[PATCH] Fix artificial overflow during GENERIC folding

[Eric Botcazou <botcazou@adacore.com>]

[-- Attachment #1: Type: text/plain, Size: 2581 bytes --]

Hi,

on the attached testcase, the Ada compiler gives a bogus warning:
storage_offset1.ads:16:52: warning: Constraint_Error will be raised at run 
time [enabled by default]

This directly comes from the GENERIC folding setting a bogus TREE_OVERFLOW on 
an INTEGER_CST during the (T)P - (T)(P + A) -> -(T) A transformation:

  /* (T)P - (T)(P + A) -> -(T) A */
  (simplify
   (minus (convert? @0)
    (convert (plus:c @@0 @1)))
   (if (INTEGRAL_TYPE_P (type)
	&& TYPE_OVERFLOW_UNDEFINED (type)
	&& element_precision (type) <= element_precision (TREE_TYPE (@1)))
    (with { tree utype = unsigned_type_for (type); }
     (convert (negate (convert:utype @1))))
    (if (element_precision (type) <= element_precision (TREE_TYPE (@1))
	 /* For integer types, if A has a smaller type
	    than T the result depends on the possible
	    overflow in P + A.
	    E.g. T=size_t, A=(unsigned)429497295, P>0.
	    However, if an overflow in P + A would cause
	    undefined behavior, we can assume that there
	    is no overflow.  */
	 || (INTEGRAL_TYPE_P (TREE_TYPE (@1))
	     && TYPE_OVERFLOW_UNDEFINED (TREE_TYPE (@1))))
     (negate (convert @1)))))
  (simplify
   (minus (convert @0)
    (convert (pointer_plus @@0 @1)))
   (if (INTEGRAL_TYPE_P (type)
	&& TYPE_OVERFLOW_UNDEFINED (type)
	&& element_precision (type) <= element_precision (TREE_TYPE (@1)))
    (with { tree utype = unsigned_type_for (type); }
     (convert (negate (convert:utype @1))))
    (if (element_precision (type) <= element_precision (TREE_TYPE (@1))
	 /* For pointer types, if the conversion of A to the
	    final type requires a sign- or zero-extension,
	    then we have to punt - it is not defined which
	    one is correct.  */
	 || (POINTER_TYPE_P (TREE_TYPE (@0))
	     && TREE_CODE (@1) == INTEGER_CST
	     && tree_int_cst_sign_bit (@1) == 0))
     (negate (convert @1)))))

Ironically enough, this occurs because of the intermediate conversion to an 
unsigned type which is supposed to hide overflows, but is counter-productive 
for constants because TREE_OVERFLOW is always set for them, so it ends up 
setting a bogus TREE_OVERFLOW when converting back to the original type.

The fix simply redirects INTEGER_CSTs to the other, direct path without the 
intermediate conversion to the unsigned type.

Tested on x86-64/Linux, OK for the mainline?


2023-05-24  Eric Botcazou <ebotcazou@adacore.com>

	* match.pd ((T)P - (T)(P + A) -> -(T) A): Avoid artificial overflow
	on constants.


2023-05-24  Eric Botcazou <ebotcazou@adacore.com>

	* gnat.dg/specs/storage_offset1.ads: New test.

-- 
Eric Botcazou

[-- Attachment #2: p.diff --]
[-- Type: text/x-patch, Size: 847 bytes --]

diff --git a/gcc/match.pd b/gcc/match.pd
index 1fe0559acfb..b9d04dd423b 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -3194,6 +3194,7 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
     (convert (plus:c @@0 @1)))
    (if (INTEGRAL_TYPE_P (type)
 	&& TYPE_OVERFLOW_UNDEFINED (type)
+	&& TREE_CODE (@1) != INTEGER_CST
 	&& element_precision (type) <= element_precision (TREE_TYPE (@1)))
     (with { tree utype = unsigned_type_for (type); }
      (convert (negate (convert:utype @1))))
@@ -3213,6 +3214,7 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
     (convert (pointer_plus @@0 @1)))
    (if (INTEGRAL_TYPE_P (type)
 	&& TYPE_OVERFLOW_UNDEFINED (type)
+	&& TREE_CODE (@1) != INTEGER_CST
 	&& element_precision (type) <= element_precision (TREE_TYPE (@1)))
     (with { tree utype = unsigned_type_for (type); }
      (convert (negate (convert:utype @1))))

[-- Attachment #3: storage_offset1.ads --]
[-- Type: text/x-adasrc, Size: 324 bytes --]

-- { dg-do compile }

with System.Storage_Elements; use System.Storage_Elements;
with System;

package Storage_Offset1 is

  type Rec is record
    I1, I2 : Integer;
  end record;

  type Ptr is access all Rec;

  R : Ptr := new Rec;

  Offset : constant Storage_Offset := R.I1'Address - R.I2'Address;

end Storage_Offset1;