On 2/15/22 05:07, Jakub Jelinek wrote:
Hi!
The comment in shorten_compare says:
/* If either arg is decimal float and the other is float, fail. */
but the callers of shorten_compare don't expect anything like failure
as a possibility from the function, callers require that the function
promotes the operands to the same type, whether the original selected
*restype_ptr one or some shortened.
So, if we choose not to shorten, we should still promote to the original
*restype_ptr.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
OK.
2022-02-15 Jakub Jelinek <ja...@redhat.com>
PR c/104510
* c-common.cc (shorten_compare): Convert original arguments to
the original *restype_ptr when mixing binary and decimal float.
* gcc.dg/dfp/pr104510.c: New test.
--- gcc/c-family/c-common.cc.jj 2022-02-04 14:36:53.998619364 +0100
+++ gcc/c-family/c-common.cc 2022-02-14 19:07:14.305068950 +0100
@@ -3174,7 +3174,11 @@ shorten_compare (location_t loc, tree *o
else if (real1 && real2
&& (DECIMAL_FLOAT_MODE_P (TYPE_MODE (TREE_TYPE (primop0)))
|| DECIMAL_FLOAT_MODE_P (TYPE_MODE (TREE_TYPE (primop1)))))
- return NULL_TREE;
+ {
+ type = *restype_ptr;
+ primop0 = op0;
+ primop1 = op1;
+ }
else if (real1 && real2
&& (TYPE_PRECISION (TREE_TYPE (primop0))
--- gcc/testsuite/gcc.dg/dfp/pr104510.c.jj 2022-02-14 19:11:05.610860035
+0100
+++ gcc/testsuite/gcc.dg/dfp/pr104510.c 2022-02-14 19:10:42.819176224 +0100
@@ -0,0 +1,12 @@
+/* PR c/104510 */
+/* { dg-do compile } */
+/* { dg-options "" } */
+
+float f;
+_Decimal64 d;
+
+int
+foo (void)
+{
+ return d > (_Decimal32) (_Decimal64) f;
+}
Jakub
