X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/82b6d266fdd393132bf7073acaea31f4f624b223..065fbd27af96e4aa82b5f3354870b0dcfee72ce8:/src/lex.c

diff --git a/src/lex.c b/src/lex.c
index 61f1c262..deb6eb7a 100644
--- a/src/lex.c
+++ b/src/lex.c
@@ -21,31 +21,37 @@
 #include "system.h"
 #include "getargs.h"
 #include "files.h"
-#include "getopt.h"		/* for optarg */
 #include "symtab.h"
 #include "options.h"
 #include "lex.h"
-#include "xalloc.h"
 #include "complain.h"
 #include "gram.h"
 #include "quote.h"
 
 /* Buffer for storing the current token.  */
-struct obstack token_obstack;
+static struct obstack token_obstack;
 const char *token_buffer = NULL;
 
-bucket *symval;
+bucket *symval = NULL;
 int numval;
 
-static int unlexed;		/* these two describe a token to be reread */
-static bucket *unlexed_symval;	/* by the next call to lex */
-
+/* A token to be reread, see unlex and lex. */
+static token_t unlexed = tok_undef;
+static bucket *unlexed_symval = NULL;
+static const char *unlexed_token_buffer = NULL;
 
 void
-init_lex (void)
+lex_init (void)
 {
   obstack_init (&token_obstack);
-  unlexed = -1;
+  unlexed = tok_undef;
+}
+
+
+void
+lex_free (void)
+{
+  obstack_free (&token_obstack, NULL);
 }
 
 
@@ -124,7 +130,7 @@ skip_white_space (void)
 | Do a getc, but give error message if EOF encountered |
 `-----------------------------------------------------*/
 
-static int
+int
 xgetc (FILE *f)
 {
   int c = getc (f);
@@ -134,75 +140,62 @@ xgetc (FILE *f)
 }
 
 
-/*------------------------------------------------------------------.
-| Read one literal character from finput.  Process \ escapes.       |
-| Append the normalized string version of the char to OUT.  Assign  |
-| the character code to *PCODE. Return 1 unless the character is an |
-| unescaped `term' or \n report error for \n.                       |
-`------------------------------------------------------------------*/
+/*---------------------------------------------------------------.
+| Read one literal character from FINPUT, process \-escapes, and |
+| return the character.                                          |
+`---------------------------------------------------------------*/
 
-/* FIXME: We could directly work in the obstack, but that would make
-   it more difficult to move to quotearg some day.  So for the time
-   being, I prefer have literalchar behave like quotearg, and change
-   my mind later if I was wrong.  */
-
-static int
-literalchar (struct obstack *out, int *pcode, char term)
+char
+literalchar (void)
 {
   int c;
-  char buf[4096];
-  char *cp;
-  int code;
-  int wasquote = 0;
+  int res;
 
   c = xgetc (finput);
   if (c == '\n')
     {
       complain (_("unescaped newline in constant"));
       ungetc (c, finput);
-      code = '?';
-      wasquote = 1;
+      res = '?';
     }
   else if (c != '\\')
     {
-      code = c;
-      if (c == term)
-	wasquote = 1;
+      res = c;
     }
   else
     {
       c = xgetc (finput);
       if (c == 't')
-	code = '\t';
+	res = '\t';
       else if (c == 'n')
-	code = '\n';
+	res = '\n';
       else if (c == 'a')
-	code = '\007';
+	res = '\007';
       else if (c == 'r')
-	code = '\r';
+	res = '\r';
       else if (c == 'f')
-	code = '\f';
+	res = '\f';
       else if (c == 'b')
-	code = '\b';
+	res = '\b';
       else if (c == 'v')
-	code = '\013';
+	res = '\013';
       else if (c == '\\')
-	code = '\\';
+	res = '\\';
       else if (c == '\'')
-	code = '\'';
+	res = '\'';
       else if (c == '\"')
-	code = '\"';
+	res = '\"';
       else if (c <= '7' && c >= '0')
 	{
-	  code = 0;
+	  res = 0;
 	  while (c <= '7' && c >= '0')
 	    {
-	      code = (code * 8) + (c - '0');
-	      if (code >= 256 || code < 0)
+	      res = (res * 8) + (c - '0');
+	      if (res >= 256 || res < 0)
 		{
 		  complain (_("octal value outside range 0...255: `\\%o'"),
-			    code);
-		  code &= 0xFF;
+			    res);
+		  res &= 0xFF;
 		  break;
 		}
 	      c = xgetc (finput);
@@ -212,21 +205,21 @@ literalchar (struct obstack *out, int *pcode, char term)
       else if (c == 'x')
 	{
 	  c = xgetc (finput);
-	  code = 0;
+	  res = 0;
 	  while (1)
 	    {
 	      if (c >= '0' && c <= '9')
-		code *= 16, code += c - '0';
+		res *= 16, res += c - '0';
 	      else if (c >= 'a' && c <= 'f')
-		code *= 16, code += c - 'a' + 10;
+		res *= 16, res += c - 'a' + 10;
 	      else if (c >= 'A' && c <= 'F')
-		code *= 16, code += c - 'A' + 10;
+		res *= 16, res += c - 'A' + 10;
 	      else
 		break;
-	      if (code >= 256 || code < 0)
+	      if (res >= 256 || res < 0)
 		{
-		  complain (_("hexadecimal value above 255: `\\x%x'"), code);
-		  code &= 0xFF;
+		  complain (_("hexadecimal value above 255: `\\x%x'"), res);
+		  res &= 0xFF;
 		  break;
 		}
 	      c = xgetc (finput);
@@ -239,84 +232,19 @@ literalchar (struct obstack *out, int *pcode, char term)
 	  badchar[0] = c;
 	  complain (_("unknown escape sequence: `\\' followed by `%s'"),
 		    quote (badchar));
-	  code = '?';
+	  res = '?';
 	}
     }				/* has \ */
 
-  /* now fill BUF with the canonical name for this character as a
-     literal token.  Do not use what the user typed, so that `\012'
-     and `\n' can be interchangeable.  */
-
-  cp = buf;
-  if (code == term && wasquote)
-    *cp++ = code;
-  else if (code == '\\')
-    {
-      *cp++ = '\\';
-      *cp++ = '\\';
-    }
-  else if (code == '\'')
-    {
-      *cp++ = '\\';
-      *cp++ = '\'';
-    }
-  else if (code == '\"')
-    {
-      *cp++ = '\\';
-      *cp++ = '\"';
-    }
-  else if (code >= 040 && code < 0177)
-    *cp++ = code;
-  else if (code == '\t')
-    {
-      *cp++ = '\\';
-      *cp++ = 't';
-    }
-  else if (code == '\n')
-    {
-      *cp++ = '\\';
-      *cp++ = 'n';
-    }
-  else if (code == '\r')
-    {
-      *cp++ = '\\';
-      *cp++ = 'r';
-    }
-  else if (code == '\v')
-    {
-      *cp++ = '\\';
-      *cp++ = 'v';
-    }
-  else if (code == '\b')
-    {
-      *cp++ = '\\';
-      *cp++ = 'b';
-    }
-  else if (code == '\f')
-    {
-      *cp++ = '\\';
-      *cp++ = 'f';
-    }
-  else
-    {
-      *cp++ = '\\';
-      *cp++ = code / 0100 + '0';
-      *cp++ = ((code / 010) & 07) + '0';
-      *cp++ = (code & 07) + '0';
-    }
-  *cp = '\0';
-
-  if (out)
-    obstack_sgrow (out, buf);
-  *pcode = code;
-  return !wasquote;
+  return res;
 }
 
 
 void
-unlex (int token)
+unlex (token_t token)
 {
   unlexed = token;
+  unlexed_token_buffer = token_buffer;
   unlexed_symval = symval;
 }
 
@@ -357,12 +285,13 @@ lex (void)
   /* Just to make sure. */
   token_buffer = NULL;
 
-  if (unlexed >= 0)
+  if (unlexed != tok_undef)
     {
+      token_t res = unlexed;
       symval = unlexed_symval;
-      c = unlexed;
-      unlexed = -1;
-      return c;
+      token_buffer = unlexed_token_buffer;
+      unlexed = tok_undef;
+      return res;
     }
 
   c = skip_white_space ();
@@ -418,27 +347,25 @@ lex (void)
     case '\'':
       /* parse the literal token and compute character code in  code  */
 
-      translations = -1;
       {
-	int code, discode;
+	int code = literalchar ();
 
 	obstack_1grow (&token_obstack, '\'');
-	literalchar (&token_obstack, &code, '\'');
+	obstack_1grow (&token_obstack, code);
 
 	c = getc (finput);
 	if (c != '\'')
 	  {
 	    complain (_("use \"...\" for multi-character literal tokens"));
-	    while (1)
-	      if (!literalchar (0, &discode, '\''))
-		break;
+	    while (literalchar () != '\'')
+	      /* Skip. */;
 	  }
 	obstack_1grow (&token_obstack, '\'');
 	obstack_1grow (&token_obstack, '\0');
 	token_buffer = obstack_finish (&token_obstack);
 	symval = getsym (token_buffer);
 	symval->class = token_sym;
-	if (!symval->user_token_number)
+	if (symval->user_token_number == SUNDEF)
 	  symval->user_token_number = code;
 	return tok_identifier;
       }
@@ -446,14 +373,17 @@ lex (void)
     case '\"':
       /* parse the literal string token and treat as an identifier */
 
-      translations = -1;
       {
 	int code;		/* ignored here */
 
 	obstack_1grow (&token_obstack, '\"');
 	/* Read up to and including ".  */
-	while (literalchar (&token_obstack, &code, '\"'))
-	  /* nothing */;
+	do
+	  {
+	    code = literalchar ();
+	    obstack_1grow (&token_obstack, code);
+	  }
+	while (code != '\"');
 	obstack_1grow (&token_obstack, '\0');
 	token_buffer = obstack_finish (&token_obstack);
 
@@ -464,32 +394,40 @@ lex (void)
       }
 
     case ',':
+      token_buffer = ",";
       return tok_comma;
 
     case ':':
+      token_buffer = ":";
       return tok_colon;
 
     case ';':
+      token_buffer = ";";
       return tok_semicolon;
 
     case '|':
+      token_buffer = "|";
       return tok_bar;
 
     case '{':
+      token_buffer = "{";
       return tok_left_curly;
 
     case '=':
+      obstack_1grow (&token_obstack, c);
       do
 	{
 	  c = getc (finput);
+	  obstack_1grow (&token_obstack, c);
 	  if (c == '\n')
 	    lineno++;
 	}
       while (c == ' ' || c == '\n' || c == '\t');
+      obstack_1grow (&token_obstack, '\0');
+      token_buffer = obstack_finish (&token_obstack);
 
       if (c == '{')
 	{
-	  token_buffer = "={";
 	  return tok_left_curly;
 	}
       else
@@ -506,6 +444,9 @@ lex (void)
       return parse_percent_token ();
 
     default:
+      obstack_1grow (&token_obstack, c);
+      obstack_1grow (&token_obstack, '\0');
+      token_buffer = obstack_finish (&token_obstack);
       return tok_illegal;
     }
 }
@@ -516,28 +457,31 @@ lex (void)
 static int
 option_strcmp (const char *left, const char *right)
 {
-    const unsigned char *l, *r;
-    int c;
-
-    assert(left != NULL && right != NULL);     
-    l = (const unsigned char *)left;
-    r = (const unsigned char *)right;
-    while (((c = *l - *r++) == 0 && *l != '\0')
-	   || ((*l == '-' || *l == '_') && (*r == '_' || *r == '-')))
-        l++;
-    return c;
+  const unsigned char *l, *r;
+  int c;
+
+  assert (left);
+  assert (right);
+  l = (const unsigned char *)left;
+  r = (const unsigned char *)right;
+  while (((c = *l - *r++) == 0 && *l != '\0')
+	 || ((*l == '-' || *l == '_') && (*r == '_' || *r == '-')))
+    l++;
+  return c;
 }
 
 /* Parse a token which starts with %.
    Assumes the % has already been read and discarded.  */
 
-int
+token_t
 parse_percent_token (void)
 {
-  int c;
-  const struct option_table_struct *tx;
+  const struct option_table_struct *tx = NULL;
+  const char *arg = NULL;
+  /* Where the ARG was found in token_buffer. */
+  size_t arg_offset = 0;
 
-  c = getc (finput);
+  int c = getc (finput);
 
   switch (c)
     {
@@ -547,6 +491,8 @@ parse_percent_token (void)
     case '{':
       return tok_percent_left_curly;
 
+      /* FIXME: Who the heck are those 5 guys!?! `%<' = `%left'!!!
+	 Let's ask for there removal.  */
     case '<':
       return tok_left;
 
@@ -569,15 +515,40 @@ parse_percent_token (void)
   obstack_1grow (&token_obstack, '%');
   while (isalpha (c) || c == '_' || c == '-')
     {
-      if (c == '-')
-	c = '_';
+      if (c == '_')
+	c = '-';
       obstack_1grow (&token_obstack, c);
       c = getc (finput);
     }
 
-  ungetc (c, finput);
+  /* %DIRECTIVE="ARG".  Separate into
+     TOKEN_BUFFER = `%DIRECTIVE\0ARG\0'.
+     This is a bit hackish, but once we move to a Bison parser,
+     things will be cleaned up.  */
+  if (c == '=')
+    {
+      /* End of the directive.  We skip the `='. */
+      obstack_1grow (&token_obstack, '\0');
+      /* Fetch the ARG if present. */
+      c = getc (finput);
+      if (c == '"')
+	{
+	  int code;
+	  arg_offset = obstack_object_size (&token_obstack);
+	  /* Read up to and including `"'.  Do not append the closing
+	     `"' in the output: it's not part of the ARG.  */
+	  while ((code = literalchar ()) != '"')
+	    obstack_1grow (&token_obstack, code);
+	}
+      /* else: should be an error. */
+    }
+  else
+    ungetc (c, finput);
+
   obstack_1grow (&token_obstack, '\0');
   token_buffer = obstack_finish (&token_obstack);
+  if (arg_offset)
+    arg = token_buffer + arg_offset;
 
   /* table lookup % directive */
   for (tx = option_table; tx->name; tx++)
@@ -585,23 +556,42 @@ parse_percent_token (void)
 	&& option_strcmp (token_buffer + 1, tx->name) == 0)
       break;
 
-  if (tx->set_flag)
-    {
-      *((int *) (tx->set_flag)) = 1;
-      return tok_noop;
-    }
+  if (arg && tx->ret_val != tok_stropt)
+    fatal (_("`%s' supports no argument: %s"), token_buffer, quote (arg));
+
 
   switch (tx->ret_val)
     {
-    case tok_setopt:
-      *((char **) (tx->set_flag)) = optarg;
+    case tok_stropt:
+      assert (tx->set_flag);
+      if (arg)
+	{
+	  /* Keep only the first assignment: command line options have
+	     already been processed, and we want them to have
+	     precedence.  Side effect: if this %-option is used
+	     several times, only the first is honored.  Bah.  */
+	  if (!*((char **) (tx->set_flag)))
+	    *((char **) (tx->set_flag)) = xstrdup (arg);
+	}
+      else
+	fatal (_("`%s' requires an argument"), token_buffer);
+      return tok_noop;
+      break;
+
+    case tok_intopt:
+      assert (tx->set_flag);
+      *((int *) (tx->set_flag)) = 1;
       return tok_noop;
       break;
 
     case tok_obsolete:
       fatal (_("`%s' is no longer supported"), token_buffer);
+      return tok_noop;
       break;
-    }
 
-  return tx->ret_val;
+    default:
+      return tx->ret_val;
+      break;
+    }
+  abort ();
 }