1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
|
From 8ebd0bf1cf02414584d15d7244b07fa97d65ba02 Mon Sep 17 00:00:00 2001
From: Changming Sun <chasun@microsoft.com>
Date: Mon, 29 Sep 2025 21:19:29 -0700
Subject: [PATCH] Fix(build): Correct maybe-uninitialized and
range-loop-construct warnings (#26201)
Forwarded: not-needed
Origin: https://github.com/microsoft/onnxruntime/pull/26201
This pull request addresses two build warnings that were treated as
errors:
1. **`maybe-uninitialized` warning in `nchwc_transformer.cc`**:
A `maybe-uninitialized` warning was triggered when accessing
`nchwc_inputs[0]` without checking if `nchwc_inputs` was empty. This
could lead to a crash if a binary node had no inputs. The fix adds a
check to ensure the vector is not empty before accessing the first
element.
2. **`range-loop-construct` warnings in `transformer_memcpy.cc`**:
The compiler warned about creating copies of objects in range-based for
loops. This was resolved by using references (`const auto&`) instead of
copies (`const auto`) for the loop variables, which is more efficient.
This fix was provided by an AI bot.
---
onnxruntime/core/optimizer/nchwc_transformer.cc | 4 ++++
onnxruntime/core/optimizer/transformer_memcpy.cc | 4 ++--
2 files changed, 6 insertions(+), 2 deletions(-)
--- a/onnxruntime/core/optimizer/nchwc_transformer.cc
+++ b/onnxruntime/core/optimizer/nchwc_transformer.cc
@@ -609,6 +609,10 @@
nchwc_inputs.push_back(nchwc_input);
}
+ if (nchwc_inputs.empty()) {
+ return;
+ }
+
auto* nchwc_input_0 = nchwc_inputs[0];
const int64_t channels = nchwc_inputs[0]->channels_;
--- a/onnxruntime/core/optimizer/transformer_memcpy.cc
+++ b/onnxruntime/core/optimizer/transformer_memcpy.cc
@@ -16,7 +16,7 @@
static ProviderTypeToProviderMap GetProvidersByType(
const InlinedVector<gsl::not_null<const IExecutionProvider*>>& providers) {
ProviderTypeToProviderMap providers_by_type{};
- for (const auto provider : providers) {
+ for (const auto& provider : providers) {
providers_by_type.emplace(provider->Type(), provider);
}
return providers_by_type;
@@ -100,7 +100,7 @@
// and mainly provides the subgraph recursion functionality
Status MemcpyTransformer::ApplyImpl(Graph& graph, bool& modified, int graph_level,
const logging::Logger& logger) const {
- for (const auto provider : providers_) {
+ for (const auto& provider : providers_) {
const auto& provider_type = provider->Type();
if (!utils::ProviderIsCpuBased(*provider)) {
TransformerMemcpyImpl copy_impl(graph, *provider, providers_by_type_);
|
