Correct handling of `position' levels with no non-IGNOREd element and

handling of NUL byte.

3 files changed, 45 insertions, 27 deletions

diff --git a/locale/weight.h b/locale/weight.h
index 424e493209..ef767bac59 100644
--- a/locale/weight.h
+++ b/locale/weight.h
@@ -55,8 +55,8 @@ typedef struct weight_t
   ((u_int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_RULES))
 
 
-static __inline int get_weight (const STRING_TYPE **str, weight_t *result);
-static __inline int
+static __inline void get_weight (const STRING_TYPE **str, weight_t *result);
+static __inline void
 get_weight (const STRING_TYPE **str, weight_t *result)
 {
   unsigned int ch = *((USTRING_TYPE *) (*str))++;
@@ -86,7 +86,7 @@ get_weight (const STRING_TYPE **str, weight_t *result)
 		  result->data[cnt].value = &__collate_extra[idx];
 		  idx += result->data[cnt].number;
 		}
-	      return 0;
+	      return;
 	    }
 	  slot += level_size;
 	}
@@ -102,7 +102,7 @@ get_weight (const STRING_TYPE **str, weight_t *result)
 	  result->data[cnt].number = 1;
 	  result->data[cnt].value = &__collate_table[slot + 1 + cnt];
 	}
-      return ch == 0;
+      return;
     }
 
   /* We now look for any collation element which starts with CH.
@@ -135,14 +135,12 @@ get_weight (const STRING_TYPE **str, weight_t *result)
 	      result->data[cnt].value = &__collate_extra[idx];
 	      idx += result->data[cnt].number;
 	    }
-	  return 0;
+	  return;
 	}
 
       /* To next entry in list.  */
       slot += __collate_extra[slot];
     }
-  /* NOTREACHED */
-  return 0;	/* To calm down gcc.  */
 }
 
 
@@ -155,7 +153,7 @@ get_weight (const STRING_TYPE **str, weight_t *result)
   do									      \
     {									      \
       weight_t *newp;							      \
-      do								      \
+      while (*str != '\0')						      \
 	{								      \
 	  newp = (weight_t *) alloca (sizeof (weight_t)			      \
 				      + (collate_nrules			      \
@@ -168,7 +166,7 @@ get_weight (const STRING_TYPE **str, weight_t *result)
 	    backw->next = newp;						      \
 	  newp->next = NULL;						      \
 	  backw = newp;							      \
+	  get_weight (&str, newp);					      \
 	}								      \
-      while (get_weight (&str, newp) == 0);				      \
     }									      \
   while (0)
diff --git a/string/strcoll.c b/string/strcoll.c
index aae5944771..c147e51f4b 100644
--- a/string/strcoll.c
+++ b/string/strcoll.c
@@ -67,16 +67,18 @@ STRCOLL (s1, s2)
       int s1idx = forward ? 0 : s1run->data[pass].number - 1;
       int s2idx = forward ? 0 : s2run->data[pass].number - 1;
 
-      do
+      while (1)
 	{
 	  int s1ignore = 0;
 	  int s2ignore = 0;
-	  u_int32_t w1, w2;
+	  u_int32_t w1 = 0;
+	  u_int32_t w2 = 0;
 
 	  /* Here we have to check for IGNORE entries.  If these are
 	     found we count them and go on with the next value.  */
-	  while ((w1 = s1run->data[pass].value[s1idx])
-		 == (u_int32_t) IGNORE_CHAR)
+	  while (s1run != NULL
+		 && ((w1 = s1run->data[pass].value[s1idx])
+		     == (u_int32_t) IGNORE_CHAR))
 	    {
 	      ++s1ignore;
 	      if ((forward && ++s1idx >= s1run->data[pass].number)
@@ -86,15 +88,19 @@ STRCOLL (s1, s2)
 		  if (nextp == NULL)
 		    {
 		      w1 = 0;
-		      break;
+		      /* No more non-INGOREd elements means lowest
+			 possible value.  */
+		      s1ignore = -1;
 		    }
+		  else
+		    s1idx = forward ? 0 : nextp->data[pass].number - 1;
 		  s1run = nextp;
-		  s1idx = forward ? 0 : s1run->data[pass].number - 1;
 		}
 	    }
 
-	  while ((w2 = s2run->data[pass].value[s2idx])
-		 == (u_int32_t) IGNORE_CHAR)
+	  while (s2run != NULL
+		 && ((w2 = s2run->data[pass].value[s2idx])
+		     == (u_int32_t) IGNORE_CHAR))
 	    {
 	      ++s2ignore;
 	      if ((forward && ++s2idx >= s2run->data[pass].number)
@@ -104,13 +110,20 @@ STRCOLL (s1, s2)
 		  if (nextp == NULL)
 		    {
 		      w2 = 0;
-		      break;
+		      /* No more non-INGOREd elements means lowest
+			 possible value.  */
+		      s2ignore = -1;
 		    }
+		  else
+		    s2idx = forward ? 0 : nextp->data[pass].number - 1;
 		  s2run = nextp;
-		  s2idx = forward ? 0 : s2run->data[pass].number - 1;
 		}
 	    }
 
+	  /* If one string is completely processed stop.  */
+	  if (s1run == NULL || s2run == NULL)
+	    break;
+
 	  /* Now we have information of the number of ignored
 	     weights and the value of the next weight.  */
 	  if ((collate_rules[pass] & sort_position) != 0
@@ -150,7 +163,6 @@ STRCOLL (s1, s2)
 	      }
 
 	}
-      while (s1run != NULL && s2run != NULL);
 
       if (s1run != s2run)
 	return s1run != NULL ? 1 : -1;
diff --git a/string/strxfrm.c b/string/strxfrm.c
index 820ad93947..4076ebb920 100644
--- a/string/strxfrm.c
+++ b/string/strxfrm.c
@@ -1,4 +1,4 @@
-/* Copyright (C) 1995, 1996 Free Software Foundation, Inc.
+/* Copyright (C) 1995, 1996, 1997 Free Software Foundation, Inc.
    This file is part of the GNU C Library.
    Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995.
 
@@ -172,14 +172,16 @@ STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n)
       const weight_t *run = forward ? forw : backw;
       int idx = forward ? 0 : run->data[pass].number - 1;
 
-      do
+      while (1)
 	{
 	  int ignore = 0;
-	  u_int32_t w;
+	  u_int32_t w = 0;
 
 	  /* Here we have to check for IGNORE entries.  If these are
 	     found we count them and go on with he next value.  */
-	  while ((w = run->data[pass].value[idx]) == (u_int32_t) IGNORE_CHAR)
+	  while (run != NULL
+		 && ((w = run->data[pass].value[idx])
+		     == (u_int32_t) IGNORE_CHAR))
 	    {
 	      ++ignore;
 	      if ((forward && ++idx >= run->data[pass].number)
@@ -189,13 +191,20 @@ STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n)
 		  if (nextp == NULL)
 		    {
 		      w = 0;
-		      break;
+		      /* No more non-INGOREd elements means lowest
+			possible value.  */
+		      ignore = -1;
 		    }
+		  else
+		    idx = forward ? 0 : nextp->data[pass].number - 1;
 		  run = nextp;
-		  idx = forward ? 0 : run->data[pass].number - 1;
 		}
 	    }
 
+	  /* Stop if all characters are processed.  */
+	  if (run == NULL)
+	    break;
+
 	  /* Now we have information of the number of ignored weights
 	     and the value of the next weight.  We have to add 2
 	     because 0 means EOS and 1 is the intermediate string end.  */
@@ -220,7 +229,6 @@ STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n)
 		  idx = run->data[pass].number - 1;
 	      }
 	}
-      while (run != NULL);
 
       /* Write marker for end of word.  */
       if (pass + 1 < collate_nrules)