[PATCH] D88088: WIP [clang] improve accuracy of ExprMutAnalyzer

[Jonas Toth via Phabricator via cfe-commits]

JonasToth added inline comments.


================
Comment at: clang/lib/Analysis/ExprMutationAnalyzer.cpp:455
+  const auto HasAnyNonConstIterator =
+      anyOf(allOf(hasMethod(allOf(hasName("begin"), unless(isConst()))),
+                  unless(hasMethod(allOf(hasName("begin"), isConst())))),
----------------
aaron.ballman wrote:
> Do we want to look for methods that end with `_?[Bb]egin` or `_?[Ee]nd` so that this would catch patterns like `foo_begin()`/`foo_end()`, `FooBegin()`/`FooEnd()`, or `Foo_Begin()`/`Foo_End()`?
This specific matcher is only applied in range-for contexts. There only the `begin(); end()` methods matter. I updated the comment above to clarify this.


================
Comment at: clang/unittests/Analysis/ExprMutationAnalyzerTest.cpp:65
+
+  std::string buffer;
   for (const auto *E = selectFirst<Expr>("expr", Results); E != nullptr;) {
----------------
aaron.ballman wrote:
> Was there a reason you hoisted this out of the `for` loop?
Jup.
```
buffer.clear()
```
The current form does proper memory-recycling (i believe at least :D)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88088/new/

https://reviews.llvm.org/D88088