================
@@ -6298,10 +6304,43 @@ static bool isImmediateSinkBlock(const CFGBlock *Blk) {
   // at least for now, but once we have better support for exceptions,
   // we'd need to carefully handle the case when the throw is being
   // immediately caught.
-  if (llvm::any_of(*Blk, [](const CFGElement &Elm) {
+  if (llvm::any_of(*Blk, [](const CFGElement &Elm) -> bool {
+        if (std::optional<CFGStmt> StmtElm = Elm.getAs<CFGStmt>())
+          return isa<CXXThrowExpr>(StmtElm->getStmt());
+        return false;
+      }))
+    return true;
+
+  auto HasNoReturnCall = [](const CallExpr *CE) {
+    if (!CE)
+      return false;
+
+    static thread_local llvm::SmallPtrSet<const FunctionDecl *, 32> InProgress;
----------------
negativ wrote:

> The use of static locals is a red-flag. Consider running gtests, where 
> multiple clang instances will be created in sequence, slowly filling up the 
> static cache with decl pointer that will get dangling after each test. Once 
> in a while after a long running unittest you would get a cache hit for a 
> dangling pointer and the disaster happens.

I've revisited the approach I initially took, so now we don't need any cache to 
collect visited functions as I added an implicit bool argument to the 
`analyzer_noreturn` attribute. These declarations are now considered equivalent:

```cpp
void foo() __attribute__((analyzer_noreturn));
void foo() __attribute__((analyzer_noreturn(true)));
```
...and I can add implicit `__attribute__((analyzer_noreturn(false)))` to 
functions that I've already checked. This gives us the following benefits:

- We don't need to revisit function calls repeatedly since the result of the 
previous check is saved in the analyzer_noreturn attribute
- Simplifies DFS search (nodes we're currently visiting are marked with` 
analyzer_noreturn(false)`)

https://github.com/llvm/llvm-project/pull/146355
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to