-/* $OpenBSD: parse.y,v 1.7 2015/07/21 11:04:06 zhuk Exp $ */
+/* $OpenBSD: parse.y,v 1.9 2015/07/22 20:15:24 zhuk Exp $ */
/*
* Copyright (c) 2015 Ted Unangst <tedu@openbsd.org>
*
};
const char *str;
};
+ int lineno;
+ int colno;
} yystype;
#define YYSTYPE yystype
struct rule **rules;
int nrules, maxrules;
+int parse_errors = 0;
void yyerror(const char *, ...);
int yylex(void);
grammar: /* empty */
| grammar '\n'
| grammar rule '\n'
+ | error '\n'
;
rule: action ident target cmd {
$$.options = $1.options | $2.options;
$$.envlist = $1.envlist;
if ($2.envlist) {
- if ($$.envlist)
- errx(1, "can't have two keepenv sections");
- else
+ if ($$.envlist) {
+ yyerror("can't have two keepenv sections");
+ YYERROR;
+ } else
$$.envlist = $2.envlist;
}
} ;
va_list va;
va_start(va, fmt);
- verrx(1, fmt, va);
+ vfprintf(stderr, fmt, va);
+ va_end(va);
+ fprintf(stderr, " at line %d\n", yylval.lineno + 1);
+ parse_errors++;
}
struct keyword {
yylex(void)
{
char buf[1024], *ebuf, *p, *str;
- int i, c, next;
+ int i, c, quotes = 0, escape = 0, qpos = -1, nonkw = 0;
p = buf;
ebuf = buf + sizeof(buf);
+
repeat:
- c = getc(yyfp);
+ /* skip whitespace first */
+ for (c = getc(yyfp); c == ' ' || c == '\t'; c = getc(yyfp))
+ yylval.colno++;
+
+ /* check for special one-character constructions */
switch (c) {
- case ' ':
- case '\t':
- goto repeat; /* skip spaces */
- case '\\':
- next = getc(yyfp);
- if (next == '\n')
- goto repeat;
- else
- c = next;
- case '\n':
- case '{':
- case '}':
- return c;
- case '#':
- while ((c = getc(yyfp)) != '\n' && c != EOF)
- ; /* skip comments */
- if (c == EOF)
+ case '\n':
+ yylval.colno = 0;
+ yylval.lineno++;
+ /* FALLTHROUGH */
+ case '{':
+ case '}':
+ return c;
+ case '#':
+ /* skip comments; NUL is allowed; no continuation */
+ while ((c = getc(yyfp)) != '\n')
+ if (c == EOF)
+ return 0;
+ yylval.colno = 0;
+ yylval.lineno++;
+ return c;
+ case EOF:
return 0;
- return c;
- case EOF:
- return 0;
}
- while (1) {
+
+ /* parsing next word */
+ for (;; c = getc(yyfp), yylval.colno++) {
switch (c) {
+ case '\0':
+ yyerror("unallowed character NUL in column %d", yylval.colno + 1);
+ escape = 0;
+ continue;
+ case '\\':
+ escape = !escape;
+ if (escape)
+ continue;
+ break;
case '\n':
+ if (quotes)
+ yyerror("unterminated quotes in column %d",
+ qpos + 1);
+ if (escape) {
+ nonkw = 1;
+ escape = 0;
+ continue;
+ }
+ goto eow;
+ case EOF:
+ if (escape)
+ yyerror("unterminated escape in column %d",
+ yylval.colno);
+ if (quotes)
+ yyerror("unterminated quotes in column %d",
+ qpos + 1);
+ goto eow;
+ /* FALLTHROUGH */
case '{':
case '}':
case '#':
case ' ':
case '\t':
- case EOF:
- goto eow;
+ if (!escape && !quotes)
+ goto eow;
+ break;
+ case '"':
+ if (!escape) {
+ quotes = !quotes;
+ if (quotes) {
+ nonkw = 1;
+ qpos = yylval.colno;
+ }
+ continue;
+ }
}
*p++ = c;
if (p == ebuf)
- yyerror("too much stuff");
- c = getc(yyfp);
+ yyerror("too long line");
+ escape = 0;
}
+
eow:
*p = 0;
if (c != EOF)
ungetc(c, yyfp);
- for (i = 0; i < sizeof(keywords) / sizeof(keywords[0]); i++) {
- if (strcmp(buf, keywords[i].word) == 0)
- return keywords[i].token;
+ if (p == buf) {
+ /*
+ * There could be a number of reasons for empty buffer, and we handle
+ * all of them here, to avoid cluttering the main loop.
+ */
+ if (c == EOF)
+ return 0;
+ else if (qpos == -1) /* accept, e.g., empty args: cmd foo args "" */
+ goto repeat;
+ }
+ if (!nonkw) {
+ for (i = 0; i < sizeof(keywords) / sizeof(keywords[0]); i++) {
+ if (strcmp(buf, keywords[i].word) == 0)
+ return keywords[i].token;
+ }
}
if ((str = strdup(buf)) == NULL)
err(1, "strdup");