[PATCH] Don't warn for signed 1-bit enum bitfields (PR c/56566)

[PATCH] Don't warn for signed 1-bit enum bitfields (PR c/56566)

[Jakub Jelinek]

Hi!

Enum with -1 and 0 values fits nicely into 1-bit signed bitfield, but
gcc was assuming that it doesn't and needs 2-bit signed bitfield for that
instead.  The only thing we really want to special case is that enum
containing just value 0 needs 1-bit bitfield instead of 0-bit bitfield.

Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?

2013-03-16  Jakub Jelinek  <jakub@redhat.com>

	PR c/56566
	* tree.c (tree_int_cst_min_precision): For integer_zerop (value)
	return 1 even for !unsignedp.

	* c-c++-common/pr56566.c: New test.

--- gcc/tree.c.jj	2013-01-24 17:04:27.000000000 +0100
+++ gcc/tree.c	2013-03-08 09:19:45.608975926 +0100
@@ -6648,8 +6648,6 @@ tree_int_cst_sgn (const_tree t)
 unsigned int
 tree_int_cst_min_precision (tree value, bool unsignedp)
 {
-  int log;
-
   /* If the value is negative, compute its negative minus 1.  The latter
      adjustment is because the absolute value of the largest negative value
      is one larger than the largest positive value.  This is equivalent to
@@ -6659,14 +6657,14 @@ tree_int_cst_min_precision (tree value,
     value = fold_build1 (BIT_NOT_EXPR, TREE_TYPE (value), value);
 
   /* Return the number of bits needed, taking into account the fact
-     that we need one more bit for a signed than unsigned type.  */
+     that we need one more bit for a signed than unsigned type.
+     If value is 0 or -1, the minimum precision is 1 no matter
+     whether unsignedp is true or false.  */
 
   if (integer_zerop (value))
-    log = 0;
+    return 1;
   else
-    log = tree_floor_log2 (value);
-
-  return log + 1 + !unsignedp;
+    return tree_floor_log2 (value) + 1 + !unsignedp;
 }
 
 /* Compare two constructor-element-type constants.  Return 1 if the lists
--- gcc/testsuite/c-c++-common/pr56566.c.jj	2013-03-08 09:31:17.265226082 +0100
+++ gcc/testsuite/c-c++-common/pr56566.c	2013-03-08 09:29:54.000000000 +0100
@@ -0,0 +1,14 @@
+/* PR c/56566 */
+/* { dg-do compile } */
+/* { dg-options "" } */
+
+struct S1 { enum E1 { N1 = -1, Z1 = 0 } e : 1; };
+struct S2 { enum E2 { N2 = -1 } e : 1; };
+struct S3 { enum E3 { Z3 = 0 } e : 1; };
+struct S4 { enum E4 { N4 = -2, Z4 = 1 } e : 2; };
+struct S5 { enum E5 { N5 = -3, Z5 = 1 } e : 3; };
+struct S6 { enum E6 { N6 = -2, Z6 = 1 } e : 1; }; // { dg-warning "too small|narrower" }
+struct S7 { enum E7 { N7 = -3, Z7 = 1 } e : 2; }; // { dg-warning "too small|narrower" }
+struct S8 { enum E8 { Z8 = 1 } e : 1; };
+struct S9 { enum E9 { Z9 = 2 } e : 2; };
+struct S0 { enum E0 { Z0 = 2 } e : 1; };	  // { dg-warning "too small|narrower" }

	Jakub

Re: [PATCH] Don't warn for signed 1-bit enum bitfields (PR c/56566)

[Jason Merrill]

Looks good to me.

Jason