static void del_token_fifo(struct token_fifo *tf)
{
size_t i;
for (i = 0; i < tf->nt; i ++)
if (S_TOKEN(tf->t[i].type)) freemem(tf->t[i].name);
if (tf->nt) freemem(tf->t);
}
/*
* Send a token to the output (a token_fifo in lexer mode, the output
* buffer in stand alone mode).
*/
void print_token(struct lexer_state *ls, struct token *t, long uz_line)
{
char *x = t->name;
if (uz_line && t->line < 0) t->line = uz_line;
if (ls->flags & LEXER) {
struct token at;
at = *t;
if (S_TOKEN(t->type)) {
at.name = sdup(at.name);
throw_away(ls->gf, at.name);
}
aol(ls->output_fifo->t, ls->output_fifo->nt, at,
TOKEN_LIST_MEMG);
return;
}
if (ls->flags & KEEP_OUTPUT) {
for (; ls->oline < ls->line;) put_char(ls, '\n');
}
if (!S_TOKEN(t->type)) x = operators_name[t->type];
for (; *x; x ++) put_char(ls, *x);
}
/*
* print the content of a macro, in #define form
*/
static void print_macro(void *vm)
{
struct macro *m = vm;
char *mname = HASH_ITEM_NAME(m);
int x = check_special_macro(mname);
size_t i;
if (x != MAC_NONE) {
fprintf(emit_output, "/* #define %s */ /* special */\n",
mname);
return;
}
fprintf(emit_output, "#define %s", mname);
if (m->narg >= 0) {
fprintf(emit_output, "(");
for (i = 0; i < (size_t)(m->narg); i ++) {
fprintf(emit_output, i ? ", %s" : "%s", m->arg[i]);
}
if (m->vaarg) {
fputs(m->narg ? ", ..." : "...", emit_output);
}
fprintf(emit_output, ")");
}
if (m->cval.length == 0) {
fputc('\n', emit_output);
return;
}
fputc(' ', emit_output);
for (i = 0; i < m->cval.length;) {
int tt = m->cval.t[i ++];
if (tt == MACROARG) {
unsigned anum = m->cval.t[i];
if (anum >= 128) anum = ((anum & 127U) << 8)
| m->cval.t[++ i];
if (anum == (unsigned)m->narg)
fputs("__VA_ARGS__", emit_output);
else
fputs(m->arg[anum], emit_output);
i ++;
}
else if (S_TOKEN(tt)) {
fputs((char *)(m->cval.t + i), emit_output);
i += 1 + strlen((char *)(m->cval.t + i));
} else fputs(operators_name[tt], emit_output);
}
fputc('\n', emit_output);
}
/*
* compare two token_fifo, return 0 if they are identical, 1 otherwise.
* All whitespace tokens are considered identical, but sequences of
* whitespace are not shrinked.
*/
int cmp_token_list(struct token_fifo *f1, struct token_fifo *f2)
{
size_t i;
if (f1->nt != f2->nt) return 1;
for (i = 0; i < f1->nt; i ++) {
if (ttMWS(f1->t[i].type) && ttMWS(f2->t[i].type)) continue;
if (f1->t[i].type != f2->t[i].type) return 1;
if (f1->t[i].type == MACROARG
&& f1->t[i].line != f2->t[i].line) return 1;
if (S_TOKEN(f1->t[i].type)
&& strcmp(f1->t[i].name, f2->t[i].name)) return 1;
}
return 0;
}
/*
* Send a token to the output at a given line (this is for text output
* and unreplaced macros due to lack of arguments).
*/
static void print_token_nailed(struct lexer_state *ls, struct token *t,
long nail_line)
{
char *x = t->name;
if (ls->flags & LEXER) {
print_token(ls, t, 0);
return;
}
if (ls->flags & KEEP_OUTPUT) {
for (; ls->oline < nail_line;) put_char(ls, '\n');
}
if (!S_TOKEN(t->type)) x = operators_name[t->type];
for (; *x; x ++) put_char(ls, *x);
}
/*
* for #unassert
*/
int handle_unassert(struct lexer_state *ls)
{
int ltww;
struct ucpp_token t;
struct token_fifo atl;
struct assert *a;
int ret = -1;
long l = ls->line;
int nnp;
size_t i;
atl.art = atl.nt = 0;
while (!next_token(ls)) {
if (ls->ctok->type == NEWLINE) break;
if (ttMWS(ls->ctok->type)) continue;
if (ls->ctok->type == NAME) {
if (!(a = HTT_get(&assertions, ls->ctok->name))) {
ret = 0;
goto handle_unassert_warp;
}
goto handle_unassert_next;
}
error(l, "illegal assertion name for #unassert");
goto handle_unassert_warp;
}
goto handle_unassert_trunc;
handle_unassert_next:
while (!next_token(ls)) {
if (ls->ctok->type == NEWLINE) break;
if (ttMWS(ls->ctok->type)) continue;
if (ls->ctok->type != LPAR) {
error(l, "syntax error in #unassert");
goto handle_unassert_warp;
}
goto handle_unassert_next2;
}
if (emit_assertions)
fprintf(emit_output, "#unassert %s\n", HASH_ITEM_NAME(a));
HTT_del(&assertions, HASH_ITEM_NAME(a));
return 0;
handle_unassert_next2:
for (nnp = 1, ltww = 1; nnp && !next_token(ls);) {
if (ls->ctok->type == NEWLINE) break;
if (ltww && ttMWS(ls->ctok->type)) continue;
ltww = ttMWS(ls->ctok->type);
if (ls->ctok->type == LPAR) nnp ++;
else if (ls->ctok->type == RPAR) {
if (!(-- nnp)) goto handle_unassert_next3;
}
t.type = ls->ctok->type;
if (S_TOKEN(t.type)) t.name = sdup(ls->ctok->name);
aol(atl.t, atl.nt, t, TOKEN_LIST_MEMG);
}
goto handle_unassert_trunc;
handle_unassert_next3:
while (!next_token(ls) && ls->ctok->type != NEWLINE) {
if (!ttWHI(ls->ctok->type) && (ls->flags & WARN_STANDARD)) {
warning(l, "trailing garbage in #unassert");
}
}
if (atl.nt && ttMWS(atl.t[atl.nt - 1].type) && (-- atl.nt) == 0)
freemem(atl.t);
if (atl.nt == 0) {
error(l, "void assertion in #unassert");
return ret;
}
for (i = 0; i < a->nbval && cmp_token_list(&atl, a->val + i); i ++);
if (i != a->nbval) {
/* we have it, undefine it */
del_token_fifo(a->val + i);
if (i < (a->nbval - 1))
mmvwo(a->val + i, a->val + i + 1, (a->nbval - i - 1)
* sizeof(struct token_fifo));
if ((-- a->nbval) == 0) freemem(a->val);
if (emit_assertions) {
fprintf(emit_output, "#unassert %s(",
HASH_ITEM_NAME(a));
print_token_fifo(&atl);
fputs(")\n", emit_output);
}
}
ret = 0;
goto handle_unassert_finish;
handle_unassert_trunc:
error(l, "unfinished #unassert");
handle_unassert_finish:
if (atl.nt) del_token_fifo(&atl);
return ret;
handle_unassert_warp:
while (!next_token(ls) && ls->ctok->type != NEWLINE);
return ret;
}
/*
* for #assert
* Assertions are not part of the ISO-C89 standard, but they are sometimes
* encountered, for instance in Solaris standard include files.
*/
int handle_assert(struct lexer_state *ls)
{
int ina = 0, ltww;
struct ucpp_token t;
struct token_fifo *atl = 0;
struct assert *a;
char *aname;
int ret = -1;
long l = ls->line;
int nnp;
size_t i;
while (!next_token(ls)) {
if (ls->ctok->type == NEWLINE) break;
if (ttMWS(ls->ctok->type)) continue;
if (ls->ctok->type == NAME) {
if (!(a = HTT_get(&assertions, ls->ctok->name))) {
a = new_assertion();
aname = sdup(ls->ctok->name);
ina = 1;
}
goto handle_assert_next;
}
error(l, "illegal assertion name for #assert");
goto handle_assert_warp_ign;
}
goto handle_assert_trunc;
handle_assert_next:
while (!next_token(ls)) {
if (ls->ctok->type == NEWLINE) break;
if (ttMWS(ls->ctok->type)) continue;
if (ls->ctok->type != LPAR) {
error(l, "syntax error in #assert");
goto handle_assert_warp_ign;
}
goto handle_assert_next2;
}
goto handle_assert_trunc;
handle_assert_next2:
atl = getmem(sizeof(struct token_fifo));
atl->art = atl->nt = 0;
for (nnp = 1, ltww = 1; nnp && !next_token(ls);) {
if (ls->ctok->type == NEWLINE) break;
if (ltww && ttMWS(ls->ctok->type)) continue;
ltww = ttMWS(ls->ctok->type);
if (ls->ctok->type == LPAR) nnp ++;
else if (ls->ctok->type == RPAR) {
if (!(-- nnp)) goto handle_assert_next3;
}
t.type = ls->ctok->type;
if (S_TOKEN(t.type)) t.name = sdup(ls->ctok->name);
aol(atl->t, atl->nt, t, TOKEN_LIST_MEMG);
}
goto handle_assert_trunc;
handle_assert_next3:
while (!next_token(ls) && ls->ctok->type != NEWLINE) {
if (!ttWHI(ls->ctok->type) && (ls->flags & WARN_STANDARD)) {
warning(l, "trailing garbage in #assert");
}
}
if (atl->nt && ttMWS(atl->t[atl->nt - 1].type) && (-- atl->nt) == 0)
freemem(atl->t);
if (atl->nt == 0) {
error(l, "void assertion in #assert");
goto handle_assert_error;
}
for (i = 0; i < a->nbval && cmp_token_list(atl, a->val + i); i ++);
if (i != a->nbval) {
/* we already have it */
ret = 0;
goto handle_assert_error;
}
/* This is a new assertion. Let's keep it. */
aol(a->val, a->nbval, *atl, TOKEN_LIST_MEMG);
if (ina) {
HTT_put(&assertions, a, aname);
freemem(aname);
}
if (emit_assertions) {
fprintf(emit_output, "#assert %s(", HASH_ITEM_NAME(a));
print_token_fifo(atl);
fputs(")\n", emit_output);
}
freemem(atl);
return 0;
handle_assert_trunc:
error(l, "unfinished #assert");
handle_assert_error:
if (atl) {
del_token_fifo(atl);
freemem(atl);
}
if (ina) {
freemem(aname);
freemem(a);
}
return ret;
handle_assert_warp_ign:
while (!next_token(ls) && ls->ctok->type != NEWLINE);
if (ina) {
freemem(aname);
freemem(a);
}
return ret;
}
/*
* We found a #define directive; parse the end of the line, perform
* sanity checks, store the new macro into the "macros" hash table.
*
* In case of a redefinition of a macro: we enforce the rule that a
* macro should be redefined identically, including the spelling of
* parameters. We emit an error on offending code; dura lex, sed lex.
* After all, it is easy to avoid such problems, with a #undef directive.
*/
int handle_define(struct lexer_state *ls)
{
struct macro *m = 0, *n;
struct token_fifo mv;
int ltwws = 1, redef = 0;
char *mname = 0;
int narg;
size_t nt;
long l = ls->line;
mv.art = mv.nt = 0;
mv.t = NULL;
/* find the next non-white token on the line, this should be
the macro name */
while (!next_token(ls) && ls->ctok->type != NEWLINE) {
if (ttMWS(ls->ctok->type)) continue;
if (ls->ctok->type == NAME) mname = sdup(ls->ctok->name);
break;
}
if (mname == 0) {
error(l, "missing macro name");
return 1;
}
if (check_special_macro(mname)) {
error(l, "trying to redefine the special macro %s", mname);
goto warp_error;
}
/*
* If a macro with this name was already defined: the K&R
* states that the new macro should be identical to the old one
* (with some arcane rule of equivalence of whitespace); otherwise,
* redefining the macro is an error. Most preprocessors would
* only emit a warning (or nothing at all) on an unidentical
* redefinition.
*
* Since it is easy to avoid this error (with a #undef directive),
* we choose to enforce the rule and emit an error.
*/
if ((n = HTT_get(¯os, mname)) != 0) {
/* redefinition of a macro: we must check that we define
it identical */
redef = 1;
n->cval.rp = 0;
freemem(mname);
mname = 0;
}
if (!redef) {
m = new_macro();
m->narg = -1;
#define mval mv
}
if (next_token(ls)) goto define_end;
/*
* Check if the token immediately following the macro name is
* a left parenthesis; if so, then this is a macro with arguments.
* Collect their names and try to match the next parenthesis.
*/
if (ls->ctok->type == LPAR) {
int i, j;
int need_comma = 0, saw_mdots = 0;
narg = 0;
while (!next_token(ls)) {
if (ls->ctok->type == NEWLINE) {
error(l, "truncated macro definition");
goto define_error;
}
if (ls->ctok->type == COMMA) {
if (saw_mdots) {
error(l, "'...' must end the macro "
"argument list");
goto warp_error;
}
if (!need_comma) {
error(l, "void macro argument");
goto warp_error;
}
need_comma = 0;
continue;
} else if (ls->ctok->type == NAME) {
if (saw_mdots) {
error(l, "'...' must end the macro "
"argument list");
goto warp_error;
}
if (need_comma) {
error(l, "missing comma in "
"macro argument list");
goto warp_error;
}
if (!redef) {
aol(m->arg, narg,
sdup(ls->ctok->name), 8);
/* we must keep track of m->narg
so that cleanup in case of
error works. */
m->narg = narg;
if (narg == 128
&& (ls->flags & WARN_STANDARD))
warning(l, "more arguments to "
"macro than the ISO "
"limit (127)");
if (narg == 32767) {
error(l, "too many arguments "
"in macro definition "
"(max 32766)");
goto warp_error;
}
} else {
/* this is a redefinition of the
macro; check equality between
old and new definitions */
if (narg >= n->narg) goto redef_error;
if (strcmp(ls->ctok->name,
n->arg[narg ++]))
goto redef_error;
}
need_comma = 1;
continue;
} else if ((ls->flags & MACRO_VAARG)
&& ls->ctok->type == MDOTS) {
if (need_comma) {
error(l, "missing comma before '...'");
goto warp_error;
}
if (redef && !n->vaarg) goto redef_error;
if (!redef) m->vaarg = 1;
saw_mdots = 1;
need_comma = 1;
continue;
} else if (ls->ctok->type == RPAR) {
if (narg > 0 && !need_comma) {
error(l, "void macro argument");
goto warp_error;
}
if (redef && n->vaarg && !saw_mdots)
goto redef_error;
break;
} else if (ttMWS(ls->ctok->type)) {
continue;
}
error(l, "invalid macro argument");
goto warp_error;
}
if (!redef) {
for (i = 1; i < narg; i ++) for (j = 0; j < i; j ++)
if (!strcmp(m->arg[i], m->arg[j])) {
error(l, "duplicate macro "
"argument");
goto warp_error;
}
}
if (!redef) m->narg = narg;
} else {
if (!ttWHI(ls->ctok->type) && (ls->flags & WARN_STANDARD))
warning(ls->line, "identifier not followed by "
"whitespace in #define");
ls->flags |= READ_AGAIN;
narg = 0;
}
if (redef) nt = 0;
/* now, we have the arguments. Let's get the macro contents. */
while (!next_token(ls) && ls->ctok->type != NEWLINE) {
struct token t;
t.type = ls->ctok->type;
if (ltwws && ttMWS(t.type)) continue;
t.line = 0;
if (t.type == NAME) {
int i;
if ((ls->flags & MACRO_VAARG)
&& !strcmp(ls->ctok->name, "__VA_ARGS__")) {
if (redef) {
if (!n->vaarg) goto redef_error;
} else if (!m->vaarg) {
error(l, "'__VA_ARGS__' is forbidden "
"in macros with a fixed "
"number of arguments");
goto warp_error;
}
t.type = MACROARG;
t.line = redef ? n->narg : m->narg;
}
for (i = 0; i < narg; i ++)
if (!strcmp(redef ? n->arg[i] : m->arg[i],
ls->ctok->name)) {
t.type = MACROARG;
/* this is a hack: we store the
argument number in the line field */
t.line = i;
break;
}
}
if (!redef && S_TOKEN(t.type)) t.name = sdup(ls->ctok->name);
if (ttMWS(t.type)) {
if (ltwws) continue;
#ifdef SEMPER_FIDELIS
t.type = OPT_NONE;
#else
t.type = NONE;
#endif
ltwws = 1;
} else ltwws = 0;
if (!redef) {
/* we ensure that each macro token has a correct
line number */
if (t.type != MACROARG) t.line = 1;
aol(mval.t, mval.nt, t, TOKEN_LIST_MEMG);
} else {
int tt;
if (n->cval.rp >= n->cval.length) {
#ifdef SEMPER_FIDELIS
if (t.type != OPT_NONE) goto redef_error;
#else
if (t.type != NONE) goto redef_error;
#endif
} else if (t.type != n->cval.t[n->cval.rp]) {
goto redef_error;
} else if (t.type == MACROARG) {
unsigned anum = n->cval.t[n->cval.rp + 1];
if (anum >= 128U) anum = ((anum & 127U) << 8)
| m->cval.t[n->cval.rp + 2];
if (anum != (unsigned)t.line) goto redef_error;
} else if (S_TOKEN(t.type) && strcmp(ls->ctok->name,
(char *)(n->cval.t + n->cval.rp + 1))) {
goto redef_error;
}
tt = n->cval.t[n->cval.rp ++];
if (S_TOKEN(tt)) n->cval.rp += 1
+ strlen((char *)(n->cval.t + n->cval.rp));
else if (tt == MACROARG) {
if (n->cval.t[++ n->cval.rp] >= 128)
n->cval.rp ++;
}
nt ++;
}
}
if (redef) {
if (n->cval.rp < n->cval.length) goto redef_error_2;
return 0;
}
/* now we have the complete macro; perform some checks about
the operators # and ##, and, if everything is ok,
store the macro into the hash table */
define_end:
#ifdef SEMPER_FIDELIS
if (mval.nt && mval.t[mval.nt - 1].type == OPT_NONE) {
#else
if (mval.nt && mval.t[mval.nt - 1].type == NONE) {
#endif
mval.nt --;
if (mval.nt == 0) freemem(mval.t);
}
if (mval.nt != 0) {
/* some checks about the macro */
if (mval.t[0].type == DSHARP
|| mval.t[0].type == DIG_DSHARP
|| mval.t[mval.nt - 1].type == DSHARP
|| mval.t[mval.nt - 1].type == DIG_DSHARP) {
error(l, "operator '##' may neither begin "
"nor end a macro");
goto define_error;
}
if (m->narg >= 0) {
size_t i;
for (i = 0; i < mval.nt; i ++)
if ((mval.t[i].type == SHARP
|| mval.t[i].type == DIG_SHARP) &&
(i == (mval.nt - 1)
|| (ttMWS(mval.t[i + 1].type) &&
(i == mval.nt - 2
|| mval.t[i + 2].type != MACROARG))
|| (!ttMWS(mval.t[i + 1].type)
&& mval.t[i + 1].type != MACROARG))) {
error(l, "operator '#' not followed "
"by a macro argument");
goto define_error;
}
}
}
{
size_t i, l;
for (i = 0, l = 0; i < mval.nt; i ++) {
l ++;
if (S_TOKEN(mval.t[i].type))
l += 1 + strlen(mval.t[i].name);
else if (mval.t[i].type == MACROARG) {
l ++;
if (mval.t[i].line >= 128) l ++;
}
}
m->cval.length = l;
if (l) m->cval.t = getmem(l);
for (i = 0, l = 0; i < mval.nt; i ++) {
m->cval.t[l ++] = mval.t[i].type;
if (S_TOKEN(mval.t[i].type)) {
size_t x = 1 + strlen(mval.t[i].name);
mmv(m->cval.t + l, mval.t[i].name, x);
l += x;
freemem(mval.t[i].name);
}
else if (mval.t[i].type == MACROARG) {
unsigned anum = mval.t[i].line;
if (anum >= 128) {
m->cval.t[l ++] = 128 | (anum >> 8);
m->cval.t[l ++] = anum & 0xFF;
} else {
m->cval.t[l ++] = anum;
}
}
}