cppfiles.c (read_and_prescan): Bump input pointer before possibly branching off to the backslash code.

	* cppfiles.c (read_and_prescan) [case SPECCASE_QUESTION]: Bump
	input pointer before possibly branching off to the backslash
	code.
	* cpphash.c (macroexpand): Correctly delete \r escapes when
	stringifying parameters.
	* cpplib.c (copy_rest_of_line): Go directly to skip_block_comment
	if we can; bail out early if we hit a line comment.
	(handle_directive): Treat '# 123' in an .S file just like
	'# <punctuation>'.  Discard the shifted '#' if we hit '#\n'.
	Return 1 for '# not_a_directive'.
	(get_directive_token): Pop macro buffers here, so that
	cpp_get_token can't sneakily move past a newline.
	Add sanity checks.
	(cpp_get_token): goto randomchar if handle_directive returns 0.

From-SVN: r31819

7 files changed, 157 insertions, 41 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index a4e9b70..e0894f9 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,5 +1,20 @@
 2000-02-05  Zack Weinberg  <zack@wolery.cumb.org>
 
+	* cppfiles.c (read_and_prescan) [case SPECCASE_QUESTION]: Bump
+	input pointer before possibly branching off to the backslash
+	code.
+	* cpphash.c (macroexpand): Correctly delete \r escapes when
+	stringifying parameters.
+	* cpplib.c (copy_rest_of_line): Go directly to skip_block_comment
+	if we can; bail out early if we hit a line comment.
+	(handle_directive): Treat '# 123' in an .S file just like
+	'# <punctuation>'.  Discard the shifted '#' if we hit '#\n'.
+	Return 1 for '# not_a_directive'.
+	(get_directive_token): Pop macro buffers here, so that
+	cpp_get_token can't sneakily move past a newline.
+	Add sanity checks. 
+	(cpp_get_token): goto randomchar if handle_directive returns 0.
+
 	* cppalloc.c: Update copyright.
 	* cpplib.c: Merge all the static function prototypes into one
 	block.
diff --git a/gcc/cppfiles.c b/gcc/cppfiles.c
index a89a0ca..36e3305 100644
--- a/gcc/cppfiles.c
+++ b/gcc/cppfiles.c
@@ -1017,6 +1017,7 @@ read_and_prescan (pfile, fp, desc, len)
 		    *op++ = '?';
 		    break;
 		  }
+		ip += 2;
 		if (CPP_OPTIONS (pfile)->warn_trigraphs)
 		  {
 		    unsigned long col;
@@ -1042,7 +1043,6 @@ read_and_prescan (pfile, fp, desc, len)
 		    *op++ = '?';
 		    *op++ = d;
 		  }
-		ip += 2;
 	      }
 	    }
 	}
diff --git a/gcc/cpphash.c b/gcc/cpphash.c
index c2244c9..64491c5 100644
--- a/gcc/cpphash.c
+++ b/gcc/cpphash.c
@@ -1158,18 +1158,17 @@ macroexpand (pfile, hp)
 
 		      if (!in_string)
 			{
+			  /* Delete "\r " and "\r-" escapes.  */
+			  if (c == '\r')
+			    {
+			      i++;
+			      continue;
+			    }
 			  /* Internal sequences of whitespace are
 			     replaced by one space except within
 			     a string or char token. */
-			  if (is_space(c))
+			  else if (is_space(c))
 			    {
-			      if (CPP_WRITTEN (pfile) > (unsigned) arg->stringified
-				  && (CPP_PWRITTEN (pfile))[-1] == '\r')
-				{
-				  /* "\r " escape markers are removed */
-				  CPP_ADJUST_WRITTEN (pfile, -1);
-				  continue;
-				}
 			      if (need_space == 0)
 				need_space = 1;
 			      continue;
diff --git a/gcc/cpplib.c b/gcc/cpplib.c
index 409eb08..40c3c3a 100644
--- a/gcc/cpplib.c
+++ b/gcc/cpplib.c
@@ -468,15 +468,18 @@ copy_rest_of_line (pfile)
 	  parse_string (pfile, c);
 	  continue;
 	case '/':
-	  if (PEEKC() == '*' && CPP_TRADITIONAL (pfile))
+	  if (PEEKC() == '*')
 	    {
-	      CPP_PUTS (pfile, "/**/", 4);
-	      skip_comment (pfile, c);
+	      if (CPP_TRADITIONAL (pfile))
+		CPP_PUTS (pfile, "/**/", 4);
+	      skip_block_comment (pfile);
 	      continue;
 	    }
 	  /* else fall through */
 	case '-':
 	  c = skip_comment (pfile, c);
+	  if (c == ' ')
+	    return;
 	  break;
 
 	case '\f':
@@ -527,10 +530,7 @@ handle_directive (pfile)
   if (c >= '0' && c <= '9')
     {
       if (CPP_OPTIONS (pfile)->lang_asm)
-	{
-	  skip_rest_of_line (pfile);
-	  return 1;
-	}
+	return 0;
 
       if (CPP_PEDANTIC (pfile)
 	  && ! CPP_PREPROCESSED (pfile)
@@ -552,21 +552,20 @@ handle_directive (pfile)
   ident_length = CPP_PWRITTEN (pfile) - ident;
   if (ident_length == 0)
     {
-      /* A line of just `#' becomes blank.  */
-      if (PEEKC() == '\n')
-	return 1;
-      else
-	return 0;
+      /* A line of just `#' becomes blank.  A line with something
+         other than an identifier after the # is reparsed as a non-
+         directive line.  */
+      CPP_SET_WRITTEN (pfile, old_written);
+      return (PEEKC() == '\n');
     }
 
-  /*
-   * Decode the keyword and call the appropriate expansion
-   * routine, after moving the input pointer up to the next line.
-   */
+  /* Decode the keyword and call the appropriate expansion routine.  */
   for (kt = directive_table; ; kt++)
     {
       if (kt->length <= 0)
-	return 0;
+	/* # identifier, but not a legit directive.  Pass onward as a
+	   CPP_DIRECTIVE token anyway - let the consumer worry about it.  */
+	return 1;
       if (kt->length == ident_length
 	  && !strncmp (kt->name, ident, ident_length)) 
 	break;
@@ -983,26 +982,43 @@ static enum cpp_token
 get_directive_token (pfile)
      cpp_reader *pfile;
 {
+  long old_written = CPP_WRITTEN (pfile);
+  enum cpp_token token;
+
   for (;;)
     {
-      long old_written = CPP_WRITTEN (pfile);
-      enum cpp_token token;
       cpp_skip_hspace (pfile);
       if (PEEKC () == '\n')
-	  return CPP_VSPACE;
+	return CPP_VSPACE;
+
       token = cpp_get_token (pfile);
-      switch (token)
-      {
-      case CPP_POP:
-	  if (! CPP_IS_MACRO_BUFFER (CPP_BUFFER (pfile)))
-	      return token;
-	  /* ... else fall though ...  */
-      case CPP_HSPACE:  case CPP_COMMENT:
+      /* token could be hspace at the beginning of a macro.  */
+      if (token == CPP_HSPACE || token == CPP_COMMENT)
+	{
 	  CPP_SET_WRITTEN (pfile, old_written);
-	  break;
-      default:
+	  continue;
+	}
+
+      /* token cannot be vspace, it would have been caught above.  */
+      if (token == CPP_VSPACE)
+	{
+	  cpp_fatal (pfile, "VSPACE in get_directive_token");
 	  return token;
-      }
+	}
+
+      /* token cannot be POP unless the buffer is a macro buffer.  */
+      if (token != CPP_POP)
+	return token;
+
+      if (! CPP_IS_MACRO_BUFFER (CPP_BUFFER (pfile)))
+	{
+	  cpp_fatal (pfile, "POP of file buffer in get_directive_token");
+	  return token;
+	}
+
+      /* We must pop the buffer by hand, else cpp_get_token might hand
+	 us whitespace or newline on the next invocation.  */
+      cpp_pop_buffer (pfile);
     }
 }
 
@@ -2246,7 +2262,7 @@ if_directive_name (pfile, ifs)
 
 /* Get the next token, and add it to the text in pfile->token_buffer.
    Return the kind of token we got.  */
-  
+
 enum cpp_token
 cpp_get_token (pfile)
      cpp_reader *pfile;
@@ -2345,7 +2361,7 @@ cpp_get_token (pfile)
 	  if (handle_directive (pfile))
 	    return CPP_DIRECTIVE;
 	  pfile->only_seen_white = 0;
-	  return CPP_OTHER;
+	  goto randomchar;
 
 	case '\"':
 	case '\'':
diff --git a/gcc/testsuite/gcc.dg/lineno-2.c b/gcc/testsuite/gcc.dg/lineno-2.c
new file mode 100644
index 0000000..64647e2
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/lineno-2.c
@@ -0,0 +1,24 @@
+/* { dg-do run } */
+
+/* Test #line with and without macros for the line number.  */
+
+#define L 90
+
+#line 44
+int i = __LINE__;
+
+#line L
+int j = __LINE__;
+
+#line 14  /* N.B. the _next_ line is line 14.  */
+
+int main(void)
+{
+  if (i != 44)
+    abort ();
+  if (j != 90)
+    abort ();
+  if (__LINE__ != 21)
+    abort ();
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.dg/strpaste.c b/gcc/testsuite/gcc.dg/strpaste.c
new file mode 100644
index 0000000..56e1484
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/strpaste.c
@@ -0,0 +1,34 @@
+/* { dg-do run } */
+
+/* Regression test for stringizing and token pasting.
+   We got internal escape markers in the strings.  */
+
+#include <string.h>
+
+#define S(x) _S(x)
+#define _S(x) #x
+
+#define I 1
+static const char s1[] = S(I.1);
+static const char t1[] = "1.1";
+
+#define f h
+#define h(a) a+f
+static const char s2[] = S( f(1)(2) );
+static const char t2[] = "1+h(2)";
+
+#undef I
+#undef f
+#undef h
+
+int
+main(void)
+{
+  if (strcmp (s1, t1))
+    abort ();
+
+  if (strcmp (s2, t2))
+    abort ();
+
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.dg/trigraphs.c b/gcc/testsuite/gcc.dg/trigraphs.c
new file mode 100644
index 0000000..62ac103
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/trigraphs.c
@@ -0,0 +1,28 @@
+/* { dg-do run } */
+/* { dg-options "-ansi" } */
+
+/* Basic tests for trigraph conversion.
+   All of them are here, but not in all possible contexts.  *??/
+/
+
+??=include <stdio.h>
+
+??=define TWELVE 1??/
+2
+
+static const char str??(??) = "0123456789??/n";
+
+int
+main(void)
+??<
+  if (sizeof str != TWELVE)
+    abort ();
+
+  if ((5 ??' 3) != 6)
+    abort ();
+
+  if ((5 ??! 3) != 7)
+    abort ();
+
+  return 0;
+??>