[clang-tools-extra] Add bugprone-loop-variable-copied-then-modified clang-tidy check. (PR #157213)
Julia Hansbrough via cfe-commits
cfe-commits at lists.llvm.org
Wed Sep 10 16:47:12 PDT 2025
================
@@ -0,0 +1,48 @@
+.. title:: clang-tidy - bugprone-loop-variable-copied-then-modified
+
+bugprone-loop-variable-copied-then-modified
+===========================================
+
+Detects when a loop variable is copied and then subsequently modified and
+suggests replacing with a ``const`` reference or an explicit copy.
+
+This pattern is considered bugprone because, frequently, programmers do not
+realize that they are modifying a *copy* rather than an underlying value,
+resulting in subtly erroneous code.
+
+For instance, the following code attempts to null out a value in a map, but only
+succeeds in
+
+.. code-block:: c++
+
+ for (auto target : target_map) {
+ target.value = nullptr;
+ }
+
+The programmer is likely to have intended this code instead:
+
+.. code-block:: c++
+
+ for (const auto& target : target_map) {
+ target.value = nullptr;
+ }
+
+This warning can be suppressed in one of two ways:
+ - In cases where the programmer did not intend to create a copy, they can
+ convert the loop variable to a ``const`` reference. A FixIt message will
+ provide a naive suggestion of how to achieve this, which works in most
+ cases.
+ - In cases where the intent is in fact to modify a copy, they may perform the
+ copy inside the body of the loop, and perform whatever operations they like
+ on that copy.
----------------
flowerhack wrote:
I'd argue that it is better and more readable, in most cases: based on code we've observed in Chromium, it's a surprisingly-common footgun for developers to perform a copy in the loop declaration without realizing it. Moving that copy inside the body of the loop takes behavior that is *implicit* and makes it *explicit*, difficult-to-ignore, demands-the-programmer-has-thought-through-what-they're-doing, etc. See e.g. https://chromium-review.googlesource.com/c/chromium/src/+/6409957 and https://chromium-review.googlesource.com/c/chromium/src/+/6512516 —in both cases it seems unlikely the bug would've gone unnoticed if they'd been forced to either make it a reference or make the copy explicit.
https://github.com/llvm/llvm-project/pull/157213
More information about the cfe-commits
mailing list