On 9/16/24 7:08 PM, Marek Polacek wrote:
On Tue, Sep 10, 2024 at 10:37:31AM -0400, Jason Merrill wrote:
On 8/29/24 12:23 PM, Marek Polacek wrote:
Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk/14?
-- >8 --
Pre r14-4793, we'd call warn_tautological_cmp -> operand_equal_p
with operands wrapped in NON_DEPENDENT_EXPR, which works, since
o_e_p bails for codes it doesn't know. But now we pass operands
not encapsulated in NON_DEPENDENT_EXPR, and crash, because the
template tree for &a[x] has null DECL_FIELD_OFFSET.
Why are we trying to compare DECL_FIELD_OFFSET in C++, anyway? I'd think we
should limit that to C and in C++ rely on FIELD_DECL identity.
This goes back to r12-7797. I suppose it's trying to handle C++ because we
can't use c_dialect_cxx () in fold-const. I know there's lang_GNU_CXX but
I suggest we do the following.
I don't think I meant this warning to be called in a template,
so let's avoid the problem this easy way.
Hmm, it seems potentially useful in a template, especially since we suppress
it in instantiations.
Ack.
How about this?
Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
-- >8 --
Pre r14-4793, we'd call warn_tautological_cmp -> operand_equal_p
with operands wrapped in NON_DEPENDENT_EXPR, which works, since
o_e_p bails for codes it doesn't know. But now we pass operands
not encapsulated in NON_DEPENDENT_EXPR, and crash, because the
template tree for &a[x] has null DECL_FIELD_OFFSET.
This patch extends r12-7797 to cover the case when DECL_FIELD_OFFSET
is null.
PR c++/116534
gcc/ChangeLog:
* fold-const.cc (operand_compare::operand_equal_p): Check
DECL_FIELD_OFFSET.
gcc/testsuite/ChangeLog:
* g++.dg/warn/Wtautological-compare4.C: New test.
---
gcc/fold-const.cc | 16 ++++++++------
.../g++.dg/warn/Wtautological-compare4.C | 21 +++++++++++++++++++
2 files changed, 31 insertions(+), 6 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/warn/Wtautological-compare4.C
diff --git a/gcc/fold-const.cc b/gcc/fold-const.cc
index 70db16759d0..3da09621584 100644
--- a/gcc/fold-const.cc
+++ b/gcc/fold-const.cc
@@ -3601,12 +3601,16 @@ operand_compare::operand_equal_p (const_tree arg0,
const_tree arg1,
/* Non-FIELD_DECL operands can appear in C++ templates. */
if (TREE_CODE (field0) != FIELD_DECL
- || TREE_CODE (field1) != FIELD_DECL
- || !operand_equal_p (DECL_FIELD_OFFSET (field0),
- DECL_FIELD_OFFSET (field1), flags)
- || !operand_equal_p (DECL_FIELD_BIT_OFFSET (field0),
- DECL_FIELD_BIT_OFFSET (field1),
- flags))
+ || TREE_CODE (field1) != FIELD_DECL)
+ return false;
+
+ if (DECL_FIELD_OFFSET (field0)
+ && DECL_FIELD_OFFSET (field1)
+ && (!operand_equal_p (DECL_FIELD_OFFSET (field0),
+ DECL_FIELD_OFFSET (field1), flags)
+ || !operand_equal_p (DECL_FIELD_BIT_OFFSET (field0),
+ DECL_FIELD_BIT_OFFSET (field1),
+ flags)))
If at least one has null FIELD_OFFSET, I'd think we still want to
compare them for being ==, this seems to assume that a field with null
offset is equal to any other field.
return false;
}
else
diff --git a/gcc/testsuite/g++.dg/warn/Wtautological-compare4.C
b/gcc/testsuite/g++.dg/warn/Wtautological-compare4.C
new file mode 100644
index 00000000000..96308f49a42
--- /dev/null
+++ b/gcc/testsuite/g++.dg/warn/Wtautological-compare4.C
@@ -0,0 +1,21 @@
+// PR c++/116534
+// { dg-do compile }
+// { dg-options "-Wall" }
+
+template <class A>
+struct Test {
+ bool foo(unsigned x, unsigned y) {
+ bool test = &a[x] == &b[y];
+ return test;
+ }
+ unsigned *a;
+ unsigned *b;
+};
+
+void
+g ()
+{
+ Test<int> t;
+ t.foo (0u, 1u);
+ t.foo (0u, 0u);
+}
base-commit: f5448384a2134f32c8733b401440da11bfe69252
