builtins.h (c_strlen_data): Add new fields and comments.

	* builtins.h (c_strlen_data): Add new fields and comments.
	* builtins.c (unterminated_array): Change field reference from
	"len" to "minlen" in c_strlen_data instance.
	* gimple-fold.c (get_range_strlen): Likewise.
	* gimple-ssa-sprintf.c (get_string_length): Likewise.

Co-Authored-By: Jeff Law <law@redhat.com>

From-SVN: r267378

5 files changed, 54 insertions, 10 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index f1f9d70..cd2a294 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,6 +1,12 @@
 2018-12-23  Martin Sebor  <msebor@redhat.com>
 	    Jeff Law  <law@redhat.com>
 
+	* builtins.h (c_strlen_data): Add new fields and comments.
+	* builtins.c (unterminated_array): Change field reference from
+	"len" to "minlen" in c_strlen_data instance.
+	* gimple-fold.c (get_range_strlen): Likewise.
+	* gimple-ssa-sprintf.c (get_string_length): Likewise.
+
 	* builtins.c (unterminated_array): Rename "data" to "lendata".  Fix
 	a few comments.
 	(expand_builtin_strnlen, expand_builtin_stpcpy_1): Likewise.
diff --git a/gcc/builtins.c b/gcc/builtins.c
index 0eb3df9..b56577e 100644
--- a/gcc/builtins.c
+++ b/gcc/builtins.c
@@ -577,11 +577,11 @@ unterminated_array (tree exp, tree *size /* = NULL */, bool *exact /* = NULL */)
      structure if EXP references a unterminated array.  */
   c_strlen_data lendata = { };
   tree len = c_strlen (exp, 1, &lendata);
-  if (len == NULL_TREE && lendata.len && lendata.decl)
+  if (len == NULL_TREE && lendata.minlen && lendata.decl)
      {
        if (size)
 	{
-	  len = lendata.len;
+	  len = lendata.minlen;
 	  if (lendata.off)
 	    {
 	      /* Constant offsets are already accounted for in LENDATA.MINLEN,
@@ -720,7 +720,7 @@ c_strlen (tree src, int only_value, c_strlen_data *data, unsigned eltsize)
 	{
 	  data->decl = decl;
 	  data->off = byteoff;
-	  data->len = ssize_int (len);
+	  data->minlen = ssize_int (len);
 	  return NULL_TREE;
 	}
 
@@ -794,7 +794,7 @@ c_strlen (tree src, int only_value, c_strlen_data *data, unsigned eltsize)
     {
       data->decl = decl;
       data->off = byteoff;
-      data->len = ssize_int (len);
+      data->minlen = ssize_int (len);
       return NULL_TREE;
     }
 
diff --git a/gcc/builtins.h b/gcc/builtins.h
index cf4f9b1..472a86d 100644
--- a/gcc/builtins.h
+++ b/gcc/builtins.h
@@ -57,10 +57,48 @@ extern bool get_pointer_alignment_1 (tree, unsigned int *,
 				     unsigned HOST_WIDE_INT *);
 extern unsigned int get_pointer_alignment (tree);
 extern unsigned string_length (const void*, unsigned, unsigned);
+
 struct c_strlen_data
 {
+  /* [MINLEN, MAXBOUND, MAXLEN] is a range describing the length of
+     one or more strings of possibly unknown length.  For a single
+     string of known length the range is a constant where
+     MINLEN == MAXBOUND == MAXLEN holds.
+     For other strings, MINLEN is the length of the shortest known
+     string.  MAXBOUND is the length of a string that could be stored
+     in the largest array referenced by the expression.  MAXLEN is
+     the length of the longest sequence of non-zero bytes
+     in an object referenced by the expression.  For such strings,
+     MINLEN <= MAXBOUND <= MAXLEN holds.  For example, given:
+       struct A { char a[7], b[]; };
+       extern struct A *p;
+       n = strlen (p->a);
+     the computed range will be [0, 6, ALL_ONES].
+     However, for a conditional expression involving a string
+     of known length and an array of unknown bound such as
+       n = strlen (i ? p->b : "123");
+     the range will be [3, 3, ALL_ONES].
+     MINLEN != 0 && MAXLEN == ALL_ONES indicates that MINLEN is
+     the length of the shortest known string and implies that
+     the shortest possible string referenced by the expression may
+     actually be the empty string.  This distinction is useful for
+     diagnostics.  get_range_strlen() return value distinguishes
+     between these two cases.
+     As the tighter (and more optimistic) bound, MAXBOUND is suitable
+     for diagnostics but not for optimization.
+     As the more conservative bound, MAXLEN is intended to be used
+     for optimization.  */
+  tree minlen;
+  tree maxlen;
+  tree maxbound;
+  /* When non-null, NONSTR refers to the declaration known to store
+     an unterminated constant character array, as in:
+     const char s[] = { 'a', 'b', 'c' };
+     It is used to diagnose uses of such arrays in functions such as
+     strlen() that expect a nul-terminated string as an argument.  */
   tree decl;
-  tree len;
+  /* Non-constant offset from the beginning of a string not accounted
+     for in the length range.  Used to improve diagnostics.  */
   tree off;
 };
 
diff --git a/gcc/gimple-fold.c b/gcc/gimple-fold.c
index 0ba1514..b1dd441 100644
--- a/gcc/gimple-fold.c
+++ b/gcc/gimple-fold.c
@@ -1343,8 +1343,8 @@ get_range_strlen (tree arg, tree length[2], bitmap *visited, int type,
 	  if (!val && lendata.decl)
 	    {
 	      *nonstr = lendata.decl;
-	      *minlen = lendata.len;
-	      *maxlen = lendata.len;
+	      *minlen = lendata.minlen;
+	      *maxlen = lendata.minlen;
 	      return type == 0 ? false : true;
 	    }
 	}
diff --git a/gcc/gimple-ssa-sprintf.c b/gcc/gimple-ssa-sprintf.c
index d627830..8284c76 100644
--- a/gcc/gimple-ssa-sprintf.c
+++ b/gcc/gimple-ssa-sprintf.c
@@ -2015,12 +2015,12 @@ get_string_length (tree str, unsigned eltsize)
     }
   else if (!slen
 	   && data.decl
-	   && data.len
-	   && TREE_CODE (data.len) == INTEGER_CST)
+	   && data.minlen
+	   && TREE_CODE (data.minlen) == INTEGER_CST)
     {
       /* STR was not properly NUL terminated, but we have
 	 length information about the unterminated string.  */
-      fmtresult res (tree_to_shwi (data.len));
+      fmtresult res (tree_to_shwi (data.minlen));
       res.nonstr = data.decl;
       return res;
     }