[clang] [analyzer] Fix uninitialized base class with initializer list when ctor is not declared in the base class (#70464) (PR #70792)
Ella Ma via cfe-commits
cfe-commits at lists.llvm.org
Tue Oct 31 09:28:30 PDT 2023
================
@@ -1222,6 +1222,15 @@ void ExprEngine::ProcessInitializer(const CFGInitializer CFGInit,
PostInitializer PP(BMI, FieldLoc.getAsRegion(), stackFrame);
evalBind(Tmp, Init, Pred, FieldLoc, InitVal, /*isInit=*/true, &PP);
}
+ } else if (BMI->isBaseInitializer() && isa<InitListExpr>(Init)) {
+ // When the base class is initialized with an initialization list, there
+ // will not be a CXXConstructExpr to initialize the base region. Hence, we
+ // need to make the bind for it.
+ StoreManager &StoreMgr = State->getStateManager().getStoreManager();
+ SVal BaseLoc = StoreMgr.evalDerivedToBase(
+ thisVal, QualType(BMI->getBaseClass(), 0), BMI->isBaseVirtual());
+ SVal InitVal = State->getSVal(Init, stackFrame);
+ evalBind(Tmp, Init, Pred, BaseLoc, InitVal, true);
----------------
Snape3058 wrote:
`evalBind` will add a `PostStore` point if not given.
IMO, the initializer list here is just to make the assignment, just like `T test = {1};`, which has an AST structure of
```
DeclStmt 0x55977c889f98 <col:14, col:20>
`-VarDecl 0x55977c889e80 <col:14, col:19> col:16 t 'T':'T' listinit
`-InitListExpr 0x55977c889f50 <col:17, col:19> 'T':'T'
`-IntegerLiteral 0x55977c889ee8 <col:18> 'int' 1
```
The `DeclStmt` here will use the `PostStore` point and wrap it with a `PostStmt` of `DeclStmt`.
So, a `PostStore` point here makes sense to me.
Besides, in the original `else` branch below, node `Pred` has a PostStmt point of CXXConstructExpr, which is also different from the `PostInitializer`.
And the `PostInitializer` point is used to notice the diagnostic system that a field assignment is here.
However, this can be done with the `PostStore` assigned with `evalBind`.
So, I think adding a `PostInitializer` here for the `evalBind` may not be necessary.
I am not very familiar with program points. If I get it incorrectly, I will correct this in the next submission.
https://github.com/llvm/llvm-project/pull/70792
More information about the cfe-commits
mailing list