Hi!
The following testcase ICEs on i686-linux, because c_finish_return wraps
c_fully_folded retval back into EXCESS_PRECISION_EXPR, but when the function
return type is void, we don't call convert_for_assignment on it that would
then be fully folded again, but just put the retval into RETURN_EXPR's
operand, so nothing removes it anymore and during gimplification we
ICE as EXCESS_PRECISION_EXPR is not handled.
This patch fixes it by not adding that EXCESS_PRECISION_EXPR in functions
returning void, the return value is ignored and all we need is evaluate any
side-effects of the expression.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2021-02-18 Jakub Jelinek <ja...@redhat.com>
PR c/99136
* c-typeck.c (c_finish_return): Don't wrap retval into
EXCESS_PRECISION_EXPR in functions that return void.
* gcc.dg/pr99136.c: New test.
--- gcc/c/c-typeck.c.jj 2021-02-04 23:41:24.200308485 +0100
+++ gcc/c/c-typeck.c 2021-02-17 19:43:30.348054337 +0100
@@ -10740,7 +10740,9 @@ c_finish_return (location_t loc, tree re
retval = TREE_OPERAND (retval, 0);
}
retval = c_fully_fold (retval, false, NULL);
- if (semantic_type)
+ if (semantic_type
+ && valtype != NULL_TREE
+ && TREE_CODE (valtype) != VOID_TYPE)
retval = build1 (EXCESS_PRECISION_EXPR, semantic_type, retval);
}
--- gcc/testsuite/gcc.dg/pr99136.c.jj 2021-02-17 19:42:17.253844514 +0100
+++ gcc/testsuite/gcc.dg/pr99136.c 2021-02-17 19:42:09.684926337 +0100
@@ -0,0 +1,9 @@
+/* PR c/99136 */
+/* { dg-do compile } */
+/* { dg-options "-w -fexcess-precision=standard" } */
+
+void
+foo (double x)
+{
+ return 1.0 / x;
+}
Jakub
