:
* - ';' is like ':' in options, except the argument is optional
* (if it isn't present, optarg is set to 0).
* Used for 'set -o'.
* - ',' is like ':' in options, except the argument always immediately
* follows the option character (optarg is set to the null string if
* the option is missing).
* Used for 'read -u2', 'print -u2' and fc -40.
* - '#' is like ':' in options, expect that the argument is optional
* and must start with a digit. If the argument doesn't start with a
* digit, it is assumed to be missing and normal option processing
* continues (optarg is set to 0 if the option is missing).
* Used for 'typeset -LZ4'.
* - accepts +c as well as -c IF the GF_PLUSOPT flag is present. If an
* option starting with + is accepted, the GI_PLUS flag will be set
* in go->info.
*/
int
ksh_getopt(const char **argv, Getopt *go, const char *optionsp)
{
char c;
const char *o;
if (go->p == 0 || (c = argv[go->optind - 1][go->p]) == '\0') {
const char *arg = argv[go->optind], flag = arg ? *arg : '\0';
go->p = 1;
if (flag == '-' && arg[1] == '-' && arg[2] == '\0') {
go->optind++;
go->p = 0;
go->info |= GI_MINUSMINUS;
return (-1);
}
if (arg == NULL ||
((flag != '-' ) && /* neither a - nor a + (if + allowed) */
(!(go->flags & GF_PLUSOPT) || flag != '+')) ||
(c = arg[1]) == '\0') {
go->p = 0;
return (-1);
}
go->optind++;
go->info &= ~(GI_MINUS|GI_PLUS);
go->info |= flag == '-' ? GI_MINUS : GI_PLUS;
}
go->p++;
if (c == '?' || c == ':' || c == ';' || c == ',' || c == '#' ||
!(o = cstrchr(optionsp, c))) {
if (optionsp[0] == ':') {
go->buf[0] = c;
go->optarg = go->buf;
} else {
warningf(true, "%s%s-%c: unknown option",
(go->flags & GF_NONAME) ? "" : argv[0],
(go->flags & GF_NONAME) ? "" : ": ", c);
if (go->flags & GF_ERROR)
bi_errorfz();
}
return ('?');
}
/* : means argument must be present, may be part of option argument
* or the next argument
* ; same as : but argument may be missing
* , means argument is part of option argument, and may be null.
*/
if (*++o == ':' || *o == ';') {
if (argv[go->optind - 1][go->p])
go->optarg = argv[go->optind - 1] + go->p;
else if (argv[go->optind])
go->optarg = argv[go->optind++];
else if (*o == ';')
go->optarg = NULL;
else {
if (optionsp[0] == ':') {
go->buf[0] = c;
go->optarg = go->buf;
return (':');
}
warningf(true, "%s%s-'%c' requires argument",
(go->flags & GF_NONAME) ? "" : argv[0],
(go->flags & GF_NONAME) ? "" : ": ", c);
if (go->flags & GF_ERROR)
bi_errorfz();
return ('?');
}
go->p = 0;
} else if (*o == ',') {
/* argument is attached to option character, even if null */
go->optarg = argv[go->optind - 1] + go->p;
go->p = 0;
} else if (*o == '#') {
/* argument is optional and may be attached or unattached
* but must start with a digit. optarg is set to 0 if the
* argument is missing.
*/
if (argv[go->optind - 1][go->p]) {
if (ksh_isdigit(argv[go->optind - 1][go->p])) {
go->optarg = argv[go->optind - 1] + go->p;
go->p = 0;
} else
go->optarg = NULL;
} else {
if (argv[go->optind] && ksh_isdigit(argv[go->optind][0])) {
go->optarg = argv[go->optind++];
go->p = 0;
} else
go->optarg = NULL;
}
}
return (c);
}
/*
* lookup variable (according to (set&LOCAL)), set its attributes
* (INTEGER, RDONLY, EXPORT, TRACE, LJUST, RJUST, ZEROFIL, LCASEV,
* UCASEV_AL), and optionally set its value if an assignment.
*/
struct tbl *
typeset(const char *var, uint32_t set, uint32_t clr, int field, int base)
{
struct tbl *vp;
struct tbl *vpbase, *t;
char *tvar;
const char *val;
size_t len;
bool vappend = false;
enum namerefflag new_refflag = SRF_NOP;
if ((set & (ARRAY | ASSOC)) == ASSOC) {
new_refflag = SRF_ENABLE;
set &= ~(ARRAY | ASSOC);
}
if ((clr & (ARRAY | ASSOC)) == ASSOC) {
new_refflag = SRF_DISABLE;
clr &= ~(ARRAY | ASSOC);
}
/* check for valid variable name, search for value */
val = skip_varname(var, false);
if (val == var) {
/* no variable name given */
return (NULL);
}
if (*val == '[') {
if (new_refflag != SRF_NOP)
errorf("%s: %s", var,
"reference variable can't be an array");
len = array_ref_len(val);
if (len == 0)
return (NULL);
/*
* IMPORT is only used when the shell starts up and is
* setting up its environment. Allow only simple array
* references at this time since parameter/command
* substitution is performed on the [expression] which
* would be a major security hole.
*/
if (set & IMPORT) {
size_t i;
for (i = 1; i < len - 1; i++)
if (!ksh_isdigit(val[i]))
return (NULL);
}
val += len;
}
if (val[0] == '=') {
strndupx(tvar, var, val - var, ATEMP);
++val;
} else if (set & IMPORT) {
/* environment invalid variable name or no assignment */
return (NULL);
} else if (val[0] == '+' && val[1] == '=') {
strndupx(tvar, var, val - var, ATEMP);
val += 2;
vappend = true;
} else if (val[0] != '\0') {
/* other invalid variable names (not from environment) */
return (NULL);
} else {
/* just varname with no value part nor equals sign */
strdupx(tvar, var, ATEMP);
val = NULL;
/* handle foo[*] => foo (whole array) mapping for R39b */
len = strlen(tvar);
if (len > 3 && tvar[len - 3] == '[' && tvar[len - 2] == '*' &&
tvar[len - 1] == ']')
tvar[len - 3] = '\0';
}
if (new_refflag == SRF_ENABLE) {
const char *qval, *ccp;
/* bail out on 'nameref foo+=bar' */
if (vappend)
errorf("appending not allowed for nameref");
/* find value if variable already exists */
if ((qval = val) == NULL) {
varsearch(e->loc, &vp, tvar, hash(tvar));
if (vp == NULL)
goto nameref_empty;
qval = str_val(vp);
}
/* check target value for being a valid variable name */
ccp = skip_varname(qval, false);
if (ccp == qval) {
if (ksh_isdigit(qval[0])) {
int c;
if (getn(qval, &c))
goto nameref_rhs_checked;
} else if (qval[1] == '\0') switch (qval[0]) {
case '$':
case '!':
case '?':
case '#':
case '-':
goto nameref_rhs_checked;
}
nameref_empty:
errorf("%s: %s", var, "empty nameref target");
}
len = (*ccp == '[') ? array_ref_len(ccp) : 0;
if (ccp[len]) {
/*
* works for cases "no array", "valid array with
* junk after it" and "invalid array"; in the
* latter case, len is also 0 and points to '['
*/
errorf("%s: %s", qval,
"nameref target not a valid parameter name");
}
nameref_rhs_checked:
/* prevent nameref loops */
while (qval) {
if (!strcmp(qval, tvar))
errorf("%s: %s", qval,
"expression recurses on parameter");
varsearch(e->loc, &vp, qval, hash(qval));
qval = NULL;
if (vp && ((vp->flag & (ARRAY | ASSOC)) == ASSOC))
qval = str_val(vp);
}
}
/* prevent typeset from creating a local PATH/ENV/SHELL */
if (Flag(FRESTRICTED) && (strcmp(tvar, "PATH") == 0 ||
strcmp(tvar, "ENV") == 0 || strcmp(tvar, "SHELL") == 0))
errorf("%s: %s", tvar, "restricted");
innermost_refflag = new_refflag;
vp = (set & LOCAL) ? local(tvar, tobool(set & LOCAL_COPY)) :
global(tvar);
if (new_refflag == SRF_DISABLE && (vp->flag & (ARRAY|ASSOC)) == ASSOC)
vp->flag &= ~ASSOC;
else if (new_refflag == SRF_ENABLE) {
if (vp->flag & ARRAY) {
struct tbl *a, *tmp;
/* free up entire array */
for (a = vp->u.array; a; ) {
tmp = a;
a = a->u.array;
if (tmp->flag & ALLOC)
afree(tmp->val.s, tmp->areap);
afree(tmp, tmp->areap);
}
vp->u.array = NULL;
vp->flag &= ~ARRAY;
}
vp->flag |= ASSOC;
}
set &= ~(LOCAL|LOCAL_COPY);
vpbase = (vp->flag & ARRAY) ? global(arrayname(tvar)) : vp;
/*
* only allow export flag to be set; AT&T ksh allows any
* attribute to be changed which means it can be truncated or
* modified (-L/-R/-Z/-i)
*/
if ((vpbase->flag & RDONLY) &&
(val || clr || (set & ~EXPORT)))
/* XXX check calls - is error here ok by POSIX? */
errorfx(2, "read-only: %s", tvar);
afree(tvar, ATEMP);
/* most calls are with set/clr == 0 */
if (set | clr) {
bool ok = true;
/*
* XXX if x[0] isn't set, there will be problems: need
* to have one copy of attributes for arrays...
*/
for (t = vpbase; t; t = t->u.array) {
bool fake_assign;
char *s = NULL;
char *free_me = NULL;
fake_assign = (t->flag & ISSET) && (!val || t != vp) &&
((set & (UCASEV_AL|LCASEV|LJUST|RJUST|ZEROFIL)) ||
((t->flag & INTEGER) && (clr & INTEGER)) ||
(!(t->flag & INTEGER) && (set & INTEGER)));
if (fake_assign) {
if (t->flag & INTEGER) {
s = str_val(t);
free_me = NULL;
} else {
s = t->val.s + t->type;
free_me = (t->flag & ALLOC) ? t->val.s :
NULL;
}
t->flag &= ~ALLOC;
}
if (!(t->flag & INTEGER) && (set & INTEGER)) {
t->type = 0;
t->flag &= ~ALLOC;
}
t->flag = (t->flag | set) & ~clr;
/*
* Don't change base if assignment is to be
* done, in case assignment fails.
*/
if ((set & INTEGER) && base > 0 && (!val || t != vp))
t->type = base;
if (set & (LJUST|RJUST|ZEROFIL))
t->u2.field = field;
if (fake_assign) {
if (!setstr(t, s, KSH_RETURN_ERROR)) {
/*
* Somewhat arbitrary action
* here: zap contents of
* variable, but keep the flag
* settings.
*/
ok = false;
if (t->flag & INTEGER)
t->flag &= ~ISSET;
else {
if (t->flag & ALLOC)
afree(t->val.s, t->areap);
t->flag &= ~(ISSET|ALLOC);
t->type = 0;
}
}
if (free_me)
afree(free_me, t->areap);
}
}
if (!ok)
errorfz();
}
if (val != NULL) {
char *tval;
if (vappend) {
tval = shf_smprintf("%s%s", str_val(vp), val);
val = tval;
} else
tval = NULL;
if (vp->flag&INTEGER) {
/* do not zero base before assignment */
setstr(vp, val, KSH_UNWIND_ERROR | 0x4);
/* done after assignment to override default */
if (base > 0)
vp->type = base;
} else
/* setstr can't fail (readonly check already done) */
setstr(vp, val, KSH_RETURN_ERROR | 0x4);
if (tval != NULL)
afree(tval, ATEMP);
}
/* only x[0] is ever exported, so use vpbase */
if ((vpbase->flag&EXPORT) && !(vpbase->flag&INTEGER) &&
vpbase->type == 0)
exportprep(vpbase, (vpbase->flag&ISSET) ? vpbase->val.s : null);
return (vp);
}
static int
getint(struct tbl *vp, mksh_ari_u *nump, bool arith)
{
mksh_uari_t c, num = 0, base = 10;
const char *s;
bool have_base = false, neg = false;
if (vp->flag & SPECIAL)
getspec(vp);
/* XXX is it possible for ISSET to be set and val.s to be NULL? */
if (!(vp->flag & ISSET) || (!(vp->flag & INTEGER) && vp->val.s == NULL))
return (-1);
if (vp->flag & INTEGER) {
nump->i = vp->val.i;
return (vp->type);
}
s = vp->val.s + vp->type;
do {
c = (unsigned char)*s++;
} while (ksh_isspace(c));
switch (c) {
case '-':
neg = true;
/* FALLTHROUGH */
case '+':
c = (unsigned char)*s++;
break;
}
if (c == '0' && arith) {
if (ksh_eq(s[0], 'X', 'x')) {
/* interpret as hexadecimal */
base = 16;
++s;
goto getint_c_style_base;
} else if (Flag(FPOSIX) && ksh_isdigit(s[0]) &&
!(vp->flag & ZEROFIL)) {
/* interpret as octal (deprecated) */
base = 8;
getint_c_style_base:
have_base = true;
c = (unsigned char)*s++;
}
}
do {
if (c == '#') {
/* ksh-style base determination */
if (have_base || num < 1)
return (-1);
if ((base = num) == 1) {
/* mksh-specific extension */
unsigned int wc;
if (!UTFMODE)
wc = *(const unsigned char *)s;
else if (utf_mbtowc(&wc, s) == (size_t)-1)
/* OPTU-8 -> OPTU-16 */
/*
* (with a twist: 1#\uEF80 converts
* the same as 1#\x80 does, thus is
* not round-tripping correctly XXX)
*/
wc = 0xEF00 + *(const unsigned char *)s;
nump->u = (mksh_uari_t)wc;
return (1);
} else if (base > 36)
base = 10;
num = 0;
have_base = true;
continue;
}
if (ksh_isdigit(c))
c = ksh_numdig(c);
else if (ksh_isupper(c))
c = ksh_numuc(c) + 10;
else if (ksh_islower(c))
c = ksh_numlc(c) + 10;
else
return (-1);
if (c >= base)
return (-1);
/* handle overflow as truncation */
num = num * base + c;
} while ((c = (unsigned char)*s++));
if (neg)
num = -num;
nump->u = num;
return (base);
}
/*
* Search for variable, if not found create globally.
*/
struct tbl *
global(const char *n)
{
struct block *l = e->loc;
struct tbl *vp;
int c;
bool array;
uint32_t h, val;
/*
* check to see if this is an array;
* dereference namerefs; must come first
*/
n = array_index_calc(n, &array, &val);
h = hash(n);
c = (unsigned char)n[0];
if (!ksh_isalphx(c)) {
if (array)
errorf("bad substitution");
vp = &vtemp;
vp->flag = DEFINED;
vp->type = 0;
vp->areap = ATEMP;
*vp->name = c;
if (ksh_isdigit(c)) {
if (getn(n, &c) && (c <= l->argc))
/* setstr can't fail here */
setstr(vp, l->argv[c], KSH_RETURN_ERROR);
vp->flag |= RDONLY;
return (vp);
}
vp->flag |= RDONLY;
if (n[1] != '\0')
return (vp);
vp->flag |= ISSET|INTEGER;
switch (c) {
case '$':
vp->val.i = kshpid;
break;
case '!':
/* if no job, expand to nothing */
if ((vp->val.i = j_async()) == 0)
vp->flag &= ~(ISSET|INTEGER);
break;
case '?':
vp->val.i = exstat & 0xFF;
break;
case '#':
vp->val.i = l->argc;
break;
case '-':
vp->flag &= ~INTEGER;
vp->val.s = getoptions();
break;
default:
vp->flag &= ~(ISSET|INTEGER);
}
return (vp);
}
l = varsearch(e->loc, &vp, n, h);
if (vp != NULL)
return (array ? arraysearch(vp, val) : vp);
vp = ktenter(&l->vars, n, h);
if (array)
vp = arraysearch(vp, val);
vp->flag |= DEFINED;
if (special(n))
vp->flag |= SPECIAL;
return (vp);
}
int
yylex(int cf)
{
Lex_state states[STATE_BSIZE], *statep, *s2, *base;
State_info state_info;
int c, c2, state;
size_t cz;
XString ws; /* expandable output word */
char *wp; /* output word pointer */
char *sp, *dp;
Again:
states[0].type = SINVALID;
states[0].ls_base = NULL;
statep = &states[1];
state_info.base = states;
state_info.end = &state_info.base[STATE_BSIZE];
Xinit(ws, wp, 64, ATEMP);
backslash_skip = 0;
ignore_backslash_newline = 0;
if (cf & ONEWORD)
state = SWORD;
else if (cf & LETEXPR) {
/* enclose arguments in (double) quotes */
*wp++ = OQUOTE;
state = SLETPAREN;
statep->nparen = 0;
} else {
/* normal lexing */
state = (cf & HEREDELIM) ? SHEREDELIM : SBASE;
while ((c = getsc()) == ' ' || c == '\t')
;
if (c == '#') {
ignore_backslash_newline++;
while ((c = getsc()) != '\0' && c != '\n')
;
ignore_backslash_newline--;
}
ungetsc(c);
}
if (source->flags & SF_ALIAS) {
/* trailing ' ' in alias definition */
source->flags &= ~SF_ALIAS;
/* POSIX: trailing space only counts if parsing simple cmd */
if (!Flag(FPOSIX) || (cf & CMDWORD))
cf |= ALIAS;
}
/* Initial state: one of SWORD SLETPAREN SHEREDELIM SBASE */
statep->type = state;
/* collect non-special or quoted characters to form word */
while (!((c = getsc()) == 0 ||
((state == SBASE || state == SHEREDELIM) && ctype(c, C_LEX1)))) {
if (state == SBASE &&
subshell_nesting_type == /*{*/ '}' &&
c == /*{*/ '}')
/* possibly end ${ :;} */
break;
Xcheck(ws, wp);
switch (state) {
case SADELIM:
if (c == '(')
statep->nparen++;
else if (c == ')')
statep->nparen--;
else if (statep->nparen == 0 && (c == /*{*/ '}' ||
c == (int)statep->ls_adelim.delimiter)) {
*wp++ = ADELIM;
*wp++ = c;
if (c == /*{*/ '}' || --statep->ls_adelim.num == 0)
POP_STATE();
if (c == /*{*/ '}')
POP_STATE();
break;
}
/* FALLTHROUGH */
case SBASE:
if (c == '[' && (cf & (VARASN|ARRAYVAR))) {
/* temporary */
*wp = EOS;
if (is_wdvarname(Xstring(ws, wp), false)) {
char *p, *tmp;
if (arraysub(&tmp)) {
*wp++ = CHAR;
*wp++ = c;
for (p = tmp; *p; ) {
Xcheck(ws, wp);
*wp++ = CHAR;
*wp++ = *p++;
}
afree(tmp, ATEMP);
break;
} else {
Source *s;
s = pushs(SREREAD,
source->areap);
s->start = s->str =
s->u.freeme = tmp;
s->next = source;
source = s;
}
}
*wp++ = CHAR;
*wp++ = c;
break;
}
/* FALLTHROUGH */
Sbase1: /* includes *(...|...) pattern (*+?@!) */
if (c == '*' || c == '@' || c == '+' || c == '?' ||
c == '!') {
c2 = getsc();
if (c2 == '(' /*)*/ ) {
*wp++ = OPAT;
*wp++ = c;
PUSH_STATE(SPATTERN);
break;
}
ungetsc(c2);
}
/* FALLTHROUGH */
Sbase2: /* doesn't include *(...|...) pattern (*+?@!) */
switch (c) {
case '\\':
getsc_qchar:
if ((c = getsc())) {
/* trailing \ is lost */
*wp++ = QCHAR;
*wp++ = c;
}
break;
case '\'':
open_ssquote_unless_heredoc:
if ((cf & HEREDOC))
goto store_char;
*wp++ = OQUOTE;
ignore_backslash_newline++;
PUSH_STATE(SSQUOTE);
break;
case '"':
open_sdquote:
*wp++ = OQUOTE;
PUSH_STATE(SDQUOTE);
break;
case '$':
/*
* processing of dollar sign belongs into
* Subst, except for those which can open
* a string: $'…' and $"…"
*/
subst_dollar_ex:
c = getsc();
switch (c) {
case '"':
goto open_sdquote;
case '\'':
goto open_sequote;
default:
goto SubstS;
}
default:
goto Subst;
}
break;
Subst:
switch (c) {
case '\\':
c = getsc();
switch (c) {
case '"':
if ((cf & HEREDOC))
goto heredocquote;
/* FALLTHROUGH */
case '\\':
case '$': case '`':
store_qchar:
*wp++ = QCHAR;
*wp++ = c;
break;
default:
heredocquote:
Xcheck(ws, wp);
if (c) {
/* trailing \ is lost */
*wp++ = CHAR;
*wp++ = '\\';
*wp++ = CHAR;
*wp++ = c;
}
break;
}
break;
case '$':
c = getsc();
SubstS:
if (c == '(') /*)*/ {
c = getsc();
if (c == '(') /*)*/ {
*wp++ = EXPRSUB;
PUSH_SRETRACE(SASPAREN);
statep->nparen = 2;
*retrace_info->xp++ = '(';
} else {
ungetsc(c);
subst_command:
c = COMSUB;
subst_command2:
sp = yyrecursive(c);
cz = strlen(sp) + 1;
XcheckN(ws, wp, cz);
*wp++ = c;
memcpy(wp, sp, cz);
wp += cz;
}
} else if (c == '{') /*}*/ {
if ((c = getsc()) == '|') {
/*
* non-subenvironment
* value substitution
*/
c = VALSUB;
goto subst_command2;
} else if (ctype(c, C_IFSWS)) {
/*
* non-subenvironment
* "command" substitution
*/
c = FUNSUB;
goto subst_command2;
}
ungetsc(c);
*wp++ = OSUBST;
*wp++ = '{'; /*}*/
wp = get_brace_var(&ws, wp);
c = getsc();
/* allow :# and :% (ksh88 compat) */
if (c == ':') {
*wp++ = CHAR;
*wp++ = c;
c = getsc();
if (c == ':') {
*wp++ = CHAR;
*wp++ = '0';
*wp++ = ADELIM;
*wp++ = ':';
PUSH_STATE(SBRACE);
PUSH_STATE(SADELIM);
statep->ls_adelim.delimiter = ':';
statep->ls_adelim.num = 1;
statep->nparen = 0;
break;
} else if (ksh_isdigit(c) ||
c == '('/*)*/ || c == ' ' ||
/*XXX what else? */
c == '$') {
/* substring subst. */
if (c != ' ') {
*wp++ = CHAR;
*wp++ = ' ';
}
ungetsc(c);
PUSH_STATE(SBRACE);
PUSH_STATE(SADELIM);
statep->ls_adelim.delimiter = ':';
statep->ls_adelim.num = 2;
statep->nparen = 0;
break;
}
} else if (c == '/') {
*wp++ = CHAR;
*wp++ = c;
if ((c = getsc()) == '/') {
*wp++ = ADELIM;
*wp++ = c;
} else
ungetsc(c);
PUSH_STATE(SBRACE);
PUSH_STATE(SADELIM);
statep->ls_adelim.delimiter = '/';
statep->ls_adelim.num = 1;
statep->nparen = 0;
break;
}
/*
* If this is a trim operation,
* treat (,|,) specially in STBRACE.
*/
if (ctype(c, C_SUBOP2)) {
ungetsc(c);
if (Flag(FSH))
PUSH_STATE(STBRACEBOURNE);
else
PUSH_STATE(STBRACEKORN);
} else {
ungetsc(c);
if (state == SDQUOTE ||
state == SQBRACE)
PUSH_STATE(SQBRACE);
else
PUSH_STATE(SBRACE);
}
} else if (ksh_isalphx(c)) {
*wp++ = OSUBST;
*wp++ = 'X';
do {
Xcheck(ws, wp);
*wp++ = c;
c = getsc();
} while (ksh_isalnux(c));
*wp++ = '\0';
*wp++ = CSUBST;
*wp++ = 'X';
ungetsc(c);
} else if (ctype(c, C_VAR1 | C_DIGIT)) {
Xcheck(ws, wp);
*wp++ = OSUBST;
*wp++ = 'X';
*wp++ = c;
*wp++ = '\0';
*wp++ = CSUBST;
*wp++ = 'X';
} else {
*wp++ = CHAR;
*wp++ = '$';
ungetsc(c);
}
break;
case '`':
subst_gravis:
PUSH_STATE(SBQUOTE);
*wp++ = COMSUB;
/*
* We need to know whether we are within double
* quotes, since most shells translate \" to "
* within "…`…\"…`…". This is not done in POSIX
* mode (§2.2.3 Double-Quotes: “The backquote
* shall retain its special meaning introducing
* the other form of command substitution (see
* Command Substitution). The portion of the
* quoted string from the initial backquote and
* the characters up to the next backquote that
* is not preceded by a <backslash>, having
* escape characters removed, defines that
* command whose output replaces "`...`" when
* the word is expanded.”; §2.6.3 Command
* Substitution: “Within the backquoted style
* of command substitution, <backslash> shall
* retain its literal meaning, except when
* followed by: '$', '`', or <backslash>. The
* search for the matching backquote shall be
* satisfied by the first unquoted non-escaped
* backquote; during this search, if a
* non-escaped backquote is encountered[…],
* undefined results occur.”).
*/
statep->ls_bool = false;
if (Flag(FPOSIX))
break;
s2 = statep;
base = state_info.base;
while (/* CONSTCOND */ 1) {
for (; s2 != base; s2--) {
if (s2->type == SDQUOTE) {
statep->ls_bool = true;
break;
}
}
if (s2 != base)
break;
if (!(s2 = s2->ls_base))
break;
base = s2-- - STATE_BSIZE;
}
break;
case QCHAR:
if (cf & LQCHAR) {
*wp++ = QCHAR;
*wp++ = getsc();
break;
}
/* FALLTHROUGH */
default:
store_char:
*wp++ = CHAR;
*wp++ = c;
}
break;
case SEQUOTE:
if (c == '\'') {
POP_STATE();
*wp++ = CQUOTE;
ignore_backslash_newline--;
} else if (c == '\\') {
if ((c2 = unbksl(true, s_get, s_put)) == -1)
c2 = s_get();
if (c2 == 0)
statep->ls_bool = true;
if (!statep->ls_bool) {
char ts[4];
if ((unsigned int)c2 < 0x100) {
*wp++ = QCHAR;
*wp++ = c2;
} else {
cz = utf_wctomb(ts, c2 - 0x100);
ts[cz] = 0;
cz = 0;
do {
*wp++ = QCHAR;
*wp++ = ts[cz];
} while (ts[++cz]);
}
}
} else if (!statep->ls_bool) {
*wp++ = QCHAR;
*wp++ = c;
}
break;
case SSQUOTE:
if (c == '\'') {
POP_STATE();
if ((cf & HEREDOC) || state == SQBRACE)
goto store_char;
*wp++ = CQUOTE;
ignore_backslash_newline--;
} else {
*wp++ = QCHAR;
*wp++ = c;
}
break;
case SDQUOTE:
if (c == '"') {
POP_STATE();
*wp++ = CQUOTE;
} else
goto Subst;
break;
/* $(( ... )) */
case SASPAREN:
if (c == '(')
statep->nparen++;
else if (c == ')') {
statep->nparen--;
if (statep->nparen == 1) {
/* end of EXPRSUB */
POP_SRETRACE();
if ((c2 = getsc()) == /*(*/ ')') {
cz = strlen(sp) - 2;
XcheckN(ws, wp, cz);
memcpy(wp, sp + 1, cz);
wp += cz;
afree(sp, ATEMP);
*wp++ = '\0';
break;
} else {
Source *s;
ungetsc(c2);
/*
* mismatched parenthesis -
* assume we were really
* parsing a $(...) expression
*/
--wp;
s = pushs(SREREAD,
source->areap);
s->start = s->str =
s->u.freeme = sp;
s->next = source;
source = s;
goto subst_command;
}
}
}
/* reuse existing state machine */
goto Sbase2;
case SQBRACE:
if (c == '\\') {
/*
* perform POSIX "quote removal" if the back-
* slash is "special", i.e. same cases as the
* {case '\\':} in Subst: plus closing brace;
* in mksh code "quote removal" on '\c' means
* write QCHAR+c, otherwise CHAR+\+CHAR+c are
* emitted (in heredocquote:)
*/
if ((c = getsc()) == '"' || c == '\\' ||
c == '$' || c == '`' || c == /*{*/'}')
goto store_qchar;
goto heredocquote;
}
goto common_SQBRACE;
case SBRACE:
if (c == '\'')
goto open_ssquote_unless_heredoc;
else if (c == '\\')
goto getsc_qchar;
common_SQBRACE:
if (c == '"')
goto open_sdquote;
else if (c == '$')
goto subst_dollar_ex;
else if (c == '`')
goto subst_gravis;
else if (c != /*{*/ '}')
goto store_char;
POP_STATE();
*wp++ = CSUBST;
*wp++ = /*{*/ '}';
break;
/* Same as SBASE, except (,|,) treated specially */
case STBRACEKORN:
if (c == '|')
*wp++ = SPAT;
else if (c == '(') {
*wp++ = OPAT;
/* simile for @ */
*wp++ = ' ';
PUSH_STATE(SPATTERN);
} else /* FALLTHROUGH */
case STBRACEBOURNE:
if (c == /*{*/ '}') {
POP_STATE();
*wp++ = CSUBST;
*wp++ = /*{*/ '}';
} else
goto Sbase1;
break;
case SBQUOTE:
if (c == '`') {
*wp++ = 0;
POP_STATE();
} else if (c == '\\') {
switch (c = getsc()) {
case 0:
/* trailing \ is lost */
break;
case '$':
case '`':
case '\\':
*wp++ = c;
break;
case '"':
if (statep->ls_bool) {
*wp++ = c;
break;
}
/* FALLTHROUGH */
default:
*wp++ = '\\';
*wp++ = c;
break;
}
} else
*wp++ = c;
break;
/* ONEWORD */
case SWORD:
goto Subst;
/* LETEXPR: (( ... )) */
case SLETPAREN:
if (c == /*(*/ ')') {
if (statep->nparen > 0)
--statep->nparen;
else if ((c2 = getsc()) == /*(*/ ')') {
c = 0;
*wp++ = CQUOTE;
goto Done;
} else {
Source *s;
ungetsc(c2);
/*
* mismatched parenthesis -
* assume we were really
* parsing a (...) expression
*/
*wp = EOS;
sp = Xstring(ws, wp);
dp = wdstrip(sp + 1, WDS_TPUTS);
s = pushs(SREREAD, source->areap);
s->start = s->str = s->u.freeme = dp;
s->next = source;
source = s;
return ('('/*)*/);
}
} else if (c == '(')
/*
* parentheses inside quotes and
* backslashes are lost, but AT&T ksh
* doesn't count them either
*/
++statep->nparen;
goto Sbase2;
/* << or <<- delimiter */
case SHEREDELIM:
/*
* here delimiters need a special case since
* $ and `...` are not to be treated specially
*/
switch (c) {
case '\\':
if ((c = getsc())) {
/* trailing \ is lost */
*wp++ = QCHAR;
*wp++ = c;
}
break;
case '\'':
goto open_ssquote_unless_heredoc;
case '$':
if ((c2 = getsc()) == '\'') {
open_sequote:
*wp++ = OQUOTE;
ignore_backslash_newline++;
PUSH_STATE(SEQUOTE);
statep->ls_bool = false;
break;
} else if (c2 == '"') {
/* FALLTHROUGH */
case '"':
PUSH_SRETRACE(SHEREDQUOTE);
break;
}
ungetsc(c2);
/* FALLTHROUGH */
default:
*wp++ = CHAR;
*wp++ = c;
}
break;
/* " in << or <<- delimiter */
case SHEREDQUOTE:
if (c != '"')
goto Subst;
POP_SRETRACE();
dp = strnul(sp) - 1;
/* remove the trailing double quote */
*dp = '\0';
/* store the quoted string */
*wp++ = OQUOTE;
XcheckN(ws, wp, (dp - sp) * 2);
dp = sp;
while ((c = *dp++)) {
if (c == '\\') {
switch ((c = *dp++)) {
case '\\':
case '"':
case '$':
case '`':
break;
default:
*wp++ = CHAR;
*wp++ = '\\';
break;
}
}
*wp++ = CHAR;
*wp++ = c;
}
afree(sp, ATEMP);
*wp++ = CQUOTE;
state = statep->type = SHEREDELIM;
break;
/* in *(...|...) pattern (*+?@!) */
case SPATTERN:
if (c == /*(*/ ')') {
*wp++ = CPAT;
POP_STATE();
} else if (c == '|') {
*wp++ = SPAT;
} else if (c == '(') {
*wp++ = OPAT;
/* simile for @ */
*wp++ = ' ';
PUSH_STATE(SPATTERN);
} else
goto Sbase1;
break;
}
}
Done:
Xcheck(ws, wp);
if (statep != &states[1])
/* XXX figure out what is missing */
yyerror("no closing quote\n");
/* This done to avoid tests for SHEREDELIM wherever SBASE tested */
if (state == SHEREDELIM)
state = SBASE;
dp = Xstring(ws, wp);
if (state == SBASE && (
#ifndef MKSH_LEGACY_MODE
(c == '&' && !Flag(FSH) && !Flag(FPOSIX)) ||
#endif
c == '<' || c == '>')) {
struct ioword *iop = alloc(sizeof(struct ioword), ATEMP);
if (Xlength(ws, wp) == 0)
iop->unit = c == '<' ? 0 : 1;
else for (iop->unit = 0, c2 = 0; c2 < Xlength(ws, wp); c2 += 2) {
if (dp[c2] != CHAR)
goto no_iop;
if (!ksh_isdigit(dp[c2 + 1]))
goto no_iop;
iop->unit = iop->unit * 10 + ksh_numdig(dp[c2 + 1]);
if (iop->unit >= FDBASE)
goto no_iop;
}
if (c == '&') {
if ((c2 = getsc()) != '>') {
ungetsc(c2);
goto no_iop;
}
c = c2;
iop->ioflag = IOBASH;
} else
iop->ioflag = 0;
c2 = getsc();
/* <<, >>, <> are ok, >< is not */
if (c == c2 || (c == '<' && c2 == '>')) {
iop->ioflag |= c == c2 ?
(c == '>' ? IOCAT : IOHERE) : IORDWR;
if (iop->ioflag == IOHERE) {
if ((c2 = getsc()) == '-')
iop->ioflag |= IOSKIP;
else if (c2 == '<')
iop->ioflag |= IOHERESTR;
else
ungetsc(c2);
}
} else if (c2 == '&')
iop->ioflag |= IODUP | (c == '<' ? IORDUP : 0);
else {
iop->ioflag |= c == '>' ? IOWRITE : IOREAD;
if (c == '>' && c2 == '|')
iop->ioflag |= IOCLOB;
else
ungetsc(c2);
}
iop->ioname = NULL;
iop->delim = NULL;
iop->heredoc = NULL;
/* free word */
Xfree(ws, wp);
yylval.iop = iop;
return (REDIR);
no_iop:
afree(iop, ATEMP);
}
if (wp == dp && state == SBASE) {
/* free word */
Xfree(ws, wp);
/* no word, process LEX1 character */
if ((c == '|') || (c == '&') || (c == ';') || (c == '('/*)*/)) {
if ((c2 = getsc()) == c)
c = (c == ';') ? BREAK :
(c == '|') ? LOGOR :
(c == '&') ? LOGAND :
/* c == '(' ) */ MDPAREN;
else if (c == '|' && c2 == '&')
c = COPROC;
else if (c == ';' && c2 == '|')
c = BRKEV;
else if (c == ';' && c2 == '&')
c = BRKFT;
else
ungetsc(c2);
#ifndef MKSH_SMALL
if (c == BREAK) {
if ((c2 = getsc()) == '&')
c = BRKEV;
else
ungetsc(c2);
}
#endif
} else if (c == '\n') {
if (cf & HEREDELIM)
ungetsc(c);
else {
gethere();
if (cf & CONTIN)
goto Again;
}
}
return (c);
}
/* terminate word */
*wp++ = EOS;
yylval.cp = Xclose(ws, wp);
if (state == SWORD || state == SLETPAREN
/* XXX ONEWORD? */)
return (LWORD);
/* unget terminator */
ungetsc(c);
/*
* note: the alias-vs-function code below depends on several
* interna: starting from here, source->str is not modified;
* the way getsc() and ungetsc() operate; etc.
*/
/* copy word to unprefixed string ident */
sp = yylval.cp;
dp = ident;
while ((dp - ident) < IDENT && (c = *sp++) == CHAR)
*dp++ = *sp++;
if (c != EOS)
/* word is not unquoted */
dp = ident;
/* make sure the ident array stays NUL padded */
memset(dp, 0, (ident + IDENT) - dp + 1);
if (!(cf & (KEYWORD | ALIAS)))
return (LWORD);
if (*ident != '\0') {
struct tbl *p;
uint32_t h = hash(ident);
if ((cf & KEYWORD) && (p = ktsearch(&keywords, ident, h)) &&
(!(cf & ESACONLY) || p->val.i == ESAC ||
p->val.i == /*{*/ '}')) {
afree(yylval.cp, ATEMP);
return (p->val.i);
}
if ((cf & ALIAS) && (p = ktsearch(&aliases, ident, h)) &&
(p->flag & ISSET)) {
/*
* this still points to the same character as the
* ungetsc'd terminator from above
*/
const char *cp = source->str;
/* prefer POSIX but not Korn functions over aliases */
while (*cp == ' ' || *cp == '\t')
/*
* this is like getsc() without skipping
* over Source boundaries (including not
* parsing ungetsc'd characters that got
* pushed into an SREREAD) which is what
* we want here anyway: find out whether
* the alias name is followed by a POSIX
* function definition
*/
++cp;
/* prefer functions over aliases */
if (cp[0] != '(' || cp[1] != ')') {
Source *s = source;
while (s && (s->flags & SF_HASALIAS))
if (s->u.tblp == p)
return (LWORD);
else
s = s->next;
/* push alias expansion */
s = pushs(SALIAS, source->areap);
s->start = s->str = p->val.s;
s->u.tblp = p;
s->flags |= SF_HASALIAS;
s->next = source;
if (source->type == SEOF) {
/* prevent infinite recursion at EOS */
source->u.tblp = p;
source->flags |= SF_HASALIAS;
}
source = s;
afree(yylval.cp, ATEMP);
goto Again;
}
}
} else if (cf & ALIAS) {
/* retain typeset et al. even when quoted */
if (assign_command((dp = wdstrip(yylval.cp, 0))))
strlcpy(ident, dp, sizeof(ident));
afree(dp, ATEMP);
}
return (LWORD);
}
/*
* Read the variable part of a ${...} expression (i.e. up to but not
* including the :[-+?=#%] or close-brace).
*/
static char *
get_brace_var(XString *wsp, char *wp)
{
char c;
enum parse_state {
PS_INITIAL, PS_SAW_HASH, PS_IDENT,
PS_NUMBER, PS_VAR1
} state = PS_INITIAL;
while (/* CONSTCOND */ 1) {
c = getsc();
/* State machine to figure out where the variable part ends. */
switch (state) {
case PS_INITIAL:
if (c == '#' || c == '!' || c == '%') {
state = PS_SAW_HASH;
break;
}
/* FALLTHROUGH */
case PS_SAW_HASH:
if (ksh_isalphx(c))
state = PS_IDENT;
else if (ksh_isdigit(c))
state = PS_NUMBER;
else if (c == '#') {
if (state == PS_SAW_HASH) {
char c2;
c2 = getsc();
ungetsc(c2);
if (c2 != /*{*/ '}') {
ungetsc(c);
goto out;
}
}
state = PS_VAR1;
} else if (ctype(c, C_VAR1))
state = PS_VAR1;
else
goto out;
break;
case PS_IDENT:
if (!ksh_isalnux(c)) {
if (c == '[') {
char *tmp, *p;
if (!arraysub(&tmp))
yyerror("missing ]\n");
*wp++ = c;
for (p = tmp; *p; ) {
Xcheck(*wsp, wp);
*wp++ = *p++;
}
afree(tmp, ATEMP);
/* the ] */
c = getsc();
}
goto out;
}
break;
case PS_NUMBER:
if (!ksh_isdigit(c))
goto out;
break;
case PS_VAR1:
goto out;
}
Xcheck(*wsp, wp);
*wp++ = c;
}
out:
/* end of variable part */
*wp++ = '\0';
ungetsc(c);
return (wp);
}
static int
getint(struct tbl *vp, mksh_ari_t *nump, bool arith)
{
int c, base, neg;
mksh_uari_t num;
const char *s;
bool have_base = false;
if (vp->flag&SPECIAL)
getspec(vp);
/* XXX is it possible for ISSET to be set and val.s to be 0? */
if (!(vp->flag&ISSET) || (!(vp->flag&INTEGER) && vp->val.s == NULL))
return (-1);
if (vp->flag&INTEGER) {
*nump = vp->val.i;
return (vp->type);
}
s = vp->val.s + vp->type;
base = 10;
num = 0;
neg = 0;
if (arith && s[0] == '0' && (s[1] | 0x20) == 'x') {
s += 2;
base = 16;
have_base = true;
}
#ifdef MKSH_LEGACY_MODE
if (arith && s[0] == '0' && ksh_isdigit(s[1]) &&
!(vp->flag & ZEROFIL)) {
/* interpret as octal (deprecated) */
base = 8;
have_base = true;
}
#endif
while ((c = *s++)) {
if (c == '-') {
neg++;
continue;
} else if (c == '#') {
if (have_base || num < 1 || num > 36)
return (-1);
base = (int)num;
if (base == 1) {
unsigned int wc;
if (!UTFMODE)
wc = *(const unsigned char *)s;
else if (utf_mbtowc(&wc, s) == (size_t)-1)
/* OPTU-8 -> OPTU-16 */
/*
* (with a twist: 1#\uEF80 converts
* the same as 1#\x80 does, thus is
* not round-tripping correctly XXX)
*/
wc = 0xEF00 + *(const unsigned char *)s;
*nump = (mksh_ari_t)wc;
return (1);
}
num = 0;
have_base = true;
continue;
} else if (ksh_isdigit(c))
c -= '0';
else if (ksh_islower(c))
c -= 'a' - 10;
else if (ksh_isupper(c))
c -= 'A' - 10;
else
return (-1);
if (c < 0 || c >= base)
return (-1);
num = num * base + c;
}
*nump = neg ? -((mksh_ari_t)num) : (mksh_ari_t)num;
return (base);
}
