Bug#210848: g++-3.3: Code miscompilation with -O0 -finline (discovered in QT)

[Aurelien Jarno <aurel32@debian.org>]

tag 210848 + patch
thanks

Hello,

Here is a .dpatch file based on the patch posted on
gcc-patches@gcc.gnu.org [1]. I built gcc with this patch, and I tried to
run the testcase. I confirm that it fixes the bug.

Cheers,
Aurelien

[1] http://gcc.gnu.org/ml/gcc-patches/2003-09/msg00853.html

#! /bin/sh -e

# DP: Unsharing lhs (leading to generate an infinite loop with inline
# DP: functions).

if [ $# -eq 3 -a "$2" = '-d' ]; then
    pdir="-d $3"
elif [ $# -ne 1 ]; then
    echo >&2 "`basename $0`: script expects -patch|-unpatch as argument"
    exit 1
fi
case "$1" in
    -patch)
        patch $pdir -f --no-backup-if-mismatch -p0 < $0
        ;;
    -unpatch)
        patch $pdir -f --no-backup-if-mismatch -R -p0 < $0
        ;;
    *)
        echo >&2 "`basename $0`: script expects -patch|-unpatch as argument"
        exit 1
esac
exit 0

--- gcc/cp/typeck.c	2003-09-19 15:53:40 +0200
+++ gcc/cp/typeck.c	2003-09-19 15:53:58 +0200
@@ -5526,9 +5526,18 @@
 	{
 	  tree result_type = TREE_TYPE (newrhs);
+        tree lhs2;
+
+        /* Ensure that the calls are not shared with the lhs kept
+           at condition.  */
+        if (TREE_SIDE_EFFECTS (lhs))
+          lhs2 = break_out_calls (lhs);
+        else
+          lhs2 = lhs;
+
 	  /* We have to convert each arm to the proper type because the
 	     types may have been munged by constant folding.  */
 	  result
 	    = build (COND_EXPR, result_type, cond,
-		     build_modify_expr (lhs, modifycode,
+                   build_modify_expr (lhs2, modifycode,
 					cp_convert (result_type,
 						    TREE_OPERAND (newrhs, 1))),