[PATCH] D99262: [analyzer] Fix dead store checker false positive

[Valeriy Savchenko via Phabricator via cfe-commits]

vsavchenko added inline comments.


================
Comment at: clang/lib/StaticAnalyzer/Checkers/DeadStoresChecker.cpp:420-421
+              // We should also allow defensive initialization of structs.
+              if (const auto *ILE =
+                      dyn_cast<InitListExpr>(E->IgnoreParenCasts())) {
+                // We can use exactly the same logic here.
----------------
NoQ wrote:
> vsavchenko wrote:
> > martong wrote:
> > > What about nested InitListExpr's?
> > > ```
> > > std::array<int, 3> a1{ {1, 2, 3} };
> > > ```
> > > 
> > > ```
> > > VarDecl 0x561b200333a0 </home/egbomrt/tmp/aaa.cc:2:1, col:34> col:20 a1 'std::array<int, 3>':'std::array<int, 3>' listinit
> > > `-InitListExpr 0x561b20036d78 <col:22, col:34> 'std::array<int, 3>':'std::array<int, 3>'
> > >   `-InitListExpr 0x561b20036dc0 <col:24, col:32> 'typename _AT_Type::_Type':'int [3]'
> > >     |-IntegerLiteral 0x561b20033408 <col:25> 'int' 1
> > >     |-IntegerLiteral 0x561b20033428 <col:28> 'int' 2
> > >     `-IntegerLiteral 0x561b20033448 <col:31> 'int' 3
> > > ```
> > I'm not sure that we'll report anything on that
> We probably won't because it's a C++ object, even though it's an aggregate so we should probably warn(?) What about a plain C object like
> ```lang=c++
> int x[2][2] = { { 0, 0 }, { 0, 0 } };
> ```
> ?
The same here, we still don't warn about stores like this, but I think I can add test with nested structs


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99262/new/

https://reviews.llvm.org/D99262