cpphash.c (collect_expansion, [...]): Make the escape character in macro buffers '\r', not '@'.

1999-04-12 14:55 -0400  Zack Weinberg  <zack@rabi.columbia.edu>
	* cpphash.c (collect_expansion, macroexpand,
	push_macro_expansion): Make the escape character in macro
	buffers '\r', not '@'.  Remove code to protect literal
	occurences of the escape character; '\r' cannot appear
	in a macro buffer unless we put it there.
	* cpplib.c (skip_comment, copy_comment, cpp_skip_hspace,
	copy_rest_of_line, cpp_get_token, parse_string,
	parse_assertion): '\r' might be a backslash-newline marker, or
	it might be a macro escape marker, depending on
	CPP_BUFFER (pfile)->has_escapes.  '@' is not a special
	character.
	* cpplib.h: Update commentary.

From-SVN: r26371

4 files changed, 125 insertions, 103 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 8f086b6..148fb57 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,18 @@
+1999-04-12 14:55 -0400  Zack Weinberg  <zack@rabi.columbia.edu>
+
+	* cpphash.c (collect_expansion, macroexpand,
+	push_macro_expansion): Make the escape character in macro
+	buffers '\r', not '@'.  Remove code to protect literal
+	occurences of the escape character; '\r' cannot appear
+	in a macro buffer unless we put it there.
+	* cpplib.c (skip_comment, copy_comment, cpp_skip_hspace,
+	copy_rest_of_line, cpp_get_token, parse_string,
+	parse_assertion): '\r' might be a backslash-newline marker, or
+	it might be a macro escape marker, depending on
+	CPP_BUFFER (pfile)->has_escapes.  '@' is not a special
+	character.
+	* cpplib.h: Update commentary.
+
 Mon Apr 12 09:30:03 1999  Richard Earnshaw (rearnsha@arm.com)
 
 	* arm.h (target_fp_name, structure_size_string, arm_cpu_select):
diff --git a/gcc/cpphash.c b/gcc/cpphash.c
index 5c2d5f3..06de9d4 100644
--- a/gcc/cpphash.c
+++ b/gcc/cpphash.c
@@ -304,10 +304,6 @@ collect_expansion (pfile, buf, limit, nargs, arglist)
      leading and trailing newline-marker and final null.  */
   maxsize = (sizeof (DEFINITION)
 	     + (limit - p) + 5);
-  /* Occurrences of '@' get doubled, so allocate extra space for them.  */
-  while (p < limit)
-    if (*p++ == '@')
-      maxsize++;
   defn = (DEFINITION *) xcalloc (1, maxsize);
 
   defn->nargs = nargs;
@@ -318,7 +314,7 @@ collect_expansion (pfile, buf, limit, nargs, arglist)
 
   /* Add one initial space escape-marker to prevent accidental
      token-pasting (often removed by macroexpand).  */
-  *exp_p++ = '@';
+  *exp_p++ = '\r';
   *exp_p++ = ' ';
 
   if (limit - p >= 2 && p[0] == '#' && p[1] == '#')
@@ -359,13 +355,6 @@ collect_expansion (pfile, buf, limit, nargs, arglist)
 		}
 	      break;
 
-	    case '@':
-	      /* An '@' in a string or character constant stands for itself,
-	         and does not need to be escaped.  */
-	      if (!expected_delimiter)
-		*exp_p++ = c;
-	      break;
-
 	    case '#':
 	      /* # is ordinary inside a string.  */
 	      if (expected_delimiter)
@@ -548,10 +537,10 @@ collect_expansion (pfile, buf, limit, nargs, arglist)
 
   if (!CPP_TRADITIONAL (pfile) && expected_delimiter == 0)
     {
-      /* If ANSI, put in a "@ " marker to prevent token pasting.
+      /* If ANSI, put in a "\r " marker to prevent token pasting.
          But not if "inside a string" (which in ANSI mode
          happens only for -D option).  */
-      *exp_p++ = '@';
+      *exp_p++ = '\r';
       *exp_p++ = ' ';
     }
 
@@ -965,7 +954,7 @@ special_symbol (hp, pfile)
       if (!buf)
 	return;
       if (*buf == '\0')
-	buf = "@ ";
+	buf = "\r ";
 
       len = strlen (buf);
       CPP_RESERVE (pfile, len + 1);
@@ -1224,9 +1213,9 @@ macroexpand (pfile, hp)
 			  if (is_space[c])
 			    {
 			      if (CPP_WRITTEN (pfile) > (unsigned) arg->stringified
-				  && (CPP_PWRITTEN (pfile))[-1] == '@')
+				  && (CPP_PWRITTEN (pfile))[-1] == '\r')
 				{
-				  /* "@ " escape markers are removed */
+				  /* "\r " escape markers are removed */
 				  CPP_ADJUST_WRITTEN (pfile, -1);
 				  continue;
 				}
@@ -1360,27 +1349,11 @@ macroexpand (pfile, hp)
 		    {
 		      if (is_space[l1[-1]])
 			l1--;
-		      else if (l1[-1] == '@')
-			{
-			  U_CHAR *p2 = l1 - 1;
-			  /* If whitespace is preceded by an odd number
-			     of `@' signs, the last `@' was a whitespace
-			     marker; drop it too. */
-			  while (p2 != p1 && p2[0] == '@')
-			    p2--;
-			  if ((l1 - p2) & 1)
-			    l1--;
-			  break;
-			}
+		      else if (l1[-1] == '\r')
+			l1--;
 		      else if (l1[-1] == '-')
 			{
-			  U_CHAR *p2 = l1 - 1;
-			  /* If a `-' is preceded by an odd number of
-			     `@' signs then it and the last `@' are
-			     a no-reexpansion marker.  */
-			  while (p2 != p1 && p2[0] == '@')
-			    p2--;
-			  if ((l1 - p2) & 1)
+			  if (l1 != p1 + 1 && l1[-2] == '\r')
 			    l1 -= 2;
 			  else
 			    break;
@@ -1392,7 +1365,7 @@ macroexpand (pfile, hp)
 
 	      /* Delete any no-reexpansion marker that precedes
 	         an identifier at the beginning of the argument. */
-	      if (p1[0] == '@' && p1[1] == '-')
+	      if (p1[0] == '\r' && p1[1] == '-')
 		p1 += 2;
 
 	      bcopy (p1, xbuf + totlen, l1 - p1);
@@ -1405,7 +1378,7 @@ macroexpand (pfile, hp)
 		  && !CPP_TRADITIONAL (pfile)
 		  && unsafe_chars (xbuf[totlen - 1], expanded[0]))
 		{
-		  xbuf[totlen++] = '@';
+		  xbuf[totlen++] = '\r';
 		  xbuf[totlen++] = ' ';
 		}
 
@@ -1416,7 +1389,7 @@ macroexpand (pfile, hp)
 		  && !CPP_TRADITIONAL (pfile)
 		  && unsafe_chars (xbuf[totlen - 1], exp[offset]))
 		{
-		  xbuf[totlen++] = '@';
+		  xbuf[totlen++] = '\r';
 		  xbuf[totlen++] = ' ';
 		}
 
@@ -1533,7 +1506,7 @@ push_macro_expansion (pfile, xbuf, xbuf_len, hp)
   mbuf->cleanup = macro_cleanup;
   mbuf->data = hp;
 
-  /* The first chars of the expansion should be a "@ " added by
+  /* The first chars of the expansion should be a "\r " added by
      collect_expansion.  This is to prevent accidental token-pasting
      between the text preceding the macro invocation, and the macro
      expansion text.
@@ -1551,7 +1524,7 @@ push_macro_expansion (pfile, xbuf, xbuf_len, hp)
      Also, we don't need the extra space if the first char is '(',
      or some other (less common) characters.  */
 
-  if (xbuf[0] == '@' && xbuf[1] == ' '
+  if (xbuf[0] == '\r' && xbuf[1] == ' '
       && (is_idchar[xbuf[2]] || xbuf[2] == '(' || xbuf[2] == '\''
 	  || xbuf[2] == '\"'))
     mbuf->cur += 2;
@@ -1560,7 +1533,7 @@ push_macro_expansion (pfile, xbuf, xbuf_len, hp)
      if this is safe.  We can do a better job here since we can know
      what the next char will be.  */
   if (xbuf_len >= 3
-      && mbuf->rlimit[-2] == '@'
+      && mbuf->rlimit[-2] == '\r'
       && mbuf->rlimit[-1] == ' ')
     {
       int c1 = mbuf->rlimit[-3];
diff --git a/gcc/cpplib.c b/gcc/cpplib.c
index d508486..3ed75db 100644
--- a/gcc/cpplib.c
+++ b/gcc/cpplib.c
@@ -260,6 +260,7 @@ skip_comment (pfile, m)
 	      return EOF;
 	    }
 	  else if (c == '\n' || c == '\r')
+	    /* \r cannot be a macro escape marker here. */
 	    CPP_BUMP_LINE (pfile);
 	  else if (c == '/' && prev_c == '*')
 	    return ' ';
@@ -288,6 +289,7 @@ skip_comment (pfile, m)
 		  return ' ';
 		}
 	      else if (c == '\r')
+		/* \r cannot be a macro escape marker here. */
 		CPP_BUMP_LINE (pfile);
 	}
     }
@@ -323,6 +325,7 @@ copy_comment (pfile, m)
 	    }
 	  else if (c == '\r')
 	    {
+	      /* \r cannot be a macro escape marker here. */
 	      CPP_BUMP_LINE (pfile);
 	      continue;
 	    }
@@ -362,6 +365,7 @@ copy_comment (pfile, m)
 	      return ' ';
 	    }
 	  else if (c == '\r')
+	    /* \r cannot be a macro escape marker here. */
 	    CPP_BUMP_LINE (pfile);
 
 	  CPP_PUTC (pfile, c);
@@ -392,7 +396,17 @@ cpp_skip_hspace (pfile)
 	}
       else if (c == '\r')
 	{
-	  CPP_BUFFER (pfile)->lineno++;
+	  /* \r is a backslash-newline marker if !has_escapes, and
+	     a deletable-whitespace or no-reexpansion marker otherwise. */
+	  if (CPP_BUFFER (pfile)->has_escapes)
+	    {
+	      if (PEEKC() == ' ')
+		FORWARD(1);
+	      else
+		break;
+	    }
+	  else
+	    CPP_BUFFER (pfile)->lineno++;
 	}
       else if (c == '/' || c == '-')
 	{
@@ -400,22 +414,12 @@ cpp_skip_hspace (pfile)
 	  if (c == EOF)
 	    return;
 	  else if (c != ' ')
-	    {
-	      FORWARD(-1);
-	      return;
-	    }
-	}
-      else if (c == '@' && CPP_BUFFER (pfile)->has_escapes
-	       && PEEKC() == ' ')
-	{
-	  FORWARD(1);
+	    break;
 	}
       else
-	{
-	  FORWARD(-1);
-	  return;
-	}
+	break;
     }
+  FORWARD(-1);
 }
 
 /* Read the rest of the current line.
@@ -437,8 +441,13 @@ copy_rest_of_line (pfile)
 	  return;
 
 	case '\r':
-	  CPP_BUFFER (pfile)->lineno++;
-	  continue;
+	  if (CPP_BUFFER (pfile)->has_escapes)
+	    break;
+	  else
+	    {
+	      CPP_BUFFER (pfile)->lineno++;
+	      continue;
+	    }
 	case '\'':
 	case '\"':
 	  parse_string (pfile, c);
@@ -2233,9 +2242,23 @@ cpp_get_token (pfile)
 		    }
 		  else if (c == '\r')
 		    {
-		      /* Backslash newline is replaced by nothing. */
-		      CPP_ADJUST_WRITTEN (pfile, -1);
-		      CPP_BUMP_LINE (pfile);
+		      if (!CPP_BUFFER (pfile)->has_escapes)
+			{
+			  /* Backslash newline is replaced by nothing. */
+			  CPP_ADJUST_WRITTEN (pfile, -1);
+			  CPP_BUMP_LINE (pfile);
+			}
+		      else
+			{
+			  /* We might conceivably get \r- or \r<space> in
+			     here.  Just delete 'em. */
+			  int d = GETC();
+			  if (d != '-' && d != ' ')
+			    cpp_fatal (pfile,
+				  "internal error: unrecognized escape \\r%c",
+				       d);
+			  CPP_ADJUST_WRITTEN (pfile, -1);
+			}			  
 		    }
 		}
 	      return CPP_STRING;
@@ -2258,33 +2281,6 @@ cpp_get_token (pfile)
 	  pfile->only_seen_white = 0;
 	  return CPP_OTHER;
 
-	case '@':
-	  if (CPP_BUFFER (pfile)->has_escapes)
-	    {
-	      c = GETC ();
-	      if (c == '-')
-		{
-		  if (pfile->output_escapes)
-		    CPP_PUTS (pfile, "@-", 2);
-		  parse_name (pfile, GETC ());
-		  return CPP_NAME;
-		}
-	      else if (c == ' ')
-		{
-		  CPP_RESERVE (pfile, 2);
-		  if (pfile->output_escapes)
-		    CPP_PUTC_Q (pfile, '@');
-		  CPP_PUTC_Q (pfile, c);
-		  return CPP_HSPACE;
-		}
-	    }
-	  if (pfile->output_escapes)
-	    {
-	      CPP_PUTS (pfile, "@@", 2);
-	      return CPP_OTHER;
-	    }
-	  goto randomchar;
-
 	case '.':
 	  c2 = PEEKC ();
 	  if (ISDIGIT(c2))
@@ -2410,13 +2406,13 @@ cpp_get_token (pfile)
 	    if (hp->type == T_DISABLED)
 	      {
 		if (pfile->output_escapes)
-		  { /* Return "@-IDENT", followed by '\0'.  */
+		  { /* Return "\r-IDENT", followed by '\0'.  */
 		    int i;
 		    CPP_RESERVE (pfile, 3);
 		    ident = pfile->token_buffer + before_name_written;
 		    CPP_ADJUST_WRITTEN (pfile, 2);
 		    for (i = ident_len; i >= 0; i--) ident[i+2] = ident[i];
-		    ident[0] = '@';
+		    ident[0] = '\r';
 		    ident[1] = '-';
 		  }
 		return CPP_NAME;
@@ -2490,13 +2486,38 @@ cpp_get_token (pfile)
 	    }
 	  return CPP_HSPACE;
 
-        case '\\':
-	  goto randomchar;
-
 	case '\r':
-	  /* Backslash newline is ignored. */
-	  CPP_BUMP_LINE (pfile);
-	  goto get_next;
+	  if (CPP_BUFFER (pfile)->has_escapes)
+	    {
+	      c = GETC ();
+	      if (c == '-')
+		{
+		  if (pfile->output_escapes)
+		    CPP_PUTS (pfile, "\r-", 2);
+		  parse_name (pfile, GETC ());
+		  return CPP_NAME;
+		}
+	      else if (c == ' ')
+		{
+		  CPP_RESERVE (pfile, 2);
+		  if (pfile->output_escapes)
+		    CPP_PUTC_Q (pfile, '\r');
+		  CPP_PUTC_Q (pfile, c);
+		  return CPP_HSPACE;
+		}
+	      else
+		{
+		  cpp_fatal (pfile,
+			     "internal error: unrecognized escape \\r%c", c);
+		  goto get_next;
+		}
+	    }
+	  else
+	    {
+	      /* Backslash newline is ignored. */
+	      CPP_BUMP_LINE (pfile);
+	      goto get_next;
+	    }
 
 	case '\n':
 	  CPP_PUTC (pfile, c);
@@ -2639,9 +2660,16 @@ parse_string (pfile, c)
 	  break;
 
 	case '\r':
-	  /* Backslash newline is replaced by nothing at all.  */
 	  CPP_ADJUST_WRITTEN (pfile, -1);
-	  CPP_BUMP_LINE (pfile);
+	  if (CPP_BUFFER (pfile)->has_escapes)
+	    {
+	      cpp_fatal (pfile,
+			 "internal error: \\r escape inside string constant");
+	      FORWARD(1);
+	    }
+	  else
+	    /* Backslash newline is replaced by nothing at all.  */
+	    CPP_BUMP_LINE (pfile);
 	  break;
 
 	case '\\':
@@ -2711,6 +2739,7 @@ parse_assertion (pfile)
 	  return 0;
 	}
       else if (c == '\r')
+	/* \r cannot be a macro escape here. */
 	CPP_BUMP_LINE (pfile);
       else
 	{
diff --git a/gcc/cpplib.h b/gcc/cpplib.h
index ccf8f92..bffd493 100644
--- a/gcc/cpplib.h
+++ b/gcc/cpplib.h
@@ -124,12 +124,17 @@ struct cpp_buffer
   char seen_eof;
 
   /* True if buffer contains escape sequences.
-     Currently there are three kinds:
-     "@-" means following identifier should not be macro-expanded.
-     "@ " means a token-separator.  This turns into " " in final output
+     Currently there are two kinds:
+     "\r-" means following identifier should not be macro-expanded.
+     "\r " means a token-separator.  This turns into " " in final output
           if not stringizing and needed to separate tokens; otherwise nothing.
-     "@@" means a normal '@'.
-     (An '@' inside a string stands for itself and is never an escape.) */
+     Any other two-character sequence beginning with \r is an error.
+
+     If this is NOT set, then \r is a one-character escape meaning backslash
+     newline.  This is guaranteed not to occur in the middle of a token.
+     The two interpretations of \r do not conflict, because the two-character
+     escapes are used only in macro buffers, and backslash-newline is removed
+     from macro expansion text in collect_expansion and/or macarg.  */
   char has_escapes;
 };