On 5/22/26 5:00 PM, Johannes Natter wrote:
Thanks for the review!
PR created:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125422
> data_addr isn't equivalent to alloc_node if cookie_size is nonzero;
> this seems to do the wrong thing in that case.
When cookie_size is nonzero, the block at lines 3618-3652 always
assigns a non-NULL value to cookie_expr. Line 3951 then wraps rval:
if (cookie_expr)
rval = build2 (COMPOUND_EXPR, TREE_TYPE (rval), cookie_expr, rval);
This makes rval != data_addr (pointer comparison), so the condition
`if (rval != data_addr)' evaluates to true and the null-check COND_EXPR
is still built. The cookie case should be handled correctly.
Ah, good point. But then we don't need any of the if (check_new) block
in this case, so we could clear check_new if we didn't add any
COMPOUND_EXPR to rval, or check rval != data_addr along with check_new?
If you prefer to make the cookie handling more explicit, I can change
the condition to:
if (rval != data_addr || cookie_size)
cookie_size != 0 always implies rval != data_addr, but it might make
the cookie case easier to see.
> Do you intend to contribute under FSF copyright assignment or the DCO?
I intend to contribute under the DCO. Here is the updated patch with
Signed-off-by added:
Great. Please also add a "PR c++/125422" line at the top of the
ChangeLog entries. 'git gcc-verify' would complain about that.
From cfdc00cb416167cbf905f6ca1e3919bf567e663f Mon Sep 17 00:00:00 2001
From: Joe Natter <[email protected]>
Date: Thu, 21 May 2026 20:20:16 +0200
Subject: [PATCH] c++: Avoid spurious -Wduplicated-branches for new
(nothrow) T[n] [PR125422]
When allocating an array of a trivial type with new (nothrow),
build_new_1() may emit a compiler-generated COND_EXPR of the form
(alloc_node != nullptr) ? rval : alloc_node
For trivial arrays without constructors, destructors, or array cookies,
rval remains equal to data_addr and is therefore equivalent to
alloc_node. This causes both branches of the generated COND_EXPR to be
identical, triggering -Wduplicated-branches on user code.
Avoid building the null-check COND_EXPR when rval == data_addr.
gcc/cp/ChangeLog:
* init.cc (build_new_1): Avoid building a null-check
COND_EXPR when rval == data_addr.
gcc/testsuite/ChangeLog:
* g++.dg/warn/Wduplicated-branches10.C: New test.
Signed-off-by: Joe Natter <[email protected]>
---
gcc/cp/init.cc | 17 +++++++++++------
.../g++.dg/warn/Wduplicated-branches10.C | 15 +++++++++++++++
2 files changed, 26 insertions(+), 6 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/warn/Wduplicated-branches10.C
diff --git a/gcc/cp/init.cc b/gcc/cp/init.cc
index d0929c35bfc..2b089524409 100644
--- a/gcc/cp/init.cc
+++ b/gcc/cp/init.cc
@@ -3961,12 +3961,17 @@ build_new_1 (vec<tree, va_gc> **placement, tree
type, tree nelts,
{
if (check_new)
{
- tree ifexp = cp_build_binary_op (input_location,
- NE_EXPR, alloc_node,
- nullptr_node,
- complain);
- rval = build_conditional_expr (input_location, ifexp, rval,
- alloc_node, complain);
+ /* Avoid generating a redundant null-check COND_EXPR
+ when rval == data_addr. */
+ if (rval != data_addr)
+ {
+ tree ifexp = cp_build_binary_op (input_location,
+ NE_EXPR, alloc_node,
+ nullptr_node,
+ complain);
+ rval = build_conditional_expr (input_location, ifexp, rval,
+ alloc_node, complain);
+ }
/* If there's no offset between data_addr and alloc_node, append it
to help -Wmismatched-new-delete at -O0. */
if (!cookie_size)
diff --git a/gcc/testsuite/g++.dg/warn/Wduplicated-branches10.C b/gcc/
testsuite/g++.dg/warn/Wduplicated-branches10.C
new file mode 100644
index 00000000000..0ebccd7d6fe
--- /dev/null
+++ b/gcc/testsuite/g++.dg/warn/Wduplicated-branches10.C
@@ -0,0 +1,15 @@
+// PR c++/125422
+// { dg-do compile }
+// { dg-options "-Wduplicated-branches" }
+
+// new (nothrow) T[n] with a runtime variable must not trigger
+// -Wduplicated-branches (false positive from compiler-generated
COND_EXPR).
+
+#include <new>
+
+void
+test_nothrow (int sz)
+{
+ char *p = new (std::nothrow) char[sz];
+ delete[] p;
+}
