Re: Bison 1.875 coredump

[Paul Eggert]

Paul Govereau <address@hidden> writes:

> I have done a little bit of investigation, and it seems like
> the problem is parse-gram.y calling into reader.c with an invalid
> pointer (loc.start.file). I have attached a stack trace below. Also, I
> noticed that if I run bison with "--trace=scan" or "--trace=parse" it
> does not crash, and seems to work fine. Once again, my inputs are:
> 
> %% A : B %%      --> core dump
> %% A : B; %%     --> works fine
> %% A : B         --> works fine

Thanks for the details.  On further investigation, I think this may
have been fixed in the latest CVS version of Bison.  Please try the
following cumulative patch relative to Bison 1.875a.

2003-03-12  Paul Eggert  <address@hidden>

        * scan-gram.l (YY_USER_INIT): Initialize code_start, too.
        (<INITIAL><<EOF>>, <SC_PRE_CODE><<EOF>>): Set *loc to the scanner
        cursor, instead of leaving it undefined.  This fixes a bug
        reported by Tim Van Holder in
        <http://mail.gnu.org/archive/html/bug-bison/2003-03/msg00023.html>.

2003-03-01  Paul Eggert  <address@hidden>

        * src/scan-gram.l (code_start): Initialize it to scanner_cursor,
        not loc->end, since loc->end might contain garbage and this leads
        to undefined behavior on some platforms.
        (id_loc, token_start): Use (IF_LINTed) initial values that do not
        depend on *loc, so that the reader doesn't give the the false
        impression that *loc is initialized.
        (<INITIAL>"%%"): Do not bother setting code_start, since its value
        does not survive the return.

2003-03-01  Akim Demaille  <address@hidden>

        * src/scan-gram.l (code_start): Always initialize it when entering
        into yylex, as SC_EPILOGUE is activated *before* the corresponding
        yylex invocation.  An alternative would be making it static, but
        then it starts with the second %%'s beginning, instead of its end.

2003-02-03  Paul Eggert  <address@hidden>

        * src/scan-gram.l (no_cr_read): Use bytes_read, not s, for size
        var.

Index: scan-gram.l
===================================================================
RCS file: /cvsroot/bison/bison/src/scan-gram.l,v
retrieving revision 1.54
retrieving revision 1.58
diff -p -u -r1.54 -r1.58
--- scan-gram.l 31 Dec 2002 02:26:51 -0000      1.54
+++ scan-gram.l 13 Mar 2003 07:07:17 -0000      1.58
@@ -1,6 +1,6 @@
 /* Bison Grammar Scanner                             -*- C -*-
 
-   Copyright (C) 2002 Free Software Foundation, Inc.
+   Copyright (C) 2002, 2003 Free Software Foundation, Inc.
 
    This file is part of Bison, the GNU Compiler Compiler.
 
@@ -43,6 +43,7 @@
       scanner_cursor.file = current_file;              \
       scanner_cursor.line = 1;                         \
       scanner_cursor.column = 1;                       \
+      code_start = scanner_cursor;                     \
     }                                                  \
   while (0)
 
@@ -136,14 +137,16 @@ splice     (\\[ \f\t\v]*\n)*
   int token_type IF_LINT (= 0);
 
   /* Location of most recent identifier, when applicable.  */
-  location id_loc IF_LINT (= *loc);
+  location id_loc IF_LINT (= empty_location);
 
-  /* Where containing code started, when applicable.  */
-  boundary code_start IF_LINT (= loc->start);
+  /* Where containing code started, when applicable.  Its initial
+     value is relevant only when yylex is invoked in the SC_EPILOGUE
+     start condition.  */
+  boundary code_start = scanner_cursor;
 
   /* Where containing comment or string or character literal started,
      when applicable.  */
-  boundary token_start IF_LINT (= loc->start);
+  boundary token_start IF_LINT (= scanner_cursor);
 %}
 
 
@@ -270,16 +273,18 @@ splice     (\\[ \f\t\v]*\n)*
   "%%" {
     static int percent_percent_count;
     if (++percent_percent_count == 2)
-      {
-       code_start = loc->start;
-       BEGIN SC_EPILOGUE;
-      }
+      BEGIN SC_EPILOGUE;
     return PERCENT_PERCENT;
   }
 
   . {
     complain_at (*loc, _("invalid character: %s"), quote (yytext));
   }
+
+  <<EOF>> {
+    loc->start = loc->end = scanner_cursor;
+    yyterminate ();
+  }
 }
 
 
@@ -530,6 +535,8 @@ splice       (\\[ \f\t\v]*\n)*
        return token_type;
       }
   }
+
+  <<EOF>>  unexpected_end_of_file (scanner_cursor, "{}");
 }
 
 
@@ -682,14 +689,14 @@ adjust_location (location *loc, char con
 static size_t
 no_cr_read (FILE *fp, char *buf, size_t size)
 {
-  size_t s = fread (buf, 1, size, fp);
-  if (s)
+  size_t bytes_read = fread (buf, 1, size, fp);
+  if (bytes_read)
     {
-      char *w = memchr (buf, '\r', s);
+      char *w = memchr (buf, '\r', bytes_read);
       if (w)
        {
          char const *r = ++w;
-         char const *lim = buf + s;
+         char const *lim = buf + bytes_read;
 
          for (;;)
            {
@@ -718,7 +725,7 @@ no_cr_read (FILE *fp, char *buf, size_t 
        }
     }
 
-  return s;
+  return bytes_read;
 }