On 1/18/26 6:04 PM, Johannes Altmanninger wrote:
On this expression:
(true ? "a" : "b")[index()]
"g++ -Wunused-value" incorrectly produces
warning: left operand of comma operator has no effect [-Wunused-value]
From the -fdump-tree-original output:
if ((void) SAVE_EXPR <index ()>, 1)
{
(void) "a"[SAVE_EXPR <index ()>];
}
else
{
(void) "b"[SAVE_EXPR <index ()>];
}
Observe that we evaluate index() (and save it) before evaluating the
ternary expression. Since "(void) SAVE_EXPR <index ()>" is ostensibly
side-effect free, we get this warning. Since SAVE_EXPR is not useless,
this is a false positive. Also the comma operator compiler-generated,
so warning about it is wrong.
Suppress this warning for this implicit expression. Test that the
warning is gone for "$ternary[index()]" but we still warn on cases like
"$ternary[(1, 0)]".
Thanks for the patch!
Alternatively, we could prevent all such warnings by changing the
definition of tem to "tree tem = cp_build_compound_expr (op0, idx,
tf_none)". That's probably the better solution but we'd need to check
if that would introduce false negatives.
Alternatively, we could set "side_effects_flag" for SAVE_EXPR, which
would also prevent this warning, but that feels wrong.
Other places in the compiler use warning_sentinel to temporarily disable
a particular warning; instead of suppress_warning we could here use
warning_sentinel w(warn_unused_value);
gcc/cp/ChangeLog:
* typeck.cc (cp_build_array_ref): suppress unused-value
We capitalize the beginning of ChangeLog sentences (i.e. "Suppress").
warning for implicit comma expression.
gcc/testsuite/ChangeLog:
* g++.dg/cpp0x/Wunused-value2.C: New test.
I'd expect this test to go in warn/ rather than cpp0x/.
Signed-off-by: Johannes Altmanninger <[email protected]>
---
gcc/cp/typeck.cc | 1 +
gcc/testsuite/g++.dg/cpp0x/Wunused-value2.C | 18 ++++++++++++++++++
2 files changed, 19 insertions(+)
create mode 100644 gcc/testsuite/g++.dg/cpp0x/Wunused-value2.C
This is my first time hacking gcc, guidance is appreciated; I haven't
run the full test suite yet (which probably takes a long time when
built in debug mode, using
../gcc/configure --with-build-config=bootstrap-debug-lean --disable-bootstrap
CFLAGS='-g -O0' CXXFLAGS='-g -O0').
It does take a while, but is still expected before submitting a patch.
Though I suppose this patch is small enough that just running Wunused*
should be sufficient.
Instead, I've been using this command validate my change:
make -j8 && timeout 5 make check-c++
RUNTESTFLAGS=dg.exp=g++.dg/cpp0x/Wunused-value2.C | grep ^FAIL
without the "timeout" the command would go on to run other tests, like the
"libgomp" ones etc.
You could make check-gcc-c++ instead to avoid that.
Probably I could use "runtest" directly.
I prefer to use make for parallel testing, but runtest is fine for a
single test.
diff --git a/gcc/cp/typeck.cc b/gcc/cp/typeck.cc
index 980f2d115a6..46bfa2fbcb8 100644
--- a/gcc/cp/typeck.cc
+++ b/gcc/cp/typeck.cc
@@ -4244,6 +4244,7 @@ cp_build_array_ref (location_t loc, tree array, tree idx,
{
idx = save_expr (idx);
op0 = save_expr (op0);
+ suppress_warning (op0, OPT_Wunused_value);
tree tem = build_compound_expr (loc, op0, idx);
op0 = build_compound_expr (loc, tem, op0);
}
diff --git a/gcc/testsuite/g++.dg/cpp0x/Wunused-value2.C
b/gcc/testsuite/g++.dg/cpp0x/Wunused-value2.C
new file mode 100644
index 00000000000..8125aa518f0
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp0x/Wunused-value2.C
@@ -0,0 +1,18 @@
+// { dg-do compile }
+// { dg-options "-Wall" }
+
+static int index() { return 0; }
+
+volatile int global;
+static int index_with_side_effect() {
+ global += 1;
+ return 0;
+}
+
+int main(int argc, char **argv)
+{
+ const bool cond = argc == 10;
+ (void)(cond ? "" : "")[index()];
+ (void)(cond ? "" : "")[index_with_side_effect()];
+ (void)(cond ? "" : "")[(1, 0)]; // { dg-warning "left operand of comma
operator has no effect" }
+}
